Systems, methods, and computer program products for detecting early-stage cancer in static medical imaging

Abstract

A system for detecting a lung cancer nodule, including: a memory module operable to store computer program code and information of a group of candidate discernible objects (CDOs) within at least one image of a lung, the group including discernible objects having associated tissue-signal response patterns corresponding to cancerous nodules. The memory, computer program code, with at least one processor are configured to cause the system to: access the information of the group of CDOs; determine for each of multiple CDOs whether that CDO satisfies an isolation requirement with respect to the rest of the CDOs, based at least on geometric remoteness from the other CDOs; based on determining that a first CDO meets the isolation criteria reporting that first CDO as a suspected nodule; and based on determining that a second CDO does not meet the isolation criteria refraining from reporting that second CDO as a suspected nodule.

Description

SYSTEMS, METHODS, AND COMPUTER PROGRAM PRODUCTS FOR DETECTING EARLY-STAGE CANCER IN STATIC MEDICAL IMAGINGRELATED APPLICATIONS

[0001] This application claims priority from U.S. provisional patent application serial number 63 / 606,758 filing date December 6, 2023, entitled "Systems, Methods, and Computer Program Products for Detecting Early-Stage Cancer in Static Medical Imaging".FIELD

[0002] The invention is related to systems, methods, and computer program products for detecting lung cancer by processing static imaging data (e.g., computed tomography (CT) scans, Magnetic Resonance Imaging (MRI) scans), and especially for detecting of early-stage lung cancer by processing of static imaging data.BACKGROUND

[0003] Different approaches were discussed in the art for using computer products (whether based on artificial intelligence, Al, or not) for analysis of static imaging data for detection of cancer, all having their shortcomings. Many such publications are known to any person who is of skill in the art. Some publications of the prior art include:

[0004] "Al-derived Computer-Aided Detection (CAD) Software for Detecting and Measuring Lung Nodules in CT Scan Images", published by the National Institute for Health (NICE) on July 5, attempts to provide evidence-based recommendations on Al- derived computer-aided detection (CAD) software for detecting and measuring lung nodules in CT scan images.

[0005] "Next Gen Radiology Al, The Journey from an Algorithm to a Clinical Solution", published by the Al company Aidoc, on 2019, discusses a proposed Medical ImagingAl ecosystem of three tiers, including an Algorithmic, Product and Solution layer, and additional material. This whitepaper attempts to review the status of Al in the field as of 2019 and to explore what it will require to bring the concept of a complete three tiers solution to fruition.

[0006] "Detection of Lung Cancer Through Low-dose CT Screening (NELSON): a Prespecified Analysis of Screening Test Performance and Interval Cancers" published on October 1, 2014 by Horeweg, N., et al, at Lancet Oncology aims to assess screening test performance, and the epidemiological, radiological, and clinical characteristics of interval cancers in NELSON trial participants assigned to the screening group.

[0007] "Validation Of a Deep Learning Computer Aided System for CT Based Lung Nodule Detection, Classification, and Growth Rate Estimation in a Routine Clinical Population", by Murchison, J.T. et al., published on May 5, 2022 by PLOS ONE, attempts to evaluate a commercially available computer assisted diagnosis system (CAD). The deep learning algorithm of the CAD was trained with a lung cancer screening cohort and developed for detection, classification, quantification, and growth of actionable pulmonary nodules on chest CT scans. Here, we evaluated the CAD in a retrospective cohort of a routine clinical population.

[0008] U.S. patent application serial number 14 / 896,386, by the inventor of the present patent application, Iliya Kusner, filed May 25, 2014 and entitled "Apparatus and Method for Automated Detection of Lung Cancer" discloses, inter alia, systems and methods of computer aided detection of cancerous nodules of lung tissue.

[0009] In light of the cited prior art, there remains a need for novel systems, methods, and computer program products for detecting lung cancer in static medical imaging. There is a further need for novel systems, methods, and computer program products for screening for lung cancer using static medical imaging.SUMMARY OF THE INVENTION

[0010] The present invention is related to systems, methods, and computer program products for detecting lung cancer— including early-stage lung cancer— by processing of static imaging data (e.g., CT scans, MRI scans). Detecting lung cancer at any stageof the disease is important for both health purposes (including accurate diagnosis, efficient treatment planning, minimization of invasive procedures, and enhanced patient comfort), as well as for financial considerations, and scientific uses. Early detection of lung cancer is especially important in saving lives by enabling timely intervention, mitigating disease progression, and improving long-term outcomes. While the invention is not limited to detecting only early-stage cancer, it enables detection of very small cancerous nodules in certain scenarios (e.g., even as little as l-3mm), which correspond to early stages of lung cancer.

[0011] Referring to detection of cancerous nodules at CT scans, at an early stage of lung cancer, cancerous nodules are very small (1-3 mm) and in most cases it looks like an unobtrusive spot on a CT-image. Such nodules can be detected only using CT- scanners with a small distance between slices. Moreover, such a nodule would only leave a detectable trace on very few slices (depending on the scanning resolution), e.g., as low as only one or two slices. Therefore, detection of such small spots requires high sensitivity of the detection and recognition system. However, high sensitivity raises the likelihood of false detection and false recognition (False Positives). Balancing the trade-off between the need for high sensitivity and between the danger of false positives is one of the most important and difficult tasks in systems for the automatic recognition of early-stage lung cancer by analyzing CT images. The systems, methods, and computer program products described below employ various novel techniques to improve the likelihood of detection while maintaining the false positive rate at an acceptably low rate. A similar discussion is also relevant to MRI scans, mutatis mutandis.

[0012] The output of CT scan is commonly provided as a series of images (a "stack"), each corresponding to one or many parallel slices or cross-sections of part of the body of the imaged patient. The color or brightness of the structures imaged in each of these image is often determined by their attenuation of the X-ray beam that the scanner emits (often measured in Hounsfield units (HU) which are a standard scale for quantifying radiodensity). Different tissues in the body have different densities and thus different attenuation values, resulting in varying shades of white to black on the CT scan. A common grayscale representation used in CT imaging is that air is displayedas black (-1000 HU), fat is slightly less black (-50 to -100 HU), water / soft tissues / blood are displayed as gray (0 to +40 HU), while bone is white (+700 to +3000 HU). However, blood vessels are not always displayed as white, as the coloration can change based on a variety of factors, such as use of contrast dye is used, which is very common for angiography and other scans that aim to highlight blood vessels, the vessels will appear white on the scan. This is because the contrast dye is highly radiodense and so attenuates the X-ray beam more than the surrounding tissues. Without contrast, blood vessels filled with blood might appear as a grey shade, similar to other soft tissues.

[0013] For example, referring to the example of CT imaging of lungs in order to detect lung cancer, a common gray scale representation of CT scans demonstrates a plurality of bright spots in the lung area against a dark background. The dark background in large parts of the lungs is the air within the lungs of the patient, while the bright spots may be either blood vessels or objects which are suspected as nodules. However, other bright spots may result from various other types of tissues, from synthetic figments resulting from the computations in the CT generation process, and so on. One challenge faced by the systems, methods, and computer program product discussed below is which of these suspicious white spots or zones should be reported as suspect nodule. This gets even more challenging as the size of suspicious objects in the CT image gets smaller. As discussed below in greater detail, the methods, systems, and computer product program products may be implemented in various scenarios, such as the specific case of detecting early-stage lung cancer when the light spots of cancer nodules are still very small. Different techniques of differentiating between suspect cancerous nodules and other types of objects in the image are implemented, e.g., taking into account different types of dynamics of movement of such objects between neighboring, taking account biological differences in different parts of the inspected organ (e.g., lungs), and so on.

[0014] In comparison to the CT scan imaging described above, the MRI (Magnetic Resonance Imaging) technique employs a magnetic field and radio waves to generate a series of images (a "stack"), showing various parallel slices or cross-sections of the body part under examination. Unlike the attenuation of X-ray beams in CT, the MRIimages derive their color or brightness from the tissues' resonance with the magnetic field, measured in Tesla (T). The grayscale representation in MRI is influenced by factors such as water and hydrogen content, leading to a different scale of shading compared to CT. In MRI, for instance, air is often displayed as black, bone typically appears dark, and the contrast between soft tissues is determined by varying intensity based on Tl or T2 weighting. In MRI imaging, Tl and T2 weighting are fundamental techniques used to manipulate the image contrast by selectively highlighting certain tissue characteristics. Tl weighting, for example, emphasizes differences in the longitudinal relaxation time (Tl) of tissues, leading to brighter signals from tissues with shorter Tl values, such as fat. T2 weighting accentuates variations in the transverse relaxation time (T2) of tissues, resulting in brighter signals from tissues with longer T2 values, like fluids and some lesions. Contrast dyes in MRI are responsive to the magnetic field, creating a differential signal that can highlight blood vessels in a manner similar to CT. While the underlying principles and representations differ between CT and MRI, the methods, systems, and computer program products may be implemented similarly, with tailored adjustments to handle the specific properties and artifacts of MRI.

[0015] According to an aspect of the invention, there is disclosed a method for detecting a lung cancer nodule by processing static medical imaging, such as CT data or MRI data. The method starts with the step of obtaining a group of candidate discernible objects within a static-imaging image stack that includes at least one image (e.g., CT slice, MRI slice) of an organ of a patient (e.g., lung). The group of candidate discernible objects includes a plurality of discernible object which qualify some preliminary criteria (such as size, brightness, contrast with background, and so on) having associated attenuation levels corresponding to cancerous nodules.

[0016] The term "static imaging data" in the context of the present disclosure pertains to medical imaging data acquired using static imaging, which provides an imaging of one or more parts of the body at a given time. In comparison, dynamic imaging data (like ultrasound imaging) provides information pertaining to progression of processes in the body over a span of time (e.g., spanning full seconds or minutes). Static imaging data, in comparison, is more like a snapshot of tissues and organs.Within the context of the present disclosure, the term "static imaging data" pertains at least to static imaging data collected by any one or more of the following type of imaging machines: a. Computed tomography (CT), including its many variations. b. Positron Emission Tomography-Computed Tomography (PET-CT). c. Magnetic Resonance Imaging (MRI), including its many variations. d. X-ray radiography machines (X-ray). e. Angiography. f. Any other currently existing statical imaging technologies, as well as any suitable medical imaging technologies that will be developed in the future.

[0017] The term "discernible object" pertains to a region of an image (or, optionally, of a region of several adjacent images which represent continuous parts of the body of the patient) which is discernible from its surrounding regions, e.g., by brightness or by other factors (e.g., brightness gradients, brightness derivatives, texture, homogeneity, and so on). While not necessarily so, a discernible object may be characterized in that some properties of it (e.g., those discussed above) are constant or approximately constant.

[0018] The term "candidate discernible object" in the context of the present disclosure pertains to such a region of an image demonstrating properties which are characteristic of how a cancerous nodule (e.g., a specific kind of lung cancer nodule which is sought in a specific implementation of the invention) is represented in the respective image— CT, MRI, or other static medical imaging image. Such properties may include, for example, HU level in the case of CT, magnetic field resonance in the case of MRI, size, shape, or any other characteristic of a cancerous nodule sought in a specific implementation. For simplicity, candidate discernible objects are also simply referred to as "candidates" in the following disclosure. Any time the term "candidate" is used, it pertains to a candidate discernible object of one or more images out of the static image data.

[0019] The term "attenuation levels" in the context of the present disclosure pertains to the degree in which different tissues (corresponding to specific CT image pixels or areas) attenuate the radiation used for the computed tomography. Such radiation depends on the specific technology of CT used. For example, many types of CT (e.g., Standard CT, Cone Beam CT (CBCT), Dual-Energy CT (DECT), and Spectral CT) utilize radiation within the X-ray part of the spectrum (e.g., wavelength range of 1- lOnm), while other types of CT (e.g., positron emission tomography CT (PET CT), singlephoton emission computed tomography (SPECT CT)) utilize radiation within the Gamma part of the spectrum (e.g., wavelength range of l-10pm). As a side note, it is noted that PET and SPECT usually provide 3D information not in distinct slices, and the following methods usually pertain to scanning procedures whose output is provided in slices.

[0020] The term "magnetic resonance levels" in the context of the present disclosure pertains to the degree in which different tissues (e.g., corresponding to specific MRI image pixels or areas) resonate with the magnetic field and radio waves used for the magnetic resonance imaging. Such resonance is influenced by the relaxation times, namely T1 and T2, which describe how protons in the tissues return to their equilibrium states after being disturbed by the magnetic field and radiofrequency pulses. These relaxation processes contribute to the contrasts seen in MRI images. Tissues with similar magnetic resonance levels might seem similar in MRI imaging.

[0021] The term "tissue-signal response patterns" in the context of the present disclosure refers to the overarching patterns in how various tissues within the body interact with and influence the signals utilized by different static medical imaging modalities to which the invention pertains. This broader term encompasses both the attenuation levels in CT, which describe the tissue's effect on radiation, and the magnetic resonance levels in MRI, reflecting the tissue's resonance with magnetic fields and radio waves. It is equally applicable to other types of static medical imaging technologies, mutatis mutandis. Beyond merely representing these interactions, "tissue-signal response patterns" also describe how these tissue-specific interactions translate into the brightness, contrasts, similarities, and other variations observed inthe resultant imaging, highlighting the intricate ways in which the imaging technology interacts with diverse types of tissues in the body.

[0022] Possibly, the stage of obtaining may start with prefiltering a larger list of identified discernible objects in the one or more images of the stack, based on one or more criteria such as any combination of one or more of the following: size (e.g., 0- 3mm), shape (e.g., discarding overly elongated discernible objects), gray level (or corresponding attenuation level, e.g., 0-70HU), intensity level (or corresponding magnetic resonance level, e.g., varying levels in T1 or T2 weighted images), inter-image behavior (e.g., little displacement between adjacent image, small number of slices, e.g., <5), and exceeding of local noise levels. It is noted that the list of candidates and / or such as larger list of identified discernible objects may be provided at the beginning of the method, but may also be generated by processing of the one or more images as part of the methods. Such image processing for identifying of discernible objects in an image may implement any suitable algorithm or process, such as the many algorithm known and practiced in the field of image processing (e.g., those belonging to the field of image processing known as "blob detection"). Some nonlimiting examples include: The Laplacian of Gaussian, The difference of Gaussians approach, The determinant of the Hessian, Hessian-Laplace, and Lindeberg's watershed-based grey-level blob detection algorithm.

[0023] Notably, the group of candidates obtained at the initial steps of the method includes many discernible objects which are not representations of cancerous nodules (such as blood vessels, figments, noise, etc.). The group of candidates may include discernible objects which are actual representations of real cancerous nodules, and possibly also of early-stage cancerous nodules, but this is not necessarily so (many CT and MRI scans are performed for cancer-free patients). Even if one or more candidates of the group do represent cancerous nodules, they are usually greatly outnumbered by discernible objects which do not represent cancerous nodules (and discernible objects which do not represent nodules at all), and for reason selecting parameters for the execution of the method which would provide substantial detection probability and maintain the false positive rate at a very low number may be favored in manyscenarios. Examples of tailoring the method to that criterion are provided below, in the detailed description of the invention.

[0024] The method continues with executing the following steps for each out of multiple candidates of the group (but not necessarily for all of them): a. Determining whether the respective candidate discernible object satisfies an isolation requirement with respect to the rest of the candidates (e.g., with respect to small nodules of 1-3 mm). This determining is based at least on geometric remoteness of the respective candidate from the other candidate discernible objects of the group. b. Conditionally reporting an existence of a suspected nodule at the lung corresponding to the respective candidate, in response to a result of the determining (and possibly to other decision criteria).

[0025] For example, if a candidate is determined to satisfy the isolation requirement (in nonlimiting ley terms - it is sufficiently lonely or remote from other candidates of the group) it may be reported as a representation of a suspected nodule (possibly subject to other tests not related to isolation). On the other hand, if the candidate is determined not to satisfy the isolation requirement (in nonlimiting lay terms - there are other neighboring candidates of the group according to the criteria of the isolation requirement), such a candidate cannot usually be reported as a suspected nodule (e.g., a small nodule of 1-3 mm), even if it satisfy other criteria which suggest it may be a nodule (e.g., size, shape, brightness, etc.). In this way, conditioning the reporting of a candidate as a suspected nodule on a geometrical isolation requirement acts as a strong filter which significantly limits the amount of false positive detections which might otherwise have been reported. While not necessarily so, such an isolation requirement may be implemented in an efficient way when including strictly deterministic criteria, without including stochastic processes or criteria that may introduce uncertainty, randomness, or chance). As discussed below, the isolation requirement may be defined in a complex way, taking into account various parameters relating to the candidate discernible object, and possibly also relating to other candidates, to the environment of the candidate, and so on. Importantly, the isolationrequirement may pertain not only to the geometric relationship of the respective candidate to other candidates, but also to other objects identified in the image. For example, the isolation requirement may also include criteria pertaining to the geometric remoteness of the respective candidate from the other objects which have characteristics of certain tissues (e.g., soft tissues, nodules), but are not considered candidates (e.g., because they are too large).

[0026] Clearly, the method is not expected to report any candidates as representing suspected cancerous nodules in cases in which the method is applied to static imaging data (e.g., CT scans, MRI scans) in cases where the method makes a final decision that the patient is healthy from the point of view of suspicion of lung cancer. This is especially important if the method is used as a screening procedure. Patients may include people and / or animals. Nevertheless, false positives might still occur, for instance, if the regions identified correspond to benign nodules or other non- cancerous growths. When applied to a patient with cancerous nodules of specific types (matching to the parameters of the specific implementation of the method), such as early-stage lung cancer, the method is expected to report one or more of the candidates as representing suspected nodules. However, even in such cases, most of the candidates in the static imaging data of the respective patient are expected to result from other conditions (and not from cancerous nodules). Therefore, in such cases, the executing of the method for the aforementioned multiple candidates in a patient having such state of cancer would likely result in: a. Reporting one or more first candidate discernible objects that meets the isolation criteria as suspected nodules (i.e., a cancerous nodule) is suspected at a location (and possibly other parameters such as size and shape) corresponding to the location of the respective candidate). b. Determining that one or more second candidates that do not meet the isolation criteria should not be reported as a suspected nodule. As aforementioned, in many cases in which one or more nodules are reported as suspect nodules, the number of candidates which are not reported assuspected nodules due to failing to meet the isolation requirement is much greater.

[0027] It is noted that the systems, methods, and computer program products discussed herein enable detection of much smaller nodules that is presently available in most if not all commercially available systems— at least in some parts of the body. For example, the disclosed systems, methods, and computer program products enable detection of detecting early-stage nodule of lung cancer, especially in case such nodules are positioned in proximity to the outer wall of the lung (wherein such nodules are common). Referring to the aforementioned method, it is noted that at least one first candidate discernible object being reported as a suspect cancerous nodule by the method may be a small cancerous nodule having an effective diameter of between 1.0-3.0mm which is positioned less than 7cm from an outer wall of the lung. The method enables the detection of that small cancerous nodule using an adaptable isolation threshold that depends on the distance between an analyzed candidate discernible object and the outer wall of the lung. The term "effective diameter" as used herein is defined as the longest straight-line distance between any two points on a trace or a surface of a 2D or a 3D shape (respectively).

[0028] The method may be implemented in many different ways, including by executing human-written code, by executing machine-learning based code, or in any other suitable manner. Optionally, the method may consist of detecting suspect cancerous nodules by performing a series by rule-based logic developed through human-engineered algorithms, and excludes employment of data-trained models. In such cases, implementing the method does not require a large number of examples for training a system, such as the training database which may be required for implementing of methods for detection of lung cancer in static imaging data which are based on Neural Networks (NN), Machine Learning (ML), and / or Deep Learning (DL). Implementing the method without being based on vast training database but rather on rule-based logic developed through human-engineered algorithms (e.g., human- written code) makes the method especially suitable to deal with rare conditions of difficult-to-detect types or conditions of cancerous nodules (for which training data would be rare). It is nevertheless noted that the method is not restricted to rule-basedlogic developed through human-engineered algorithms, and may also include (partly or wholly) parts which are developed using NN, ML, DL, Al, or any other similar fashion.

[0029] According to an aspect of the invention, there is disclosed a system for detecting a lung cancer nodule by processing static imaging data, the system including:(i) at least one tangible memory module operable to store computer program code and information of a group of CDOs within a static-imaging image stack that includes at least one image of a lung, the group of CDOs includes discernible objects having associated tissue-signal response patterns corresponding to cancerous nodules; and (ii) at least one processor. The at least one memory and the computer program code are configured, with the at least one processor, to cause the system to at least: (a) access the information of the group of CDOs stored on the at least one tangible memory module; (b) determine for each CDOs out of multiple CDOs of the group of CDOs whether the respective CDO satisfies an isolation requirement with respect to the rest of the CDOs, based at least on geometric remoteness of the respective CDO from the other CDOs; (c) based on determining that at least one first CDO out of the multiple CDOs meets the isolation criteria, reporting the first CDO as a suspected nodule; and (d) based on determining that at least one second CDO out of the multiple CDOs does not meet the isolation criteria, refraining from reporting the second CDO as a suspected nodule.

[0030] According to a further aspect of the invention, the isolation criteria may be further dependent upon a distance of the respective CDO from an outer wall of the lung.

[0031] According to a further aspect of the invention in which a specific first CDO is closer to the wall of the lung than a specific second CDO, a distance between the a specific second CDO to its nearest CDO of the group of CDOs may be at least twice a distance between the a specific first CDO to its nearest CDO of the group of CDOs.

[0032] According to a further aspect of the invention, the at least one tangible memory module may be further operable to store second information of a second group of CDOs within a second static-imaging image stack that includes at least one second pulmonary image, the second group of CDOs includes discernible objectshaving associated tissue-signal response patterns corresponding to cancerous nodules; and the at least one memory and the computer program code may be configured, with the at least one processor to access the second information stored on the at least one tangible memory module and to process the second information like the information. In that case, both the group of CDOs and the second group of CDOs may include a subspace which consist of a subgroup of CDOs that includes a subject discernible object. The first CDO may be the subject discernible object in the image stack, and a third CDO of the second group of CDOs may be the subject discernible object of the other image stack. A pleural distance of the third CDO in that case may be at least twice a pleural distance of the first CDO. wherein the at least one memory and the computer program code may be configured, with the at least one processor, to determine that the third CDO does not meet the isolation criteria and subsequently to refrain from reporting the third CDO as a suspected nodule.

[0033] According to a further aspect of the invention, the at least one memory and the computer program code may be configured, with the at least one processor, to report analyzed CDOs as suspected nodules only for analyzed CDOs for which there is up to one neighboring CDO of the group of CDOs within a circle of a given radius which includes only the analyzed CDO and the neighboring CDO. The at least one memory and the computer program code in such case may be configured, with the at least one processor, to refrain from reporting any analyzed CDO as a suspected nodule if there are at least two neighboring CDO of the group of CDOs within a circle of the given radius which includes the analyzed CDO.

[0034] According to a further aspect of the invention, the at least one memory and the computer program code may be configured, with the at least one processor, to refrain from reporting any analyzed CDO from being reported as a suspected nodule if the respective analyzed CDO belongs to a subgroup of N>5 CDOs of the group of CDOs, each of which complying with the rule that at least three other CDOs of the subgroup are positioned within a circle of a predetermined radius Rocentered on the respective CDO.

[0035] According to a further aspect of the invention, the at least one memory and the computer program code may be configured, with the at least one processor, to determine whether an analyzed CDO satisfies an isolation requirement with respect to the rest of the CDOs by initially determining whether the respective CDO meets an intra-slice isolation criteria with respect to the rest of the CDOs in the same image as the respective CDO, based at least on geometric remoteness of the respective CDO from the other CDOs; and only if the respective CDOs meets the intra-slice isolation criteria, determining whether the respective CDO meets a 3D isolation criteria with respect to CDOs of at least one other image of the static-imaging image stack. The at least one memory and the computer program code in such case may be further configured, with the at least one processor, to refrain from reporting an existence of a suspected nodule at the lung corresponding to any analyzed respective CDO that do not meet both the intra-slice isolation criteria and the inter-slice isolation criteria.

[0036] According to a further aspect of the invention, the first CDO may be a small cancerous nodule having an effective diameter of between 0.5-1.5mm which is positioned less than 5cm from an outer wall of the lung. The at least one memory and the computer program code in such case may be configured, with the at least one processor, to implement an adaptable isolation threshold that depends on the distance between an analyzed CDO and the outer wall of the lung for detecting the small cancerous nodule.

[0037] According to a further aspect of the invention, the at least one memory and the computer program code may be configured, with the at least one processor, to perform a series by rule-based logic developed through human-engineered algorithms, and excludes employment of data-trained models for detecting the suspect cancerous nodules.

[0038] According to an aspect of the invention, there is disclosed computer implemented method for detecting a lung cancer nodule by processing static imaging data, the method including: (i) obtaining a group of candidate discernible objects (CDOs) within a static-imaging image stack that includes at least one image of a lung, the group of CDOs includes discernible objects having associated tissue-signalresponse patterns corresponding to cancerous nodules; and (ii) executing for each out of multiple CDOs out of the group of CDOs: (a) determining whether the respective CDO satisfies an isolation requirement with respect to the rest of the CDOs, based at least on geometric remoteness of the respective CDO from the other CDOs; and (c) conditionally reporting an existence of a suspected nodule at the lung corresponding to the respective CDO, in response to a result of the determining. The aforementioned executing includes at least: (a) reporting a first CDO that meets the isolation criteria as a suspected nodule; and (b) determining that a second CDO that does not meet the isolation criteria should not be reported as a suspected nodule.

[0039] According to a further aspect of the invention, the isolation criteria may be further dependent upon a distance of the respective CDO from an outer wall of the lung.

[0040] According to a further aspect of the invention, the first CDO may be closer to the wall of the lung than the second CDO, and a distance between the second CDO to its nearest CDO of the group of CDOs may be at least twice a distance between the first CDO to its nearest CDO of the group of CDOs.

[0041] According to a further aspect of the invention, the method may include repeating the steps of obtaining and executing for a second group of CDOs within another static-imaging image stackthat includes at least one pulmonary image. In such case, both the group of CDOs and the second group of CDOs may include a subspace which consist of a subgroup of CDOs that includes a subject discernible object. In such case, the first CDO may be the subject discernible object in the image stack, and a third CDO of the second group of CDOs may be the subject discernible object of the other image stack, and a pleural distance of the third CDO may be at least twice a pleural distance of the first CDO. In such case, the step of executing may be carried out for the third CDO includes determining that the third CDO does not meet the isolation criteria and should not be reported as a suspected nodule.

[0042] According to a further aspect of the invention, the isolation criteria may permit an analyzed CDO to be reported as a suspected nodule if there is up to one neighboring CDO of the group of CDOs within a circle of a given radius which includesonly the analyzed CDO and the neighboring CDO. In such case, the isolation criteria may prevent an analyzed CDO from being reported as a suspected nodule if there are at least two neighboring CDO of the group of CDOs within a circle of the given radius which includes the analyzed CDO.

[0043] According to a further aspect of the invention, the isolation criteria may prevent an analyzed CDO from being reported as a suspected nodule if the respective analyzed CDO belongs to a subgroup of N>5 CDOs of the group of CDOs, each of which complying with the rule that at least three other CDOs of the subgroup are positioned within a circle of a predetermined radius Rocentered on the respective CDO.

[0044] According to a further aspect of the invention, the executing may include: (a) determining whether the respective CDO meets an intra-slice isolation criteria with respect to the rest of the CDOs in the same image as the respective CDO, based at least on geometric remoteness of the respective CDO from the other CDOs; (b) only if the respective CDOs meets the intra-slice isolation criteria, determining whether the respective CDO meets a 3D isolation criteria with respect to CDOs of at least one other image of the static-imaging image stack; and (c) selectively reporting an existence of a suspected nodule at the lung corresponding to the respective CDO, in response to determining that the respective CDO meets both the intra-slice isolation criteria and the inter-slice isolation criteria.

[0045] According to a further aspect of the invention, the first CDO may be a small cancerous nodule having an effective diameter of between 0.5-1.5mm which is positioned less than 5cm from an outer wall of the lung. The method in such case may include detecting the small cancerous nodule using an adaptable isolation threshold that depends on the distance between an analyzed CDO and the outer wall of the lung.

[0046] According to a further aspect of the invention, the method may consist of detecting suspect cancerous nodules by performing a series by rule-based logic developed through human-engineered algorithms, and excludes employment of data- trained models.

[0047] According to a further aspect of the invention, there is disclosed a computer program product for detecting a lung cancer nodule by processing static imaging dataincluding instructions which, when executed by a computer, cause the computer to execute: (a) obtaining a group of candidate discernible objects (CDOs) within a staticimaging image stack that includes at least one image of a lung, the group of CDOs includes discernible objects having associated tissue-signal response patterns corresponding to cancerous nodules; and (b) executing for each out of multiple CDOs out of the group of CDOs: (a) determining whether the respective CDO satisfies an isolation requirement with respect to the rest of the CDOs, based at least on geometric remoteness of the respective CDO from the other CDOs; and (b) conditionally reporting an existence of a suspected nodule at the lung corresponding to the respective CDO, in response to a result of the determining. The executing in such case may include at least: (a) reporting a first CDO that meets the isolation criteria as a suspected nodule; and (b) determining that a second CDO that does not meet the isolation criteria should not be reported as a suspected nodule.

[0048] According to a further aspect of the invention, the isolation criteria may be further dependent upon a distance of the respective CDO from an outer wall of the lung.

[0049] According to a further aspect of the invention, the first CDO may be closer to the wall of the lung than the second CDO, and a distance between the second CDO to its nearest CDO of the group of CDOs may be at least twice a distance between the first CDO to its nearest CDO of the group of CDOs.

[0050] According to a further aspect of the invention, the computer program product may further include instructions which, when executed by the computer, cause the computer to: repeat the steps of obtaining and executing for a second group of CDOs within another static-imaging image stack that includes at least one pulmonary image. In such case, the following may optionally be implemented: (a) both the group of CDOs and the second group of CDOs include a subspace which consist of a subgroup of CDOs that includes a subject discernible object; (b) the first CDO is the subject discernible object in the image stack, and a third CDO of the second group of CDOs is the subject discernible object of the other image stack; and (c) a pleural distance of the third CDO is at least twice a pleural distance of the first CDO. Thedetermining in such case may include determining that the third CDO does not meet the isolation criteria and should not be reported as a suspected nodule.

[0051] According to a further aspect of the invention, the isolation criteria may permit an analyzed CDO to be reported as a suspected nodule if there is up to one neighboring CDO of the group of CDOs within a circle of a given radius which includes only the analyzed CDO and the neighboring CDO. The isolation criteria in such case may prevent an analyzed CDO from being reported as a suspected nodule if there are at least two neighboring CDO of the group of CDOs within a circle of the given radius which includes the analyzed CDO.

[0052] According to a further aspect of the invention, the isolation criteria may prevent an analyzed CDO from being reported as a suspected nodule if the respective analyzed CDO belongs to a subgroup of N>5 CDOs of the group of CDOs, each of which complying with the rule that at least three other CDOs of the subgroup are positioned within a circle of a predetermined radius Rocentered on the respective CDO.

[0053] According to a further aspect of the invention, the computer program product may further include instructions which, when executed by the computer, cause the computer to: (a) determine whether the respective CDO meets an intra-slice isolation criteria with respect to the rest of the CDOs in the same image as the respective CDO, based at least on geometric remoteness of the respective CDO from the other CDOs; (b) selectively determine, only if the respective CDOs meets the intra- slice isolation criteria, whether the respective CDO meets a 3D isolation criteria with respect to CDOs of at least one other image of the static-imaging image stack; and (c) selectively report an existence of a suspected nodule at the lung corresponding to the respective CDO, in response to determining that the respective CDO meets both the intra-slice isolation criteria and the inter-slice isolation criteria.BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to understand the invention and to see how it may be carried out in practice, embodiments will now be described, by way of non-limiting examples only, with reference to the accompanying drawings, in which:

[0055] Figs. 1A-1D illustrate a flow chart illustrating an example of a method for detecting suspicious nodules based on processing of static medical imaging;

[0056] Figs. 2A and 2B represent execution of steps of the aforementioned method on a schematic representation of a such an image;

[0057] Fig. 3 illustrates a candidate discernible object and a peer object, and a technique to determine distance between these two objects;

[0058] Fig. 4 illustrates an example of a cluster of candidates and their spatial relationship to other candidates;

[0059] Fig. 5 illustrates an example of a cluster that includes a number of candidates;

[0060] Fig. 6 illustrates two modeled 3D objects and their spatial relationships;

[0061] Fig. 7 provides an example of a CT image of a cross-section of lungs of a person;

[0062] Fig. 8 is a schematic transverse section through a thorax of a person, viewed from above;

[0063] Fig. 9 is a schematic diagram of different types of blood vessels in the lungs;

[0064] Fig. 10 illustrates a flow chart illustrating an example of a method for detecting cancerous nodules based on processing of static imaging data;

[0065] Fig. 11 includes four diagrams schematically illustrating different ways of implementing isolation requirement which is dependent on the spatial relationship of the candidate;

[0066] Fig. 12 illustrates assessing isolation of candidates at different distances from the outer wall of the lung; and

[0067] Figs. 13A and 13B are functional block diagrams illustrating an example of a system for detecting a lung cancer nodule by processing Static imaging data.

[0068] It will be appreciated that for simplicity and clarity of illustration and description, certain elements in the figures may not have been drawn to scale. This could include the exaggeration of certain element dimensions relative to others.Additionally, corresponding or analogous elements may be identified using repeated reference numerals in the figures.DETAILED DESCRIPTION OF EMBODIMENTS

[0069] In the following detailed description, numerous specific details are set forth in order to provide a thorough understanding of the invention. While specific details are provided to enable a thorough understanding of the invention, those skilled in the art will appreciate that the invention may be practiced without these details. Additionally, well-known methods, procedures, and components are not described in detail to avoid obscuring the invention. Finally, any reference to a method, system, or non-transitory computer readable medium should be interpreted as including related aspects of the invention.

[0070] The terms "computer", "processor", and "controller" should be expansively construed to cover any kind of electronic device with data processing capabilities, including, by way of non-limiting example, a personal computer, a server, a computing system, a communication device, a processor (e.g. digital signal processor, DSP), a microcontroller, a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), a smartphone, an electronic control unit (ECU) of a vehicle, cloud computing servers, an and so on. Unless stated otherwise, the terms "computer", "processor", and "controller" may also include a combination of several modules (e.g., several central processing units, CPUs), which operate together toward a goal. Unless specifically stated otherwise, as apparent from the following discussions, it is appreciated that throughout the specification discussions utilizing terms such as "processing", "calculating", "computing", "determining", "generating", "setting", "configuring", "selecting", "defining", or the like, include actions and / or processes of a computer that manipulate and / or transform data into other data. That data is represented as physical quantities, e.g., such as electronic or electromagnetic quantities, and / or said data representing physical objects.

[0071] It is appreciated that certain features of the presently disclosed subject matter, which are, for clarity, described in the context of separate embodiments, mayalso be provided in combination in a single embodiment. Conversely, various features of the presently disclosed subject matter, which are, in the interest of concision, described in the context of a single embodiment, may also be provided separately or in any suitable sub-combination. In embodiments of the presently disclosed subject matter one or more steps illustrated in the figures may be executed in a different order and / or one or more groups of steps may be executed simultaneously. The figures illustrate a general schematic of the system architecture in accordance with an embodiment of the presently disclosed subject matter. Each module in the figures can be made up of any combination of software, hardware and / or firmware that performs the functions as defined and explained herein. The modules in the figures may be centralized in one location or dispersed over more than one location.

[0072] Any reference in the specification to a method should be applied mutatis mutandis to a system capable of executing the method. Any reference in the specification to a method which can be executed by a computer should be applied mutatis mutandis to a non-transitory computer readable medium that stores instructions that once executed by a computer result in the execution of the method. All the details, variations, optional features, optional steps which are discussed with respect to a system are also applicable, mutatis mutandis, to such a corresponding method (and non-transitory computer readable medium, where applicable), and vice versa.

[0073] Figs. 1A-1D illustrate a flow chart illustrating an example of method 400, in accordance with the presently disclosed subject matter. Method 400 is a computer implemented method for detecting suspicious nodules based on processing of static medical imaging, such as CT images or MRI images. Such static medical imaging methods (and especially static multi-slice diagnostic imaging method) usually provide information in stacks of adjacent images, each corresponding to a parallel crosssection of the body of the patient. While the in-slice resolutions of the scan and the distance between slices may vary depending on various factors (e.g., the specific scanner, the imaging protocol used, and the particular requirements of the examination), common in-plane resolution of diagnostic CT scanning is often in the range of 0.3-lmm per pixel (roughly 10-33 pixels per cm), and the displacementbetween parallel slices often varies between 0.5-10 mm. Higher resolution CT scans can be acquired but may expose the patient to higher levels of radiation. Specialized scanners and protocols may be used to achieve higher resolutions. Common in-plane resolution of MRI scans is typically in the range of 0.3 -1.5mm per pixel (roughly 7-33 pixels per cm), and the displacement between parallel slices commonly varies between l-5mm. That means that even relatively small nodules (e.g., l-3mm) are represented by tens of pixels in each slice in which they are manifested.

[0074] Considering method 400 as a whole, it is noted that method 400 may be used for detecting lung cancer in patients in which there is already a suspicion for lung cancer, or for following the development of lung cancer in patients in which lung cancer was already diagnosed. However, the ability for detecting small nodules— especially in proximity to the walls of the lungs where they are likely to be located — may also be implemented for screening static imaging data of large populations of people for which there is greater uncertainty regarding the presence of lung cancer (and especially - of early state lung cancer). Method 400— as well as method 500, system 200, and the disclosed computer program products — may be implemented sufficiently cheaply and efficiently to be implemented for static imaging data collected for a vast number of people, allowing to screen, detect, and hopefully to allow treatment and cure of lung cancer in large population. Large masses of the population (for example, specific risk groups, and / or people whose lungs were examined for any other reason) may be examined, even when there is no preliminary data on suspicion of lung cancer and people who are presumably healthy are examined.

[0075] Optionally, method 400 may be implemented on a general computer or processor. Alternatively, method 400 may be executed using dedicated hardware, such as an ASIC or specialized architecture. As CT scans and MRI scans become more affordable and thus more common, and as cancer prevalence grows, there is a growing incentive for mass implementation of method 400 for processing growing amount of static medical imaging scans done with the intention of detecting lung cancer (either as a primary goal of the static medical imaging, or as secondary processing of static medical imaging scans taken for other purposes). These trends also promote the development of dedicated hardware for efficient execution of method 400 if it is to beimplemented in large multitude. Nevertheless, as stated above, method 400 may also be implemented without such dedicated hardware. It is noted that the discussion of method 400 commonly uses the example of processing CT data. However, as discussed above, method 400 is also applicable to MRI scans and to any other type of medical static imaging, mutatis mutandis.

[0076] Step 410 of method 400 includes obtaining a group of identified discernible objects within a CT image stack that includes at least one CT image (e.g., of a lung). This group of identified discernible objects includes discernible objects having associated attenuation levels corresponding to cancerous nodules. For example, the brightness level (or grayscale level or other color parameters, if applicable) may correspond to HU levels of cancerous nodules of the kind sought in the specific implementation of method 400 (e.g., early state lung cancer nodules). Referring to the examples set forth with respect to other diagrams of this disclosure, step 410 may optionally be executed by processor 220, e.g., by filtering module 222 or by isolation analysis module 224.

[0077] Optionally, step 410 may include generating the group of identified discernible objects, by applying various techniques of image processing. Many techniques and algorithms for identifying discernible objects are known in the art, and various algorithms may be selected depending on the specific implementation of method 400 (e.g., types of CT images processed, types of lung cancer nodules sought, types of medical procedures performed prior to the CT scan, such as whether contrast dye was used). For example, a simple way of generating the group of identified discernible objects is to identify regions (possibly qualifying some geometrical constraints like size) whose brightness is between a predefined range. In other examples, some more complex algorithms may take into account additional factors such as brightness gradients, brightness derivatives, texture, homogeneity, and so on. Alternatively (or additionally), step 410 may include receiving the group of discernible objects as an input to method 400 (e.g., provided by another system, another processor, etc.).

[0078] Step 410 is followed by step of 420 of filtering the group of identified discernible objects to provide a group of candidate discernible objects (also referred to as "group of candidates") that retains a plurality of discernible objects of the filtered group, while excluding other discernible objects of the filtered group. It is noted that filtering out of identified discernible objects may be done for various reasons. One example includes filtering out identified discernible objects whose parameters render them to have very low likelihood (even negligible likelihood) of resulting from an actual nodule. Another example includes filtering out identified discernible objects which might result from an actual nodule, but for which there is a better algorithm to improve the detection rate to false positives rate compared to method 400. Other considerations may also lead to filtering out of discernible objects in step 420. Referring to the examples set forth with respect to other diagrams of this disclosure, step 420 may optionally be executed by processor 220, e.g., by filtering module 222.

[0079] Referring to step 420 as a whole, the filtering may include any one or more of the following: a. Optional step (not denoted) of filtering out discernible objects based on their validity with respect to a noise threshold, where the noise threshold may optionally be determined locally, pertaining only to a parts of the respective CT slice. Filtering out of borderline discernible objects (which are not clearly above the noise threshold) may be implemented in order to reduce the likelihood of false positive errors, at the potential expense of missing out some very early-stage nodules. b. Optional step 421 of filtering out identified discernible objects based on their size. For example, step 421 may include filtering out any discernible object having a two points more remote from one another than a predefined length. For example, discernible objects larger than 3mm, 4mm, or 5mm (or any fractional measurement between 2-10mm) may be filtered out. c. Optional step 422 of filtering out identified discernible objects based on their shape. For example, step 422 may include filtering out overly elongated discernible objects (e.g., because such objects are not likely to representcancerous nodules, and are very likely to represent an imprint of a blood vessel in the imaged CT slice). For example, objects with a length-to-width ratio higher than a predefined ratio (e.g., 1:3, 1:4, 1:5, etc.) may be filtered out. d. Optional step 423 of filtering out identified discernible objects based on their relationship to information in other CT images, e.g., information in adjacent CT slices, such as information relating to discernible objects identified in adjacent CT slices. For example, step 423 may include filtering out identified discernible object which have counterpart discernible objects (i.e., in similar or overlapping location) in more than a predefined number of slices (e.g., >4), e.g., because such objects are not likely to represent cancerous nodules, and are very likely to represent an imprint of a blood vessel across multiple CT slices. In another example, step 423 may include filtering out identified discernible object whose counterpart discernible objects in adjacent slices are displaced (or "shifted") beyond a predefined measure (e.g., >lmm, >50% of the measure of the respective discernible object), e.g., because such objects are not likely to represent cancerous nodules, and are very likely to represent an imprint of a blood vessel across multiple CT slices. e. Optional step 424 of filtering out identified discernible objects based on having characteristics of other tissues or synthetic causes. A few examples were provided above with relating to filtering out identified discernible objects with high likelihood of being blood vessels, e.g., because these identified discernible objects have high likelihood of causing false positive reporting of suspected cancer. The same principles may be extended to other types of tissues or of synthetic causes for such small bright patches in the CT image (figments), e.g., depending on the type of CT scanning implemented, type of medical procedures performed prior to the CT scan, and so on.

[0080] It is noted that steps 421, 422, 423, 424, are just examples of the types of filtering which may be carried out in step 420, and that step 420 may include any combination of one or more of these steps, or none of these examples. Additionally,step 420 may implement any other suitable form of filtering, e.g., as required by the specific implementation. Regarding step 420 as a whole, it is noted that step 420 may serve an important role in the aforementioned attempt of minimizing occurrence of false positives.

[0081] It is noted that if the CT stack includes more than one CT image, step 420 may be executed for one, some, or all of the CT images of the stack. If implemented for more than one CT image of the stack, step 420 may be executed for each slice separately, in a slice-by-slice manner but using information of other CT images of the stack (e.g., sub-step 423), for a plurality of CT images together implementing crosslayer algorithms, or in any other suitable fashion.

[0082] Step 420 is followed by step 430 of obtaining a group of candidate discernible objects within the CT image stack, resulting from the filtering of step 420. Referring to the examples set forth with respect to other diagrams of this disclosure, step 430 may optionally be executed by processor 220, e.g., by filtering module 222. It is noted that information pertaining to the various candidate discernible objects may be received in many different formats and include many types of data, depending on the specific implementation. Such types of data may include, for example, any combination of one or more of the following: clipped images, location of center, dimensional information, shape information, shape parameterization (e.g., roughness of the circumference, irregularity of the edge), texture parameters, color parameters, and so on.

[0083] It is noted that method 400 (either in step 430 or in previous steps) may optionally include a step of improving quality or otherwise processing information the individual discernible objects. Such a step may include improving contrast, determining derivatives, determining characteristics parameters (e.g., texture, irregularity of edge), exacting location information, deriving geometrical parameters, and so on.

[0084] It is noted that if the CT stack includes more than one CT image, step 430 may be executed for one, some, or all of the CT images of the stack. If implemented for more than one CT image of the stack, step 430 may be executed for each slice separately, in a slice-by-slice manner but using information of other CT images of thestack, for a plurality of CT images together implementing cross-layer algorithms, or in any other suitable fashion.

[0085] Step 440 includes generating a group of peer objects for analysis (e.g., geometrical analysis and optionally additional types of analysis), based at least one the group of candidate discernible objects, and possibly on the group of identified discernible objects. Referring to the examples set forth with respect to other diagrams of this disclosure, step 440 may optionally be executed by processor 220, e.g., by filtering module 222. As mentioned above the group includes a plurality of candidates but may optionally also include other objects which are not candidates (e.g., corresponding to blood vessels or to potential larger nodules for whose detection is potentially handled by another algorithm). It is noted that optionally, step 440 may include generating a plurality of groups of peer object, to perform the geometrical analysis of group independently. Such groups may correspond to the different CT slices, to different types or classes of objects, or any other ways of classification. Such groups may be mutually exclusive (i.e., a single object may not be represented in two or more groups), but this is not necessarily so. Step 440 may optionally include generating a different group of peer objects for each of the one or more CT images of the stack. Step 440 may optionally include generating at least one multilayer group of peer objects that includes objects (whether candidates or not) from multiple slices (e.g., adjacent CT slices).

[0086] Figs. 2A and 2B represent execution of steps 410, 420, 430, and 440 on a schematic representation of a CT image, in accordance with examples of the presently disclosed subject matter. It is noted that the order in which the sub-steps of filtering (e.g., by shape, by size, by color) is merely an example, and the filtering may be done in any order, possibly combining different types of filtering into a single computation.

[0087] Diagram 801 represents the group of identified discernible objects within a single CT image obtained in step 410, (for example, either by processing of a CT image, by processing of pre-processed CT image data, or by receiving from an external entity). In the illustrated examples, the discernible objects have different sizes, shapes, anddifferentiation in textures and / or colors (represented by the white and gray fill colors of the different shapes).

[0088] Diagram 802 represents an instance of executing step 422 on the output of the last step, by filtering out overly elongated shapes. Diagram 803 represents an instance of executing step 424 on the output of the last step, by filtering out shapes whose color and / or texture indicates lower match to the type of cancerous nodules sought in the example implementation of method 400. Diagram 804 represents a second instance of executing step 422 on the output of the last step, by filtering out overly large shapes. The output of that last step of filtering in the illustrated example is the group of candidate discernible objects. Diagram 805 represent generating of the peer group of objects in relation to which the candidates of diagram 804 will be evaluated (at least in the relative geometric aspects). The group of diagram 805 includes all of the candidates of diagram 804, and some of the objects of the original groups of identified discernible objects of diagram 801. As mentioned above (and further discussed below), the group (or groups) of peer objects to which a candidate may be compared may also include objects (candidates or not) of other slices; this option is not illustrated in the examples of Figs. 2A and 2B.

[0089] Reverting to the discussion of step 440, It is noted that if the CT stack includes more than one CT image, step 440 may be executed for one, some, or all of the CT images of the stack. If implemented for more than one CT image of the stack, step 440 may be executed for each slice separately, in a slice-by-slice manner but using information of other CT images of the stack, for a plurality of CT images together implementing cross-layer algorithms, or in any other suitable fashion. As discussed below in greater detail, outputs of step 440 may include one or more single-layer groups of peer objects and / or cross-layer groups of peer objects.

[0090] Method 400 continues with step 450 of applying single-layer geometric analysis to candidates of the group of candidates. Referring to the examples set forth with respect to other diagrams of this disclosure, step 450 may optionally be executed by processor 220, e.g., by isolation analysis module 224. For example, step 450 may include determining whether each respective candidate discernible object satisfies anisolation requirement with respect to other objects of one or more peer groups to which the respective candidate belongs. The isolation requirement is based at least on geometric remoteness of the respective candidate discernible object from the other candidate discernible objects. The geometric computations may optionally involve explicitly determining distances between the respective candidates to other objects of the peer group, but techniques which do not require explicit determining of distance to each other object of the group may also be implemented (e.g., only searching for other peer object within a small area around the respective candidate). Step 450 may optionally include determining the isolation requirement, using a predefined isolation requirement, or amending a predefined isolation requirement. Some examples for determining different single-layer isolation requirements for different candidates are discussed below.

[0091] The isolation of a discernible object suspected of representing a cancerous nodule (especially an early-stage nodule) may be implemented as a leading factor that make it possible to suppress False Positive. Optionally, a relatively unnuanced implementation of method 400 may include a simple implementation in which the geometric isolation requirement is the main (and possibly only) filter used to differentiate between candidates which are reported as indicative-of or suspicious-as cancer to those which are not reported as such. That is, method 400 can be useful if only including the filtering of isolation with respect to other objects (especially if including more advanced variations of this isolation requirements, e.g., as discussed below). However, while steps 420 and 430 are optional in this sense, method 400 may perform even better if additional steps of filtering are performed.

[0092] Importantly, the isolation requirement may pertain not only to the geometric relationship of the respective candidate to other candidates, but also to other objects identified in the image, included in a group of peer objects in which the candidate object is included. For example, the isolation requirement may also include criteria pertaining to the geometric remoteness of the respective candidate from the other objects which have characteristics of certain tissues (e.g., soft tissues, nodules), but are not considered candidates (e.g., because they are too large).

[0093] It should be noted that the isolation requirement might include several isolation conditions, criteria, or sub-requirements, for a candidate discernible object to meet. The isolation requirement might also implement a complex system of conditions requirements (e.g., the candidate must meet conditions A and B or condition C, but not both, in order to be considered isolated, and so on). Several examples of criteria for isolation are discussed below, and method 400 may include any combination of one or more of these criteria, or implement additional criteria of isolation (in addition or instead of these examples).

[0094] A first example of an isolation requirement is that a candidate might be considered isolated with respect to the other peer objects (of one or more group) in the respective slice if the distance D from that candidate to the nearest member of the group of peer objects is greater than a threshold limit.

[0095] An example algorithm for finding the distance between two discernible objects on the same CT slice is described with respect to Fig. 3. Fig. 3 illustrates a candidate discernible object A and a peer object B, and a technique to determine distance between these two objects. It should be noted that this is just one example, and many other metrics for determining distance between objects may be implemented. The illustrated example algorithm includes the following steps: a. Finding a center of mass of candidate A (denoted 101 in the diagram). Many ways of determining center of mass of 2D shapes are known in the art, and any suitable algorithm may be used; b. Finding point 102 on the contour of peer object B, which is the point on the contour of B which is closest to point 101; c. Finding point 103 on the contour of candidate A, which is the point on the contour of B which is closest to point 102; d. Determining the distance between points 102 and 103, and using this distance to represent the distance between candidate A and peer object B.

[0096] Reverting to the discussion of step 450, the threshold distance limit may be determined according to various factors, such as any one or more of the following nonlimiting examples: a. The maximum size of nodule for which the algorithm is searching (such as the maximum size of early-stage lung cancer to be found; e.g., D0=3mm, 4mm, or 5mm, or any fractional measurement between 2-10mm); b. A distance between the candidate and an outer wall of the lung (for example, may be the is the shortest distance from the center of the candidate to the outer wall of the lung, the shortest distance between any point on the edge of the candidate and the outer wall of the lung, and so on); c. Scan parameters used by in the CT scan in which the CT image was created; d. Information about the patient (e.g., age, gender, weight, BMI, years of smoking).

[0097] For example, according to this first example of isolation criterion, a candidate discernible object will be considered isolated if the shortest distance between this candidate and its nearest pear (e.g., center to center, edge to edge) is larger than a threshold distance determined based on parameters such as the aforementioned examples, e.g., "candidate A is isolated if at least or "candidate A isisolated if and only if Functions can be built in differentways, e.g., experimentally, based on the experience of radiologists, based on machine learning, or in any other suitable way. Two nonlimiting examples for such rules are "candidate A is isolated if at least and "candidate A is isolated if and only if

[0098] Sub-step 451 of step 450 includes assessing the isolation of a candidate in response to maximal size of nodule sought (the aforementioned Do) .

[0099] Sub-step 452 of step 450 includes assessing the isolation of a candidate in response to its distance from an outer wall of the lung (the aforementioned D1A). Optionally, D1A may be a distance within a single CT slice. Alternatively, D1A may be a cross-layer distance (for example, the closest part of the wall of the lung may beimaged in another CT slice). The distribution of cancerous nodules of specific parameters (e.g., size) may vary depending on the distance from the outer wall of the lung). Additionally, or alternatively, the distribution of other specific types of tissues or physiological objects (e.g., blood vessels) might also vary depending on their distance from the outer wall of the lung. Therefore, the likelihood of a candidate of being indicative of a cancerous nodule might depend on its distance from the outer wall of the lung. Also, the likelihood of a candidate being indicative of a more dangerous cancerous nodule might depend on its distance from the outer wall of the lung.

[0100] Therefore, in at least some types of lung cancer, the likelihood of a candidate being indicative of a cancerous nodule under specific condition of isolation might depend on its distance from the outer wall of the lung. Therefore, step 450 may include applying different isolation requirements to different candidates based on their distance from the outer wall of the lung.

[0101] Optionally, given two candidates located in different distances from the outer wall of the lung, step 452 might include determining different threshold distances for the two candidates in order to allow them to be considered isolated. For example, step may include determining two threshold distances DAminand DBminsuch that DAmin<DBmin, where DAminis the threshold distance to candidate A which is closer to the outer wall of the lung and DBminis the threshold distance to candidate B which is more remote from the outer wall of the lung. In such case, candidate A might be considered isolated and possibly reported as indicative of suspect cancerous nodule even though a distance DA between candidate A and the peer object closest to candidate A is shorter than the distance DB between candidate B and a peer object closest to candidate B; candidate B in such case might not be considered isolated, even in cases in which distance DB between candidate B and its closest neighbor is significantly larger than the corresponding distance for candidate A, DA (e.g., >xl.5 larger, >x2 larger, >x5 larger, etc.). For clarity only, it is noted that such condition may involve calculating the distances between a candidate and the outer wall of the lung, but not necessarily so. For example, implicit or implied assessment of the geometric relationship between the respective candidate and the outer wall of the lung may alsobe optionally implemented. Isolation criteria which may be implemented, and which are based on the proximity of the candidate to the outer wall of the lung may follow the following guiding principle: the closer is the candidate is to the outer wall of the lung, the fewer restrictions are imposed on his loneliness, and the further the candidate is from the outer wall of the lung, the greater the restrictions are placed on his loneliness. The goal of these actions is to reduce False Positives while maintaining sensitivity.

[0102] Optionally, step 450 might include assessing the isolation of a candidate in response to its position within the lung. In a way, this might be considered a generalization of sub-step 452 discussed above, and all the variations and nuances discussed with respect to sub-step 452 might be applied to assessing the isolation of a candidate in response to its position within the organ, mutatis mutandis. For example, step 450 might include assigning different isolation criteria to discernible objects that are positioned in a lower part of the lung than to discernible objects that are positioned in a higher part of the lung.

[0103] Optional substep 453 of step 450 includes assessing the isolation of a candidate in response to its inclusion status in a cluster of proximate peer objects that is remote from other peer objects in the group of peer objects. Optionally the cluster may include only candidates. Alternatively, the cluster may include at least one candidate and at least one peer object that is not a candidate. Referring to the examples set forth with respect to the drawings, step 453 may optionally be executed by processor 220, e.g., by clustering analysis module 226. Fig. 4 illustrates an example of cluster 117 that includes four candidates (denoted 111, 112, 113, and 114) which are relatively remote from the rest of the candidates and the peer objects. One example reason to assess whether a candidate is included in such a cluster is that in many cases, few peer objects of small size (e.g., corresponding to early-stage cancerous nodule, such as under 3mm) may be a result of a branching of a "tree" of blood vessels, not far from the branching node. Therefore, the detection of such a phenomenon might reduce the probability of false positives.

[0104] An example criteria to determine the inclusion status of a candidate in a cluster of proximate candidates is a candidate belongs to a cluster of candidates if there are at least three candidates whose respective mutual distances are all smaller than D2, where D2=K2-D0. As established before, Dois the maximum size of nodule for which the algorithm is searching, and K2is a number larger than 1 (K2> 1). The value of K2might be determined in different ways, e.g., experimentally, based on the experience of radiologists, based on machine learning, or in any other suitable way. For example, values such as K2=1.25, K2=1.74, and K2=5 (as well as any other constant, matching to the relevant type of lung cancer analyzed, for example) might be used. Other functions to establish Z)2can also be built in different ways, e.g., experimentally, based on the experience of radiologists, based on machine learning, or in any other suitable way. Similar criteria may optionally also be used for a cluster which might also include non-candidate peer object, if applicable.

[0105] The criteria for the required remoteness of an established cluster 117 from other peer objects might also be determined in similar ways. For example, the distance between any of the candidates of cluster 117 to any other peer object of the group of peer object might be required to exceed a minimal threshold D3determined according to one or more parameters (examples of relevant parameters were discussed above). For example, optionally, D3= K3-Dowhere K3> K2. The value of K3might be determined in different ways, e.g., experimentally, based on the experience of radiologists, based on machine learning, or in any other suitable way. For example, values such as K3=1.77, K3=2, K3=2.55, and K3=6 (as well as any other constant, matching to the relevant type of lung cancer analyzed, for example) might be used. For example, different ratios between K3and K2might be implemented, such as K3>1.5 K2, K3>2 K2IK3>5 K2,etc. Again, different criteria might be useful for different types of lung cancer, different types of CT scans, patients of different characteristics, and so on. Other functions to establish D3can also be built in different ways, e.g., experimentally, based on the experience of radiologists, based on machine learning, or in any other suitable way. Similar criteria may optionally also be used for a cluster which might also include non-candidate peer object, if applicable. A circle 119 has a radius of K3.

[0106] Step 453 may be used for reducing the likelihood of False Positives, e.g., by requiring that candidates that are members of clusters will be subjected to stricter requirements for loneliness. A possible rational for implementing step 453 is that belonging to a cluster indicates a pattern which unite a group of objects, and thus the likelihood that early-stage lung cancer has such a pattern is negligible, and the likelihood that an early-stage cancerous nodule would develop right next to such a cluster is also very low. For clarity only, it is noted that such condition may involve calculating the distances between the candidates to themselves and to the outer wall of the lung, but not necessarily so. Implicit or implied assessment of the geometric relationship between the respective candidate and the respective part of the outer part of the wall might also be optionally implemented.

[0107] Optional substep 454 of step 450 includes assessing the isolation of a candidate in response to its inclusion status in a cluster of at least N proximate peer objects. Optionally the cluster may include only candidates. Alternatively, the cluster may include at least one candidate and at least one peer object that is not a candidate. Referring to the examples set forth with respect to the drawings, step 454 may optionally be executed by processor 220, e.g., by clustering analysis module 226. Fig. 5 illustrates an example of cluster 129 that includes more than N candidates (in the illustrated example eight candidates denoted 121-128, e.g., N=7). As illustrated, the requirements about the remoteness of the cluster itself from other peer objects are more relaxed in comparison to step 453, if at all implemented. One example reason to assess whether a candidate is included in such a cluster is that in many cases, the number of candidates clustered together in a relatively small part of the CT image might indicate that there is a relatively high likelihood that such candidates are a manifestation of some other phenomenon, and not early-stage cancer. Example criteria for determining a presence of a cluster 129 include: for each out of the at least N candidates there are at least three other candidates within a radius of R from this candidate; a density of candidates within an area of predefined size (e.g., a circle with a radius of ft) the dot density remains above a predefined density P. The parameters of this algorithm (N, R, ft, P) can might be determined in different ways, e.g.,experimentally, based on the experience of radiologists, based on machine learning, or in any other suitable way.

[0108] It is noted that if the CT stack includes more than one CT image, step 450 may be executed for one, some, or all of the CT images of the stack. If implemented for more than one CT image of the stack, step 450 may be executed for each slice separately, in a slice-by-slice manner but using information of other CT images of the stack (for example, if a cluster of at least N objects was identified in a certain CT image, any candidates in the corresponding area of one or more adjacent layers— e.g., ±2 layers— might be assumed belonging to the same cluster and thus not indicative of cancerous nodules, without further calculations), for a plurality of CT images together implementing cross-layer algorithms, or in any other suitable fashion.

[0109] Regarding step 450 as a whole, it is noted that the single-layer isolation requirements pertain only to the information included in a single CT slice. This means that a candidate might be considered as demonstrating single-layer isolation, even if that candidates has a very near neighbor in another slice (could be a neighboring slice, or a somewhat removed slice, e.g., two, three, or four slices way). For that reason, method 400 may optionally continue with step 460 of applying cross-layer geometric analysis of candidates of the group of candidates. As discussed later in greater detail, while implementing steps 450 and 460 serially might have advantages in certain scenarios (e.g., computation cost, computation speed, etc.), in some cases more intertwined assessments of single-layer and cross-layer geometric isolation aspects might be implemented.

[0110] Step 460 includes applying cross-layer geometric analysis of candidates of the group of candidates. Referring to the examples set forth with respect to other diagrams of this disclosure, step 460 may optionally be executed by processor 220, e.g., by isolation analysis module 224. For example, step 460 may include determining whether each respective candidate discernible object satisfies an isolation requirement with respect to objects of one or more cross-layer peer groups to which the respective candidate belongs. It is noted that method 400 may also include generating at least one cross-layer group of peer objects for cross-layer geometricalanalysis, based on the at least one group of candidate discernible objects of each respective layer and / or based on the outputs of the single-layer isolation assessment of step 450. Any detail discussed with respect to generating group of peer objects in step 440 might be applied to the generation of the one or more cross-layer groups of peer objects, mutatis mutandis. The generation of the one or more cross-layer groups of peer object might be executed as part of step 440, before step 450, after step 450, or in any other suitable fashion. Step 450 may optionally include determining the cross-layer isolation requirement, using a predefined cross-layer isolation requirement, or amending a cross-layer predefined isolation requirement. Some examples for determining different cross-layer isolation requirements for different 3D objects are discussed below.

[0111] The cross-layer isolation requirement is based at least on geometric remoteness of the respective candidate discernible object from the other candidate discernible objects of different layers. Some of the requirements (e.g., threshold distances, minimal number of items in a cluster, etc.) might be similar or identical to these of the single-layer isolation requirement discussed above with respect to step 450, but this is not necessarily so. Any detail discussed with respect to the applying of single-layer isolation requirement in step 450 might be applied to the applying of the cross-layer isolation requirement in step 460, mutatis mutandis, and many such details are not repeated in the interest of concision. For example, step 460 may include the cross-layer equivalent of assessing of isolation of a candidate in response to any combination of one or more of: (a) the maximal size of nodule sought, (b) a cross-layer distance of the candidate from the outer wall of the lung, (c) an inclusion status of the candidate in a cross-layer cluster of proximate peer objects that is remote from other peer objects in the group of peer objects, and (d) an inclusion status of the candidate in a cross-layer cluster of at least N proximate peer objects. All of these sub-steps are collectively denoted step 461 in the diagram of Fig. IB.

[0112] The cross-layer geometric computations may optionally involve explicitly determining distances between the respective candidates to other objects of the peer group, but techniques which do not require explicit determining of distance to each other object of the group may also be implemented (e.g., only searching for other peerobject within a small area around the respective candidate. It should be noted that the cross-layer isolation requirement might include several isolation conditions, criteria, or sub-requirements, for a candidate discernible object to meet. The isolation requirement might also implement a complex system of conditions requirements. Several examples of criteria for isolation are discussed below, and method 400 may include any combination of one or more of these criteria, or implement additional criteria of isolation (in addition or instead of these examples).

[0113] Step 460 may include a multi-step subprocess which includes first substep 462 of modeling three-dimensional (3D) objects which correspond to discernible objects of separate slices, second substep 464 of assessing 3D loneliness of such 3D objects with respect to other 3D objects and / or to discernible objects of one or more CT slices, and third substep 465 of filtering out 3D objects based on their 3D loneliness status. The multi-step subprocess of step 460 may also include optional substep 463 of Filtering out 3D objects based on their 3D parameters.

[0114] Step 462 of modeling 3D objects which correspond to discernible objects of separate slices might be executed in different ways, and may optionally be based on various algorithm known in the art for assessing a shape of a 3D object from a plurality of images, at least some of which including cross-sections of that object. Any suitable format of output might be implemented, such as (but not limited to): a stack of 2D discernible objects, a stack of 2D discernible object with additional 3D parameters, a 3D model of the shape (also indicative of perimeter of the shape between adjacent slices, for example), and so on. It should be noted that the distance between neighboring slices of the stack is usually provided with the stack of images or otherwise provided, and may be used in the preparing of the 3D model.

[0115] Step 462 may implement different ways for determining whether discernible objects of two neighboring slices (e.g., immediately adjacent slices) refer to a single 3D dimensional physical object. Both such objects may be required to be candidates, but optionally one of them may be a candidate and the other may be another object (e.g., a previously eliminated discernible object). For example, if a candidate in slice M is determined to belong to a same 3D object of a discernible object of slice M+l whichwas filtered out as being too large, the candidate in M might be considered a tapering off of that larger object, and therefore be eliminated.

[0116] Optionally, step 462 may include determining that two discernible objects of neighboring slice belong to the same 3D object if the orthogonal projections of these discernible objects (on a plane parallel to the plane of the slices) are separated from each other by less than where: D14and D24are the size ofeach the respective two discernible objects. Different metrics of separations may be used to determine the separation between the orthogonal projections of these two discernible objects on a single plane, such as - the distance between the centers of mass of these two objects, the largest distances between their perimeters, the difference between their perimeters on the line defined by the two centers of mass, whether a center of mass of one is included within the perimeter of the other, the distance of the center of mass of one from the perimeter of the other, and so on. Once two or more discernible objects of different layers are determined to belong to the same 3D object, the modeling may include generating a model of a 3D object corresponding to these two or more discernible objects, including information pertaining to the 3D object as a whole, and potentially also to the individual original 2D discernible objects.

[0117] Optional substep 463 includes filtering out 3D objects based on their 3D parameters, such as 3D shape (e.g., elongated 3D shapes may be filtered out), 3D size, 3D orientation, and so on.

[0118] Substep 464 includes assessing 3D loneliness of such 3D objects with respect to other 3D objects and / or to discernible objects of one or more CT slices. It is noted that 3D equivalents of any of the loneliness criteria discussed with respect to step 450 may be implemented as the loneliness criteria of step 460, mutatis mutandis. Any variation or nuance discussed above with respect to the loneliness criteria may be applied, mutatis mutandis, to the loneliness assessment of step 460, and especially to substep 464.

[0119] The following discussion provides an example of an optional way of assessing 3D loneliness of a modeled 3D object as part of substep 464. According to theproposed example, a distance D in 3D space between two modeled 3D objects A andB might be defined as follows:

[0120] Where M(A) is the set of all points on the outer surface of 3D object A (e.g., at a pixel level, at equal intervals, etc.), Mi(A) is an ithpoint selected from the set M(A), fV(B) is the set of all points on the surface of the 3D object B, and fVj(B) is the jthpoint selected from the setFor example, is the distancebetween point number 3 of the set M(A) and point number 5 of the set 7V(B).

[0121] Based on the aforementioned definition, a metric for a minimal distance 6 is defined as the shortest distance of all possible distances in a pair of points, where one point lies on the surface of object A, and the other point lies on the surface of object B. Since calculating 6 directly is very time consuming, thefollowing algorithm demonstrates an optional way of calculating an estimate D of the value D.

[0122] Let D be the distance between (a) trace SpAof 3D object A on slice P out of a plurality of slices in which 3D object resides, and (b) trace of SqBof 3D object B on its slice Q of that plurality of slices. If KAAis the number of slices on which 3D object A left a trace and KBBis the number of slices on which 3D object B left a trace, then

[0123] An example of an optional algorithm for calculating the quantityIs discussed with respect to Fig. 6, which illustrates two modeled 3D objects: first 3D object 131 which left traces on three layers the traces being denoted132, 133, and 134, respectively), and second object 136 which also left traces on three layers the trace onbeing denoted 137). These sets of layers may beoverlapping, partly overlapping, or totally different. The proposed algorithm includes the following steps: a. For each slice SpAwhich includes a part of first 3D object 131, determine a center of the trace of first object 131 on the respective slice. The center may be center of mass of the respective trace, but other forms of determining center may also be used (e.g., in order to reduce computation load). Therespective trace (e.g., 132, 133, 134) may be the exact trace of the corresponding discernible object on that slice, or a somewhat different trace (e.g., if smoothing of the 3D object was applied). For example, center 135 in the diagram is the center of mass of slice S3A).For example, the (x,y) coordinates on the Pthslice (XAP, YAP) may be determined by the formulas:b. For each such center (e.g., 135), the nearest trace point SQBis searched for in any of the slices which include a part of second 3D object 136. This single point is denoted 138 in the diagram, on trace 137 of slice S1A. c. Next, the closest point to point 138 on any of the traces of first object 131 in all slices SQAis found. This point is denoted 139 in the diagram.

[0124] All of the distances in 3D space (as well as the distances within 2D) between any two points may be computed using standard Euclidean geometry, e.g.,

[0125] The coordinates of XA, XB, YA, YB, ZA, ZB are easy to calculate, because the pixel size is known, and the distance between slices (slice spacing) is known.

[0126] Substep 465 includes filtering out 3D objects based on their 3D loneliness status. Substep 465 may optionally be performed in the 3D level (e.g., if method 400 switched for completely 3D assessment of nodules in stage 460). Substep 465 may optionally include removing 2D candidates from the group of remaining candidates (at the end of step 450).

[0127] It is noted that steps 450 and 460 may be performed sequentially, with step 460 starting after stage 450 concluded, and using its output. Optionally, however, method 400 may include executing steps 450 and 460 in some form of combined manner. For example, after executing both 2D and 3D filtering of a first slice, the outputs may be used to execute step 450 for another slice (e.g., filtering out some of the potential candidate of that other slice based on the 3D processing of the first slice).

[0128] It is noted that implementations of method 400 in which step 460 is carried out only after the significant narrowing down of the group of discernible object into candidates and then the further narrowing of the group of candidates based on at least the 2D isolation requirement in stage 450 may require significantly less computations when compared to methods for detecting suspicious nodules by applying 3D geometrical analysis of CT images (whether other versions of method 400 or other methods such as some prior art methods), without doing the preliminary isolation filtering as suggested above.

[0129] Following step 460, method 400 continues with step 470 of providing a list of at least one suspected nodule, based on the outputs of the cross-layer geometric analysis. Referring to the examples set forth with respect to other diagrams of this disclosure, step 470 may optionally be executed by processor 220, e.g., by cancer evaluation module 228. For example, step 470 may include providing a list of at least one suspected nodule, each of which corresponds to any remaining candidate and / or modeled 3D object following the execution of steps 450 and 460. Optionally, step 470 may include providing a list of at least one suspected nodule as suspected early-stage lung cancer.

[0130] Fig. 7 provides an example of a CT image of a cross-section of lungs of a person (denoted 901), and an enlarged image 903 of part of image 901 (the part being denoted 902). A few of the candidates are denoted 905 in the diagram.

[0131] Fig. 8 is a schematic transverse section through a thorax of a person, viewed from above. The diagram illustrates different parts of the human body which may be imaged in a static medical imaging process such as CT or MRI. The diagram shows: a. The left lung 910 and the right lung 918. b. Parietal pleura 911, which is the outer membrane that lines the inner surface of the thoracic cavity, diaphragm, and mediastinum. c. Visceral pleura 912, which is an inner membrane that is intimately related to the outer wall of the lung as it directly covers and adheres to the lung'ssurface. The visceral pleura provides a protective covering and helps in the lung's expansion and contraction. d. Thorax cavity 913, which is the space within the chest that contains the lungs, heart, and other thoracic organs. e. Pericardial membranes 914: These membranes are a complex structure that surrounds and protects the heart within the pericardial cavity. They consist of the visceral layer that closely adheres to the heart muscle and the parietal layer lining the inner surface of the fibrous pericardium. f. Sternum 915, the flat bone located in the center of the chest, connecting to the rib bones and supporting the clavicles. g. Esophagus 916, the muscular tube connecting the pharynx to the stomach, facilitating the movement of swallowed food and liquids. h. Left pulmonary artery 917, A branch of pulmonary trunk 919, responsible for transporting deoxygenated blood from the heart to the left lung. i. Pulmonary trunk 919, the large blood vessel that splits into the left and right pulmonary arteries, carrying deoxygenated blood from the heart to the lungs. j. Thoracic wall 920, the boundary of the thorax, including the rib cage, muscles, and skin, enclosing and protecting the organs within.

[0132] As illustrated schematically in Fig. 8, the diameter of blood vessels in the lungs (as is the case in other parts of the body) change drastically. For simplicity of illustration, only a very cursory mapping of the blood vessels was offered in Fig. 8.

[0133] Fig. 9 is a schematic diagram of different types of blood vessels in the lungs. The diagram illustrates six different types of blood vessels in the lungs, starting from the main pulmonary artery down to the minuscule pulmonary capillaries. It should be noted that the illustration is not intended to exhaust all types of blood vessels in the lungs, but rather to illustrate the different environments in which a cancerous nodule 140 might reside, and help illustrate how such nodule 140 would be imaged in the cross-section imaging over the different backgrounds of different blood vessels environments. The following table illustrates the different diameters of the differenttypes of blood vessels in Fig. 9. It should be noted that the diagram itself is not to scale, and that the widths of the different types of blood vessels are not illustrated using a linear scale.

[0134] Considering the function of the lungs (breathing and exchange of gases between the respiratory system and the blood system), it is noted that the pulmonary capillaries are intricately related to the alveoli and in both structure and function. The alveoli are tiny air sacs clustered at the end of the bronchial tree, where oxygen enters the lungs and carbon dioxide is expelled. Pulmonary capillaries are the smallest blood vessels that intimately surround the alveoli. Their extremely thin walls are closely apposed to the alveolar walls, allowing for efficient gas exchange. Oxygen from the alveoli diffuses into the blood in the capillaries, while carbon dioxide from the blood diffuses into the alveoli to be exhaled. This relationship between the alveoli and pulmonary capillaries is central to the process of respiration, enabling the exchange of gases between the blood and the external environment. Since the alveoli make up the bulk of the lung tissue, they extend to the periphery of the lungs, close to the outer wall, known as the visceral pleura. The visceral pleura is a thin membrane that tightly covers the surface of the lungs, including the alveoli.

[0135] As exemplified in the diagram, next to the outer wall of the lungs (not shown) the blood vessels are very small. Especially, these blood vessels are much smaller than the size of early-state lung cancer nodules whose presence is sought in many implementations of the systems, methods, and computer program products disclosed here. Therefore, if a discernible object is detected in that area, it is relatively safe to assume that it is not a blood vessel, and may therefore be considered a nodule under a relatively relaxed isolation requirement. Considering, by way of example, a discernible object detected further away from the outer wall of the lung, e.g., around subsegmental arteries whose diameter is between 1-5 mm. Many such discernible objects would turn out to result from imaging of blood vessels (e.g., in CT imaging). The structure of the vascular system in the lungs, and especially the degree of branching, means that discernible objects which result from imaging of blood vessels are often grouped together, even in the same image. Therefore, stricter isolation requirements may be implemented in such areas, requiring that a given candidatewould be remote from other discernible object, in order to reduce the likelihood of erroneously reporting a blood vessel as a potential nodule. It should be noted that while Fig. 9 was discussed with respect to the lungs, similar structural aspects of the vascular system (or other circulatory systems like the lymphatic system in MRI imaging) are also applicable to other parts of the body, and to different types of cancer.

[0136] Fig. 10 illustrates a flow chart illustrating an example of method 500, in accordance with the presently disclosed subject matter. Method 500 is a computer implemented method for detecting lung cancer nodules based on processing of static imaging data, such as CT images or MRI images. Optionally, method 500 may be implemented on a general computer or processor. Alternatively, method 500 may be executed using dedicated hardware, such as an ASIC or specialized architecture. As CT scans and MRI scans become more affordable and thus more common, and as cancer prevalence grows, there is a growing incentive for mass implementation of method 500 for processing growing amount of static medical imaging scans done with the intention of detecting cancer (either as a primary goal of the static medical imaging, or as secondary processing of static medical imaging scans taken for other purposes). These trends also promote the development of dedicated hardware for efficient execution of method 500 if it is to be implemented in large multitude. Nevertheless, as stated above, method 500 may also be implemented without such dedicated hardware. It is noted that the discussion of method 500 commonly uses the example of processing CT data. However, as discussed above, method 500 is also applicable to MRI scans and to any other type of static medical imaging, mutatis mutandis.

[0137] In some ways, method 400 may be considered an example of method 500, as the further discussion of method 500 will show. Thus, many details discussed with respect to method 400 are not repeated with respect for method 500, in the interest of concision. It is noted that any step, substep, combination of several steps (or substeps), variation or discussion mentioned above with respect to method 400 may be implemented as part of method 500, mutatis mutandis, even if not explicitly stated. It is noted also that any step, substep, combination of several steps (or substeps),variation or discussion mentioned above with respect to method 500 may be implemented as part of method 400, mutatis mutandis, even if not explicitly stated.

[0138] Method 500 starts with step 510 of obtaining a group of candidate discernible objects within a static-imaging image stack that includes at least one image of a body part of a patient that includes at least part of a lung of the patient. Referring to the examples set forth with respect to the drawings, step 510 may optionally be executed by processor 220, e.g., by filtering module 222 or by isolation analysis module 224.

[0139] The static-imaging image stack may include all images taken during the imaging process (e.g., CT scan, MRI scan), but this is not necessarily so. The staticimaging image stack may include only adjacent images (of adjacent cross-sections as imaged by the imaging machine), but this is not necessarily so. For example, the staticimaging image stack may include between 1-5 images, between 5-15 images, between 15-40 images, between 30-70 images, between 70-150 images, between 150-500 slices, and so on. For example, a standard CT scan of lungs of an adult may include some 30-70 images ("slices"), each corresponding, for example, to a thickness of 1 to 10mm and a High-Resolution CT scan (HRCT) of lungs of an adult may include some 100-300 images, each corresponding to a thickness of 1 to 2mm.

[0140] The group of candidate discernible objects includes discernible objects having associated tissue-signal response patterns corresponding to cancerous nodules (e.g., cancerous nodules). The candidate discernible objects (also referred to as "candidates") may resemble cancerous nodules (or specific types of cancerous nodules, especially ones sought in method 500) in other ways in addition to their tissue-signal response patterns. Such other forms of similarity may include similar size, specific shapes, and so on. It is noted that step 510 may include any one or more of the following: capturing the images; receiving the images from an external entity; applying various algorithms in order detect discernible objects in the stack; preprocessing discernible objects detected in the stack (e.g., in order to improve their quality, their defining trace, etc.); filtering out some detection discernible objects, and so on. Many of such steps were discussed with respect to steps 410, 420, 430, and 440, and may be incorporated as part of step 510, mutatis mutandis.

[0141] Method 500 continues with executing for each out of multiple candidates out of the group of candidate discernible objects steps 520 and 530. These steps might be performed for any candidate of the group, but this is not necessarily so.

[0142] Step 520 includes determining whether the respective candidate discernible object satisfies an isolation requirement with respect to the rest of the candidate discernible objects, based at least on geometric remoteness of the respective candidate discernible object from the other candidate discernible objects. Optionally, step 520 may include determining whether the respective candidate discernible object satisfies an isolation requirement with respect to the rest of the candidate discernible objects and at least one other discernible object that is not a candidate, based at least on geometric remoteness of the respective candidate discernible object from the other candidates and from the at least one other discernible object. Referring to the examples set forth with respect to the drawings, step 520 may optionally be executed by processor 220, e.g., by isolation analysis module 224.

[0143] Step 520 may optionally include implementing step 450 and any one or more of its substeps and / or variations, mutatis mutandis. In some nonlimiting examples, step 520 may include any combination of one or more of the following: assessing the isolation of a candidate in response to maximal size of nodule sought (e.g., as discussed with respect to substep 451), assessing the isolation of a candidate in response to its distance from an outer wall of the lung (e.g., as discussed with respect to substep 452), assessing the isolation of a candidate in response to its inclusion status in a cluster of proximate peer objects that is remote from other peer objects in the group of peer objects (e.g., as discussed with respect to substep 453), and assessing the isolation of a candidate in response to its inclusion status in a cluster of at least N proximate peer objects (e.g., as discussed with respect to substep 454). While not necessarily so, such an isolation requirement may be implemented in an efficient way when including strictly deterministic criteria, without including stochastic processes or criteria that may introduce uncertainty, randomness, or chance).

[0144] It should be noted that different formats of output for the determining of step 520 may be implemented. Optionally, step 520 may conclude in any one of thefollowing nonlimiting examples: a binary output (e.g., "1" or "0", "isolated" or "not isolated"), a classified output (e.g., group i out of N predefined groups, such as "isolated", "not isolated", "borderline"), a natural number (e.g., an integer on a scale of 1-5, a natural number between 0-10). The different types of output may be utilized in corresponding different manners in the following steps of method 500 (and especially in step 530). It is noted that similarly varied outputs and utilization thereof may also be implemented in steps 450 and 460 of method 400, mutatis mutandis.

[0145] Step 530 includes conditionally reporting an existence of a suspected nodule corresponding to the respective candidate discernible object, in response to a result of the determining whether the respective candidate satisfies the isolation requirement in step 520. Optionally, step 530 may include conditionally reporting existence of suspected nodule at the lung corresponding to the respective candidate discernible object, in response to a result of the determining of stage 520 (as aforementioned) and further based on additional parameters or conditions. The reporting may include additional information regarding the suspected nodule, such as information pertaining to location, size, shape, and composition of the suspected nodule, snippets of one or more images corresponding to the suspected nodule, and so on. Referring to the examples set forth with respect to the drawings, step 530 may optionally be executed by processor 220, e.g., by cancer evaluation module 228.

[0146] The term "Conditionally" as used herein refers to the execution of a step or action based on the fulfillment of one or more predetermined conditions. For example, in relation to step 530, the reporting of an existence of a suspected nodule occurs if the determining of step 520 meets one or more specified condition. As discussed above, the conditional nature of this action does not exclude the possibility of other conditions influencing the execution of the step. In some cases, an "if and only if" condition may be applied in step 520 with respect to the result of the determining of step 520. Optionally, the condition may be implemented including strictly deterministic criteria, without including stochastic processes or criteria that may introduce uncertainty, randomness, or chance.

[0147] The execution of step 530 for the multiple candidates would result in withholding issuing false-positive reports pertaining to most of the multiple candidates, and— subject for determining that a certain candidate fulfills certain criteria, including meeting the isolation criteria — reporting one or more candidates as suspected nodules. Clearly, if no candidate meets the criteria (which might be expected in most healthy patients), step 530 would not include reporting of any candidate as suspected nodule.

[0148] In case one or more candidates do meet the reporting criteria, step 530 may include the substep 532 of determining whether the respective candidate should be reported, based at least on the result of the determining, and method 500 may include step 540 of reporting at least candidate discernible object that meets the isolation criteria as a suspected nodule, and step 550 of determining that one or more candidate discernible objects that do not meet the isolation criteria should not be reported as a suspected nodules. Candidates which are reported are also referred to as "first candidates", and candidates which are not reported are also referred to as "second candidates". Referring to the examples set forth with respect to the drawings, either one of steps 540 and 550 may optionally be executed by processor 220, e.g., by cancer evaluation module 228.

[0149] It is noted that step 530 might be preceded by processing image information of the respective slice (or any other required information) in order to obtain any information required in order to determine whether a given candidate should be reported as a suspected nodule or not.

[0150] Reverting to step 520, step 520 might optionally include applying a two- stages analysis of the isolation status of each respective candidate, first applying a single-slice isolation analysis (also referred to as "intra-slice isolation analysis"), and then applying a more taxing 3D cross-layer isolation analysis only to the candidates which pass an intra-slice isolation criteria. A long discussion of such step-by-step 2D- then-3D analysis was provided above, e.g., in the context of steps 450 and 460 of method 400, and all of it may apply to method 500, mutatis mutandis.

[0151] Optionally, method 500 may include as part of step 520:a. Determining whether the respective candidate discernible object meets an intra-slice isolation criteria with respect to the rest of the candidate discernible objects in the same image as the respective candidate discernible object (and optionally in further respect to other discernible objects which are not candidates), based at least on geometric remoteness of the respective candidate discernible object from the other candidate discernible objects (and possibly further based on geometric remoteness of the respective candidate discernible object from other discernible objects in the image); and b. Only if the respective candidate discernible objects meets the intra-slice isolation criteria, determining whether the respective candidate discernible object meets a 3D isolation criteria with respect to candidate discernible objects of at least one other image of the static-imaging image stack. As discussed above with respect to method 400, the 3D isolation-based analysis may optionally be preceded by a 3D pre-filtering stage (e.g., removing elongated modeled 3D objects from consideration).

[0152] Step 530 in such case may include selectively reporting an existence of a suspected nodule at the lung corresponding to the respective candidate discernible object, in response to determining that the respective candidate discernible object meets both the intra-slice isolation criteria and the inter-slice isolation criteria (and possibly also in response to additional criteria, e.g., as discussed above).

[0153] For example, method 500 may include analyzing two candidates of a single slice to determine whether any of the candidates should be reported as being associated with a suspected cancerous nodule. Optionally, both the first candidate (which is being reported) and the second candidate (which is not being reported) are isolated within their slice, and 3D isolation analysis is performed to both. In that example, the 3D analysis concludes with determining that the second candidate should not be reported (e.g., by applying any of the processing discussed with respect to step 460), while the same 3D analysis executed for the first candidate concludes with determining that the first candidate demonstrates both intra-slice and cross-sliceisolation, and can therefore be reported as suspect nodule (optionally, subject to additional criteria).

[0154] Reverting to step 520, it is noted that different isolation criteria may be used for different candidates. For example, method 500 may include implementing different isolation criteria for different candidates based on differences such as: candidate parameters (e.g., shape, size, shape, composition, tissue-signal response pattern), location parameters (e.g., defined with respect to the physiological environment in different parts of the lung), and so on. Optionally, method 500 may include optional step 515 of determining isolation requirements for step 520. The determining of step 515 may include determining different isolation requirements for different candidates (e.g., as discussed above), such as: for individual candidates, for different candidate groups, for different areas in the image, for different areas in the image stack, and so on. Optionally, step 515 may include determining isolation criteria based on any one or more of the following nonlimiting examples: a. The maximum size of nodule for which the algorithm is searching; b. Spatial relationship with respect to the outer wall of the lung; c. Scan parameters used by in the imaging process in which the image stack was created (e.g., operational parameters of the imaging machine); d. Information about the patient (e.g., age, gender, weight, BMI, years of smoking).

[0155] It is noted that while different criteria might be defined for different candidates, a single set of isolation criteria may nevertheless be manifested in different ways for different candidates. For example, the criteria may include a threshold which is defined as a function of a distance of the candidate from the outer wall of the lung, such that this same function is applied to the different candidates differently.

[0156] Fig. 11 includes four diagrams schematically illustrating different ways of implementing isolation requirement which is dependent on the spatial relationship with respect to the outer wall of the lung, in accordance with examples of thepresently disclosed subject matter. In the illustrated example, the spatial relationship is represented by a one-dimensional distance (e.g., a single-layer distance, a crosslayer distance) to the nearest part of the outer side of the wall, but other distance metrics may also be implemented.

[0157] Each of the four diagrams graphically illustrates an isolation criterion in which the isolation requirement includes a minimal distance threshold for the candidate with respect to other candidates (or possibly, other discernible objects) in order to be considered isolated: in diagram 151 the minimal threshold is a continuous monotonous function; in diagram 152 the minimal threshold is defined as a linear monotonous function until a certain distance, and then remains constant; in diagram 153 the minimal threshold is defined as a binary choice between two values based on the distance from the wall of the lung; and in diagram 154 the minimal threshold is defined as a combination of monotonously rising function near the wall of the lung, and as a series of constant values at greater distances.

[0158] Similarly to the discussion with respect to step 515, it is noted that method 500 may optionally include an optional step of determining reporting requirements for step 530. For example, such determination of reporting criteria may be based on any one or more of the following nonlimiting examples: a. The maximum size of nodule for which the algorithm is searching; b. Spatial relationship with respect to the outer wall of the lung; c. Scan parameters used by in the imaging process in which the image stack was created (e.g., operational parameters of the imaging machine); d. Information about the patient (e.g., age, gender, weight, BMI, years of smoking).

[0159] Fig. 12 illustrates assessing isolation of candidates at different distances from the outer wall of the lung, in accordance with examples of the presently disclosed subject matter. Diagram 171 illustrates a lung in which a plurality 172 of candidates is positioned. These candidates are colored black simply to ease the identification of these candidates in the diagram. When assessing the isolation of candidate 173, itsisolation requirement is correlated to the distance 175 between that candidate 173 and the outer wall of the lung. As discussed with respect to method 400 above, this distance may optionally be established as part of method 500, but method 500 may also be executed without explicitly calculating distance 175. The isolation requirement in diagram 171 is represented by circle 176 whose radius depends on distance 175. As can be seen, none of the other adjacent candidates 174 (collectively denoted with candidate 173 as group 172) is positioned within the threshold requirement represented by circle 176, and therefore candidate 173 is considered isolated from other candidates of the image.

[0160] Diagram 181 focus on another group of candidates (collectively denoted 182) having the exact same internal spatial relationship as group 172, and thus the distances of candidate 183 from its neighboring candidates 184 are identical to those of candidate 173. However, the distance 185 between candidate 183 and the outer wall of the lung is greater than distance 175, and therefore the minimal required threshold for isolation (represented by the radius of circle 186) is also significantly larger. As can be seen, there are adjacent candidates 184 which are positioned within the threshold requirement represented by circle 186, and therefore - despite having the same group spatial relationship as candidate 173 with respect to group 175, candidate 183 is not considered isolated. For example, group 182 may represent the footprints of a branched blood vessel, e.g., as discussed above.

[0161] While Fig. 12 illustrated the difference in isolation requirement for two groups of similar internal geometry, the difference in isolation requirement used for candidates having different distances to the outer wall of the lung may be applied in additional manners. For example, method 500 may optionally be applied to a group of candidate discernible objects which includes a first candidate discernible object that is closer to the wall of the lung than a second candidate discernible object, when a distance between the second candidate discernible object to its nearest candidate discernible object of the group of candidate discernible objects is at least twice a distance between the first candidate discernible object to its nearest candidate discernible object of the group of candidate discernible objects. Nevertheless, the difference in isolation requirement may result in the first candidate discernible objectbeing reported as suspected nodule, while the second candidate is not reported (e.g., considered a false positive).

[0162] Referring to the distinction made in the discussion of steps 450 and 460 of method 400 between single-layer isolation requirement and cross-section isolation requirement, it is possible (in at least some versions of method 500) that for a candidate of a first slice there will be a candidate of an adjacent slice with very close 3D distance between the two, but the first candidate will nevertheless be considered isolated in single-layer analysis (and possibly also in cross-layer analysis, e.g., if the inlayer translation between their two projections means that they are not modeled as a single 3D object).

[0163] Optionally, method 500 may include a variation of substep 453 of method 400, which includes assessing the isolation of a candidate in response to its inclusion status in a cluster of proximate peer objects that is remote from other peer objects in the group of peer objects (e.g., as exemplified in relation to Fig. 4). Such a step might be implemented, for example, to overcome false positive errors resulting from erroneous misidentification of branching blood vessels.

[0164] Optionally, the isolation criteria used in method 500 might prevent an analyzed candidate discernible object from being reported as a suspected nodule if there are at least P neighboring candidates of the group of candidates located within a circle of a given threshold radius which includes the analyzed candidate discernible object. The same given threshold radius may be implemented uniformly to all candidates in a certain slice, or different threshold radii might be used (e.g., in different parts of the lung). In this paragraph, P is a predetermined number that is equal to or greater than 2. The same isolation criteria would permit an analyzed candidate discernible object to be reported as a suspected nodule if there less than P neighboring candidate discernible object of the group of candidate discernible objects (e.g., is up to one for P=2) within a circle of the given radius. That circle, in such case, would include only the analyzed candidate discernible object and the neighboring candidate discernible object.

[0165] Optionally, method 500 may include a variation of optional substep 454 of method 400, which includes assessing the isolation of a candidate in response to its inclusion status in a cluster of at least N proximate peer objects. Optionally, the cluster may include only candidates (e.g., as exemplified in relation to Fig. 5). For example, N might be equal to any natural number between 5 and 12. Such a step might be implemented, for example, to overcome false positive errors resulting from erroneous misidentification of another phenomenon in a case which is unlikely to represent a cancerous nodule.

[0166] Optionally, the isolation criteria used in method 500 (especially in step 520) might be one that prevents an analyzed candidate discernible object from being reported as a suspected nodule if it belongs to a subgroup of N>5 candidate discernible objects selected from the group of candidate discernible objects. Such a group might be defined, for example, as a group in which each of the candidates complies with a rule that at least three other candidate discernible objects of the subgroup are positioned within a sphere of a predetermined radius Rocentered on the respective candidate discernible object. Any other detail discussed with respect to step 545 and / or to Fig. 5 might be implemented for method 500, mutatis mutandis.

[0167] Pertaining to both method 400 and method 500, it is noted that different steps of filtering might be applied to discernible objects, candidates, and / or modeled 3D objects (extrapolated from a plurality of candidates, possibly including other discernible objects) at different points during the execution of these options. Several examples of such steps of filtering were discussed above.

[0168] Method 500 may optionally include filtering any type of filterable entity (including any combination of one or more of: discernible objects, candidates, and / or modeled 3D objects) based on any one or more of the following types of filters: a. Noise based filter (e.g., filtering out filterable entities which are insufficiently differentiable from noise, e.g., based on their tissue-signal response patterns, size, or combination thereof); b. Tissue-signal response patterns (e.g., filtering out filterable entities whose tissue-signal response values deviate from a predefined range, filterableentities whose intra-entity variation in tissue-signal response values exceeds a predefined threshold, filterable entities whose tissue-signal response values create specific spatial patterns or textures); c. Size (e.g., filtering out filterable entities having an effective diameter which is smaller than a predefined threshold value, such as 4mm); d. Shape (e.g., filtering out filterable entities which are overly elongated, for example because such entities may be a depiction of blood vessels); e. 3D continuity (e.g., filtering out filterable entities which are continues over a larger number of slices, such as more than 5-6 slices, for example because such entities may be a depiction of blood vessels); f. Translation between slices (e.g., filtering out filterable entities which demonstrate lateral translation between adjacent slices beyond a specific lateral distance and / or beyond a threshold number of slices, for example because such entities may be a depiction of blood vessels);

[0169] Optionally, method 500 may include implementing combined filtering of discernible objects, candidates, or modeled 3D objects (or any combination of two or more of the above) based on two or more of the above identified filters. For example, the combined filtering may include implementing filtering based on any of the following combinations of aforementioned filtering types: a. Combined filtering based on both (a) size, (b) shape, and (c) tissue-signal response patterns (e.g., CT gray level or MRI gray level). b. Combined filtering based on both (a) size, (b) shape, and (c) 3D continuity. c. Combined filtering based on both (a) size, (b) shape, and (c) translation between slices. d. Combined filtering based on both (a) size, (b) shape, (c) 3D continuity, and (d) translation between slices. e. Combined filtering based on both (a) size, (b) shape, and (c) noise levels.f. Combined filtering based on both (a) shape, (b) 3D continuity, and (c) noise levels. g. Combined filtering based on both (a) shape, (b) translation between slices, and (c) noise levels. h. Combined filtering based on both (a) shape, (b) 3D continuity, (c) translation between slices, and (d) noise levels. i. Combined filtering based on both (a) shape, (b) tissue-signal response patterns, and (c) 3D continuity. j. Combined filtering based on both (a) shape, (b) tissue-signal response patterns, and (c) translation between slices. k. Combined filtering based on both (a) shape, (b) tissue-signal response patterns, (c) 3D continuity, and (d) translation between slices.

[0170] Pertaining to both method 400 and to method 500, it is noted that optionally, different types of filters and / or different filtering parameters may be applied to different subsets of filterable entities (e.g., based on characteristics of the respective entities, of their spatial locations within the body or within the cross-section image, and so on). Several non-limiting examples were provided above.

[0171] Referring to both of method 400 and method 500, it is noted that these methods may be applied to static medical imaging product which are captured by a corresponding machine (e.g., CT machine, MRI machine), and includes imaging data of the body of an actual person (or animal, in some cases). However, these methods may also be applied for static medical imaging products which were generated by machines, or which were manipulated by machines. For example, method 400 and / or method 500 may be applied to synthetic CT images which were generated by a computer (e.g., based on a source CT data which was manipulated, using artificial intelligence or in any other way). In another example, method 400 and / or method 500 may be applied to synthetic MRI images which were generated by a computer (e.g., based on a source MRI data which was manipulated, using artificial intelligence or in any other way). Applying these methods to such synthetic images may beimplemented for different reasons, such as to train machine learning system (e.g., create truth values for a training database) or to train radiologist (e.g., creating analyzed reference data and images for rare conditions which are not commonly encountered in the everyday practice of most radiologists).

[0172] Optionally, method 500 (and likewise method 400) may further include repeating steps 510, 520, and 530 for a second group of candidate discernible objects within another static-imaging image stack that includes at least one other image (e.g., both the first image stack and the second image stacks including pulmonary CT or MRI images of the lungs). Both groups of candidate discernible objects include a subspace which consist of a subgroup of candidate discernible objects that includes a subject discernible object. For example, the areas 172 and 182 in Fig. 12 may be considered such subspaces (even though not both have to appear in the same image stack in the present case). The subspace may be a 2D subspace or a 3D subspace. For example, the size of the subspace may be l-4cm2(2D case) or l-20cm3(3D case). It is noted that obtaining image stacks with the same subspace consisting of a specific subgroup of candidates in a certain spatial relationship is very easily achievable when creating synthetic imaging. It is noted that both image stacks may include synthetic images or only one of them may include synthetic images generated by manipulating images of the other stack (other options are also possible).

[0173] Continuing the sample example, a first candidate of the first image stack which is reported as a suspected cancerous nodule may be the subject discernible object in the first image stack, and another candidate (referred to as a "third candidate discernible object") of the second group of candidate discernible objects may be the subject discernible object of the second image stack. Considering such two stacks of sectional images, some interesting scenarios might occur, e.g., when implementing isolation criteria which is responsive to the distances of the different candidates from the outer wall of the lung (also referred to as "pleural distance").

[0174] For example, optionally a pleural distance of the third candidate may be much larger than the pleural distance of the first candidate (e.g., at least twice larger), but nevertheless method 500 would include reporting the first candidate as a suspectcancerous nodule and determining that the third candidate discernible object does not meet the isolation criteria and should not be reported as a suspected nodule. Understanding of this scenario might be facilitated when considered in view of the somewhat different example of Fig. 12.

[0175] Figs. 13A and 13B are functional block diagrams illustrating an example of system 200, in accordance with the presently disclosed subject matter. System 200 is a system for detecting a lung cancer nodule by processing static imaging data. While not necessarily so, system 200 may be operable to execute method 400 and / or method 500, or any combination of any one or more steps from one or both of methods 400 and 500. Any detail discussed with respect to method 400 and / or to method 500 might be performed by system 200 (e.g., by the least one tangible memory module 210 or by processor 220), mutatis mutandis. Especially, processor 220 may optionally be configured and operable to execute any processing step, processing substep, combination of several processing steps (or processing substeps) of method 400 and / or of method 500. It is noted that any step, substep, combination of several steps (or substeps), variation or discussion provided above with respect to method 400 and / or to method 500 may be implemented as part of system 200, mutatis mutandis, even if not explicitly stated, and that many such details are not repeated with respect to system 200 in the interest of concision.

[0176] System 200 includes at least one tangible memory module 210 that is operable to store information of a group of CDOs within a static-imaging image stack that includes at least one image of a part of the body of the patient that includes at least a part of the lung. The group of CDOs includes discernible objects having associated tissue-signal response patterns corresponding to cancerous nodules. It is noted that the at least one tangible memory module 210 may optionally store the entire static-imaging image stack or part of it (e.g., image clippings corresponding to the CDOs), but this is not necessarily so. Memory module 210 also stores computer program code which stores instructions, which processor 220 can execute for carrying out any combination of one or more of the functionalities discussed below with respect to processor 200, and / or any one of the steps and / or functionalities discussed above with respect to any one of methods 400 and 500. While any one or more of thefunctionalities of processor 220 may be implemented using dedicated circuitry designed specifically to execute one or more steps or functionalities in the hardware and / or firmware level, other functionalities may be implemented by circuitry which is adapted to execute different instructions stored as a computer program code on the at least one memory module 210. Throughout the disclosure, whenever processor 220 is described as having functionality or capability, that functionality may be implemented using any combination of hardware (dedicated or not), firmware (dedicated or not), and software (including computer program code stored on the at least one memory module 210).

[0177] System 200 also includes at least one processor 220, that are collectively operable to: a. access the information of the group of CDOs stored on the at least one tangible memory module; b. determine for each CDOs out of multiple CDOs of the group of CDOs whether the respective CDO satisfies an isolation requirement with respect to the rest of the CDOs, based at least on geometric remoteness of the respective CDO from the other CDOs; c. based on determining that at least one first CDO out of the multiple CDOs meets the isolation criteria, reporting the first CDO as a suspected nodule; and d. based on determining that at least one second CDO out of the multiple CDOs does not meet the isolation criteria, refraining from reporting the second CDO as a suspected nodule

[0178] The at least one processor 220 may optionally be configured to conditionally report an existence of a suspected nodule at the lung corresponding to the respective CDO, in response to a result of the determining for each respective CDO for which such a determining is performed by the at least one processor 220.

[0179] Optionally, system 200 may include a communication module by which the at least one processor 220 reports the existence of suspected nodules (and optionallyprovide additional data). Communication module 230 may optionally be connected to another component of system 200 (e.g., a user interface component such as a computer screen, a speaker, a printer, and so on) and / or to an external system (e.g., a remote computer, a television, a smartphone, and so on). Communication module 230 may be a standard communication module (e.g., Ethernet Module, Wi-Fi Module, Bluetooth Module, USB Communication Module, Serial Communication Module) or a dedicated communication module. Communication module may implement standard communication protocols (e.g., Transmission Control Protocol (TCP), Internet Protocol (IP), Hypertext Transfer Protocol (HTTP), File Transfer Protocol (FTP), Wireless Application Protocol (WAP)) and / or dedicated communication protocol. Communication module 230 may implement wired and / or wireless communication.

[0180] Pertaining to the communication between the at least one processor 220 and the at least one memory module 210, and / or between the at least one processor 220 and communication module 230: the respective communication channel may be wired or wireless and implement standard or dedicated protocols. The at least one processor 220 and the at least one memory module 210 (or communication module 230, respectively) may be located in varying proximity to each other, such as in a single computer, the same building, or at a remote location, including modules of remote servers. A few non-limiting examples of standard communication protocols which may be implemented in such a communication channels include: Double Data Rate (DDR), QuickPath Interconnect (QPI), Transmission Control Protocol / lnternet Protocol (TCP / IP), File Transfer Protocol (FTP), Hypertext Transfer Protocol / Secure (HTTP / HTTPS), and System Management Bus (SMBus). A few non-limiting examples of interfaces facilitating communication within system 200 include: Peripheral Component Interconnect Express (PCIe), Serial ATA (SATA), Universal Serial Bus (USB), and HyperTransport (HT).

[0181] Optionally, the isolation criteria may be further dependent upon a distance of the respective CDO from an outer wall of the lung.

[0182] Optionally, the at least one processor 220 may be configured and operable to determine the isolation criteria. Optionally, the at least one processor 220 may beconfigured and operable to determine the isolation criteria to dynamically modify the isolation criteria based on various parameters and / or considerations (e.g., as discussed with respect to method 400 and to method 500 above). Optionally, the at least one processor 220 may implement different isolation criteria to different candidate discernible objects (e.g., as discussed with respect to method 400 and to method 500 above). Optionally, the at least one processor 220 may be configured and operable to determine different isolation criteria to different candidate discernible objects.

[0183] Optionally, a specific first CDO is closer to the wall of the lung than a specific second CDO, and a distance between the a specific second CDO to its nearest CDO of the group of CDOs is at least twice a distance between the a specific first CDO to its nearest CDO of the group of CDOs.

[0184] Optionally, at least one tangible memory module 210 may be further operable to store second information of a second group of CDOs within a second static-imaging image stack that includes at least one second pulmonary image, the second group of CDOs includes discernible objects having associated tissue-signal response patterns corresponding to cancerous nodules, and the at least one processor 220 of that system, in such case, may be operable to access the second information stored on at least one tangible memory module 210 and to process the second information like the information of the first stack. In a case in which each of the group of CDOs and the second group of CDOs includes a subspace which consist of a subgroup of CDOs that includes a subject discernible object (a first CDO of the first stack is the subject discernible object in the image stack, and a third CDO of the second group of CDOs is the subject discernible object of the other image stack, such that a pleural distance of the third CDO is at least twice a pleural distance of the first CDO), the at least one processor 230 may be configured and operable to determine that the third CDO does not meet the isolation criteria and subsequently to refrain from reporting the third CDO as a suspected nodule.

[0185] Optionally, the at least one processor 220 may be is operable to report analyzed CDOs as suspected nodules only for analyzed CDOs for which there is up toone neighboring CDO of the group of CDOs within a circle of a given radius which includes only the analyzed CDO and the neighboring CDO (but not necessarily for every analyzed CDO which meat that criterion). The at least one processor 220 in such case may be further operable to refrain from reporting any analyzed CDO as a suspected nodule if there are at least two neighboring CDO of the group of CDOs within a circle of the given radius which includes the analyzed CDO.

[0186] Optionally, the at least one processor 220 may be configured and operable to refrain from reporting any analyzed CDO from being reported as a suspected nodule if the respective analyzed CDO belongs to a subgroup of N>5 CDOs of the group of CDOs, each of which complying with the rule that at least three other CDOs of the subgroup are positioned within a circle of a predetermined radius Rocentered on the respective CDO.

[0187] Optionally, the at least one processor 220 may be configured and operable to determine whether an analyzed CDO satisfies an isolation requirement with respect to the rest of the CDOs by initially determining whether the respective CDO meets an intra-slice isolation criteria with respect to the rest of the CDOs in the same image as the respective CDO, based at least on geometric remoteness of the respective CDO from the other CDOs; and only if the respective CDOs meets the intra-slice isolation criteria, determining whether the respective CDO meets a 3D isolation criteria with respect to CDOs of at least one other image of the static-imaging image stack. The at least one processor 220 in such case may be operable to refrain from reporting an existence of a suspected nodule at the lung corresponding to any analyzed respective CDO that do not meet both the intra-slice isolation criteria and the inter-slice isolation criteria.

[0188] Optionally, one or more of the at least one first CDO (being reported as a suspect cancerous nodule) may be a small cancerous nodule having an effective diameter of between 0.5-1.5mm which is positioned less than 5cm from an outer wall of the lung, and the at least one processor 220 is operable to implement an adaptable isolation threshold that depends on the distance between an analyzed CDO and the outer wall of the lung for detecting the small cancerous nodule.

[0189] Optionally, the at least one processor 220 may be configured and operable to perform a series by rule-based logic developed through human-engineered algorithms, and excludes employment of data-trained models for detecting the suspect cancerous nodules.

[0190] Referring to the example of Fig. 13B, it is noted that the at least one memory module 210 (represented as a single memory module 210, for the sake of simplicity) may store any information required for the detection of suspect cancerous nodules, as described with respect to system 200, to method 400, and to method 500. For example, the at least one memory module 210 may optionally include any one or more of the following: Image stack data, candidate discernible objects data, data of other discernible objects, isolation requirement data, operational parameters of the system, dedicated code (e.g., instructions required for execution of the processing by the at least one processor 220, e.g., instructions for executing any one or more steps of method 400 and / or method 500).

[0191] Referring to the example of Fig. 13B, it is noted that the at least one processor 220 (represented as a single processor 220, for the sake of simplicity) may optionally implement multiple processing modules, which may be configured to operate on the outputs of one another, communicate with memory modules 210, with communication module 230, and so on. For example, the at least one processor 220 may Implement any one or more of the following modules: discernible objects filtering module 222, isolation analysis module 224, clustering analysis module 226, and cancer evaluation module 228.

[0192] Referring to method 400, it is noted that a computer program product for detecting a lung cancer nodule by processing static imaging data is disclosed, including instructions which, when executed by a computer (which in this context may optionally be implemented by one or more computers), cause that computer to execute the steps of method 400. A computer program is also disclosed, which makes a computer execute method 400.

[0193] Referring to method 500, it is noted that a computer program product for detecting a lung cancer nodule by processing static imaging data is disclosed, includinginstructions which, when executed by a computer (which in this context may optionally be implemented by one or more computers), cause that computer to execute the steps of method 500. A computer program is also disclosed, which makes a computer execute method 500.

[0194] A designated computer program product for detecting a lung cancer nodule by processing static imaging data including instructions which, when executed by a computer, cause the computer to execute: a. obtaining a group of candidate discernible objects (CDOs) within a staticimaging image stack that includes at least one image of a lung, the group of CDOs includes discernible objects having associated tissue-signal response patterns corresponding to cancerous nodules; and b. executing for each out of multiple CDOs out of the group of CDOs: (a) determining whether the respective CDO satisfies an isolation requirement with respect to the rest of the CDOs, based at least on geometric remoteness of the respective CDO from the other CDOs; and (b) conditionally reporting an existence of a suspected nodule at the lung corresponding to the respective CDO, in response to a result of the determining. Optionally, the executing may include at least: (a) reporting a first CDO that meets the isolation criteria as a suspected nodule; and (b) determining that a second CDO that does not meet the isolation criteria should not be reported as a suspected nodule.

[0195] Likewise, a designated computer-readable storage medium for detecting a lung cancer nodule by processing static imaging data is disclosed, including instructions which, when executed by a computer, cause the computer to execute: a. obtaining a group of candidate discernible objects (CDOs) within a staticimaging image stack that includes at least one image of a lung, the group of CDOs includes discernible objects having associated tissue-signal response patterns corresponding to cancerous nodules; andb. executing for each out of multiple CDOs out of the group of CDOs: (a) determining whether the respective CDO satisfies an isolation requirement with respect to the rest of the CDOs, based at least on geometric remoteness of the respective CDO from the other CDOs; and (b) conditionally reporting an existence of a suspected nodule at the lung corresponding to the respective CDO, in response to a result of the determining. Optionally, that executing may include at least: (a) reporting a first CDO that meets the isolation criteria as a suspected nodule; and (b) determining that a second CDO that does not meet the isolation criteria should not be reported as a suspected nodule.

[0196] Pertaining to the designated computer program product and the designated computer-readable storage medium, any one of them may be implemented such that the isolation criteria is further dependent upon a distance of the respective CDO from an outer wall of the lung.

[0197] Pertaining to the designated computer program product and the designated computer-readable storage medium, any one of them may be implemented such that the first CDO is closer to the wall of the lung than the second CDO, and a distance between the second CDO to its nearest CDO of the group of CDOs is at least twice a distance between the first CDO to its nearest CDO of the group of CDOs.

[0198] Pertaining to the designated computer program product and the designated computer-readable storage medium, any one of them may further include instructions which, when executed by the computer, cause the computer to: (a) repeat the steps of obtaining and executing for a second group of CDOs within another static-imaging image stack that includes at least one pulmonary image. In a case in which each of the group of CDOs and the second group of CDOs includes a subspace which consist of a subgroup of CDOs that includes a subject discernible object (a first CDO of the first stack is the subject discernible object in the image stack, and a third CDO of the second group of CDOs is the subject discernible object of the other image stack, such that a pleural distance of the third CDO is at least twice a pleural distance of the first CDO), the instructions may optionally be such that cause the computer to determine thatthe third CDO does not meet the isolation criteria and subsequently to refrain from reporting the third CDO as a suspected nodule.

[0199] Pertaining to the designated computer program product and the designated computer-readable storage medium, any one of them may be implemented such that the isolation criteria permits an analyzed CDO to be reported as a suspected nodule if there is up to one neighboring CDO of the group of CDOs within a circle of a given radius which includes only the analyzed CDO and the neighboring CDO; and such that the isolation criteria prevents an analyzed CDO from being reported as a suspected nodule if there are at least two neighboring CDO of the group of CDOs within a circle of the given radius which includes the analyzed CDO.

[0200] Pertaining to the designated computer program product and the designated computer-readable storage medium, any one of them may be implemented such that the isolation criteria prevents an analyzed CDO from being reported as a suspected nodule if the respective analyzed CDO belongs to a subgroup of N>5 CDOs of the group of CDOs, each of which complying with the rule that at least three other CDOs of the subgroup are positioned within a circle of a predetermined radius Rocentered on the respective CDO.

[0201] Pertaining to the designated computer program product and the designated computer-readable storage medium, any one of them may further include instructions which, when executed by the computer, cause the computer to: (a) determine whether the respective CDO meets an intra-slice isolation criteria with respect to the rest of the CDOs in the same image as the respective CDO, based at least on geometric remoteness of the respective CDO from the other CDOs; (b) selectively determine, only if the respective CDOs meets the intra-slice isolation criteria, whether the respective CDO meets a 3D isolation criteria with respect to CDOs of at least one other image of the static-imaging image stack; and (c)selectively report an existence of a suspected nodule at the lung corresponding to the respective CDO, in response to determining that the respective CDO meets both the intra-slice isolation criteria and the inter-slice isolation criteria.

[0202] Each of the computer programs may be stored internally on a non-transitory computer readable medium. All or some of the computer program may be provided on computer readable media permanently, removably, or remotely coupled to an information processing system. The computer readable media may include, for example and without limitation, any number of the following: magnetic storage media including disk and tape storage media; optical storage media such as compact disk media (e.g., CD-ROM, CD-R, etc.) and digital video disk storage media; nonvolatile memory storage media including semiconductor-based memory units such as FLASH memory, EEPROM, EPROM, ROM; ferromagnetic digital memories; MRAM; volatile storage media including registers, buffers or caches, main memory, RAM, etc.

[0203] A computer process typically includes an executing (running) program or portion of a program, current program values and state information, and the resources used by the operating system to manage the execution of the process. An operating system (OS) is the software that manages the sharing of the resources of a computer and provides programmers with an interface used to access those resources. An operating system processes system data and user input, and responds by allocating and managing tasks and internal system resources as a service to users and programs of the system.

[0204] While the embodiments described above are provided as examples, it should be understood that various modifications and substitutions may be made without departing from the scope of the invention as defined in the appended claims.

Claims

CLAIMSWhat is claimed is:

1. A computer implemented method for detecting a lung cancer nodule by processing static imaging data, the method comprising: obtaining a group of candidate discernible objects (CDOs) within a staticimaging image stack that comprises at least one image of a lung, the group of CDOs comprises discernible objects having associated tissue-signal response patterns corresponding to cancerous nodules; and executing for each out of multiple CDOs out of the group of CDOs: determining whether the respective CDO satisfies an isolation requirement with respect to the rest of the CDOs, based at least on geometric remoteness of the respective CDO from the other CDOs; and conditionally reporting an existence of a suspected nodule at the lung corresponding to the respective CDO, in response to a result of the determining; wherein the executing comprises at least: reporting a first CDO that meets the isolation criteria as a suspected nodule; and determining that a second CDO that does not meet the isolation criteria should not be reported as a suspected nodule.

2. The method according to claim 1, wherein the isolation criteria is further dependent upon a distance of the respective CDO from an outer wall of the lung.

3. The method according to claim 2, wherein the first CDO is closer to the wall of the lung than the second CDO, wherein a distance between the second CDO to its nearest CDO of the group of CDOs is at least twice a distance between the first CDO to its nearest CDO of the group of CDOs.

4. The method according to claim 1, further comprising repeating the steps of obtaining and executing for a second group of CDOs within another static-imaging image stack that comprises at least one pulmonary image; wherein both the group of CDOs and the second group of CDOs comprise a subspace which consist of a subgroup of CDOs that includes a subject discernible object; wherein the first CDO is the subject discernible object in the image stack, and a third CDO of the second group of CDOs is the subject discernible object of the other image stack; wherein a pleural distance of the third CDO is at least twice a pleural distance of the first CDO; wherein the step of executing is carried out for the third CDO comprises determining that the third CDO does not meet the isolation criteria and should not be reported as a suspected nodule.

5. The method according to claim 1, wherein the isolation criteria permits an analyzed CDO to be reported as a suspected nodule if there is up to one neighboring CDO of the group of CDOs within a circle of a given radius which comprises only the analyzed CDO and the neighboring CDO; wherein the isolation criteria prevents an analyzed CDO from being reported as a suspected nodule if there are at least two neighboring CDO of the group of CDOs within a circle of the given radius which comprises the analyzed CDO.

6. The method according to claim 1, wherein the isolation criteria prevents an analyzed CDO from being reported as a suspected nodule if the respective analyzed CDO belongs to a subgroup of N>5 CDOs of the group of CDOs, each of which complying with the rule that at least three other CDOs of the subgroup are positioned within a circle of a predetermined radius Rocentered on the respective CDO.

7. The method according to claim 1, wherein the executing comprises: determining whether the respective CDO meets an intra-slice isolation criteria with respect to the rest of the CDOs in the same image as therespective CDO, based at least on geometric remoteness of the respective CDO from the other CDOs; only if the respective CDOs meets the intra-slice isolation criteria, determining whether the respective CDO meets a 3D isolation criteria with respect to CDOs of at least one other image of the static-imaging image stack; and selectively reporting an existence of a suspected nodule at the lung corresponding to the respective CDO, in response to determining that the respective CDO meets both the intra-slice isolation criteria and the inter-slice isolation criteria.

8. The method according to claim 1, wherein the first CDO is a small cancerous nodule having an effective diameter of between 0.5-1.5mm which is positioned less than 5cm from an outer wall of the lung, wherein the method comprises detecting the small cancerous nodule using an adaptable isolation threshold that depends on the distance between an analyzed CDO and the outer wall of the lung.

9. The method according to claim 1, wherein the method consists of detecting suspect cancerous nodules by performing a series by rule-based logic developed through human-engineered algorithms, and excludes employment of data- trained models.

10. A system for detecting a lung cancer nodule by processing static imaging data, the system comprising: at least one tangible memory module operable to store computer program code and information of a group of candidate discernible objects (CDOs) within a static-imaging image stack that comprises at least one image of a lung, the group of CDOs comprises discernible objects having associated tissue-signal response patterns corresponding to cancerous nodules; and at least one processor; wherein the at least one memory and the computer program code are configured, with the at least one processor, to cause the system to at least:access the information of the group of CDOs stored on the at least one tangible memory module; determine for each CDOs out of multiple CDOs of the group of CDOs whether the respective CDO satisfies an isolation requirement with respect to the rest of the CDOs, based at least on geometric remoteness of the respective CDO from the other CDOs; based on determining that at least one first CDO out of the multiple CDOs meets the isolation criteria, reporting the first CDO as a suspected nodule; and based on determining that at least one second CDO out of the multiple CDOs does not meet the isolation criteria, refraining from reporting the second CDO as a suspected nodule.

11. The system according to claim 10, wherein the isolation criteria is further dependent upon a distance of the respective CDO from an outer wall of the lung.

12. The system according to claim 11, wherein a specific first CDO is closer to the wall of the lung than a specific second CDO, wherein a distance between the a specific second CDO to its nearest CDO of the group of CDOs is at least twice a distance between the a specific first CDO to its nearest CDO of the group of CDOs.

13. The system according to claim 10, wherein the at least one tangible memory module is further operable to store second information of a second group of CDOs within a second static-imaging image stack that comprises at least one second pulmonary image, the second group of CDOs comprises discernible objects having associated tissue-signal response patterns corresponding to cancerous nodules; wherein the at least one memory and the computer program code are configured, with the at least one processor are operable to access the second information stored on the at least one tangible memory module and to process the second information like the information, wherein both the group of CDOs and the second group of CDOs comprise a subspace which consist of a subgroup of CDOs that includes a subjectdiscernible object; wherein the first CDO is the subject discernible object in the image stack, and a third CDO of the second group of CDOs is the subject discernible object of the other image stack; wherein a pleural distance of the third CDO is at least twice a pleural distance of the first CDO; wherein the at least one memory and the computer program code are configured, with the at least one processor, to determine that the third CDO does not meet the isolation criteria and subsequently to refrain from reporting the third CDO as a suspected nodule.

14. The system according to claim 10, wherein the at least one memory and the computer program code are configured, with the at least one processor, to report analyzed CDOs as suspected nodules only for analyzed CDOs for which there is up to one neighboring CDO of the group of CDOs within a circle of a given radius which comprises only the analyzed CDO and the neighboring CDO; wherein the at least one memory and the computer program code are configured, with the at least one processor, are further operable to refrain from reporting any analyzed CDO as a suspected nodule if there are at least two neighboring CDO of the group of CDOs within a circle of the given radius which comprises the analyzed CDO.

15. The system according to claim 10, wherein the at least one memory and the computer program code are configured, with the at least one processor, to refrain from reporting any analyzed CDO from being reported as a suspected nodule if the respective analyzed CDO belongs to a subgroup of N>5 CDOs of the group of CDOs, each of which complying with the rule that at least three other CDOs of the subgroup are positioned within a circle of a predetermined radius Rocentered on the respective CDO.

16. The system according to claim 10, wherein the at least one memory and the computer program code are configured, with the at least one processor, to determine whether an analyzed CDO satisfies an isolation requirement with respect to the rest of the CDOs by initially determining whether the respective CDO meets an intra-slice isolation criteria with respect to the rest of the CDOs inthe same image as the respective CDO, based at least on geometric remoteness of the respective CDO from the other CDOs; and only if the respective CDOs meets the intra-slice isolation criteria, determining whether the respective CDO meets a 3D isolation criteria with respect to CDOs of at least one other image of the staticimaging image stack; wherein the at least one memory and the computer program code are configured, with the at least one processor, to refrain from reporting an existence of a suspected nodule at the lung corresponding to any analyzed respective CDO that do not meet both the intra-slice isolation criteria and the inter-slice isolation criteria.

17. The system according to claim 10, wherein the first CDO is a small cancerous nodule having an effective diameter of between 0.5-1.5mm which is positioned less than 5cm from an outer wall of the lung, wherein the at least one memory and the computer program code are configured, with the at least one processor, to implement an adaptable isolation threshold that depends on the distance between an analyzed CDO and the outer wall of the lung for detecting the small cancerous nodule.

18. The system according to claim 10, wherein the at least one memory and the computer program code are configured, with the at least one processor, to perform a series by rule-based logic developed through human-engineered algorithms, and excludes employment of data-trained models for detecting the suspect cancerous nodules.

19. A computer program product for detecting a lung cancer nodule by processing static imaging data comprising instructions which, when executed by a computer, cause the computer to execute: obtaining a group of candidate discernible objects (CDOs) within a staticimaging image stack that comprises at least one image of a lung, the group of CDOs comprises discernible objects having associated tissue-signal response patterns corresponding to cancerous nodules; and executing for each out of multiple CDOs out of the group of CDOs: (a) determining whether the respective CDO satisfies an isolation requirementwith respect to the rest of the CDOs, based at least on geometric remoteness of the respective CDO from the other CDOs; and (b) conditionally reporting an existence of a suspected nodule at the lung corresponding to the respective CDO, in response to a result of the determining; wherein the executing comprises at least: (a) reporting a first CDO that meets the isolation criteria as a suspected nodule; and (b) determining that a second CDO that does not meet the isolation criteria should not be reported as a suspected nodule.

20. The computer program product according to claim 19, wherein the isolation criteria is further dependent upon a distance of the respective CDO from an outer wall of the lung.

21. The computer program product according to claim 20, wherein the first CDO is closer to the wall of the lung than the second CDO, wherein a distance between the second CDO to its nearest CDO of the group of CDOs is at least twice a distance between the first CDO to its nearest CDO of the group of CDOs.

22. The computer program product according to claim 19, further comprising instructions which, when executed by the computer, cause the computer to: (a) repeat the steps of obtaining and executing for a second group of CDOs within another static-imaging image stack that comprises at least one pulmonary image; wherein both the group of CDOs and the second group of CDOs comprise a subspace which consist of a subgroup of CDOs that includes a subject discernible object; wherein the first CDO is the subject discernible object in the image stack, and a third CDO of the second group of CDOs is the subject discernible object of the other image stack; wherein a pleural distance of the third CDO is at least twice a pleural distance of the first CDO; and (b) determine that the third CDO does not meet the isolation criteria and should not be reported as a suspected nodule.

23. The computer program product according to claim 19, wherein the isolation criteria permits an analyzed CDO to be reported as a suspected nodule if there is up to one neighboring CDO of the group of CDOs within a circle of a given radiuswhich comprises only the analyzed CDO and the neighboring CDO; wherein the isolation criteria prevents an analyzed CDO from being reported as a suspected nodule if there are at least two neighboring CDO of the group of CDOs within a circle of the given radius which comprises the analyzed CDO.

24. The computer program product according to claim 19, wherein the isolation criteria prevents an analyzed CDO from being reported as a suspected nodule if the respective analyzed CDO belongs to a subgroup of N>5 CDOs of the group of CDOs, each of which complying with the rule that at least three other CDOs of the subgroup are positioned within a circle of a predetermined radius Rocentered on the respective CDO.

25. The computer program product according to claim 19, further comprising instructions which, when executed by the computer, cause the computer to: determine whether the respective CDO meets an intra-slice isolation criteria with respect to the rest of the CDOs in the same image as the respective CDO, based at least on geometric remoteness of the respective CDO from the other CDOs; selectively determine, only if the respective CDOs meets the intra-slice isolation criteria, whether the respective CDO meets a 3D isolation criteria with respect to CDOs of at least one other image of the static-imaging image stack; and selectively report an existence of a suspected nodule at the lung corresponding to the respective CDO, in response to determining that the respective CDO meets both the intra-slice isolation criteria and the inter-slice isolation criteria.

Images