Cancer mapping using machine learning

Abstract

A method of mapping cancer includes inputting data elements from medical images, biopsy, and biopsy pathology tags into a machine learning model to estimate a likelihood that a patient suffers from clinically significant cancer, and outputting an estimate of the likelihood that clinically significant cancer is present at each voxel of the three-dimensional image through the machine learning model.

Description

[0001] Cross - Reference to Related Applications

[0002] This application claims priority to U.S. Non - Provisional Application No. 18 / 473,101, filed on September 22, 2023, which claims priority to U.S. Provisional Patent Application No. 63 / 376,938, filed on September 23, 2022, and U.S. Provisional Patent Application No. 63 / 385,757, filed on December 1, 2022. The disclosures of each of the above - mentioned applications are hereby incorporated by reference in their entirety. Technical Field

[0003] The present disclosure generally relates to systems and methods for detecting cancerous lesions. More specifically, the present disclosure relates to systems and methods for identifying cancerous lesions using machine - learning algorithms. Background Art

[0004] Traditionally, the approximate location of cancer within an organ or other human tissue has been determined by medical imaging techniques (such as magnetic resonance imaging (MRI)) and treated by surgery, radiotherapy, chemotherapy, hormone therapy, and / or other methods. Many of these treatments can cause harm to healthy tissue within the organ or healthy tissue surrounding the cancerous tissue, causing irreparable and unnecessary damage if introduced into areas outside the cancerous region. It is difficult for doctors to accurately predict the extent of cancerous tissue within a region, including determining the precise boundary between cancer cells and healthy tissue. Often, doctors must overestimate the size of the cancerous lesion to ensure that all cancer cells are removed, which results in the removal of some surrounding healthy tissue. However, depending on the size of the cancerous lesion and the size of the organ in which the cancer is located, this removal of healthy tissue can have an adverse effect on the continued function of the organ.

[0005] The prostate is an example of such an organ. The prostate is a small, oval gland in the male body that produces semen. Prostate cancer is the most common cancer in men, after skin cancer. Prostate cancer can be detected at the local or regional stage, representing stages I, II, and III. The location and anatomy of the prostate make cancerous lesions on or within the prostate difficult to treat, often damaging the prostate region or adjacent organs outside the cancerous lesion.

[0006] For this and other reasons, improvements are needed in the field of detecting and characterizing cancerous lesions on or within the prostate. Summary of the Invention

[0007] In at least one example of the present disclosure, a device for mapping cancer may include a processor electrically coupled to a memory component that stores electronic instructions which, when executed by the processor, cause the device to perform a machine learning algorithm configured to receive an input and generate an output based on the input, where the input includes data elements from a medical image and the output includes an estimate of the likelihood of clinically significant cancer at each voxel of a three-dimensional image.

[0008] In one example, the input further includes prostate-specific antigen (PSA), and the likelihood of clinically significant cancer includes the likelihood of clinically significant prostate cancer (csPCa). In one example, the output further includes a cancer estimation map (CEM). In one example, the CEM shows a color-coded heat map that represents the likelihood of cancer at each voxel of the three-dimensional image. In one example, the medical image is an MRI image of a patient's anatomy. In one example, the anatomy includes the prostate. In one example, the CEM includes a lesion contour that represents the lesion size of a cancerous lesion shown in the three-dimensional image. In one example, the output further includes a visualization curve representing the relationship between an encapsulation confidence score and the lesion size. In one example, the visualization curve includes points representing a specific lesion size and a specific encapsulation confidence score. In one example, the points are configured to be visually manipulated along the visualization curve to change the specific lesion size and the specific encapsulation confidence score represented by the points, where manipulating the points changes the lesion contour.

[0009] In at least one example of the present disclosure, a method for mapping cancer includes inputting data elements from a medical image into a machine learning model to estimate the likelihood of a patient having clinically significant cancer and outputting, via the machine learning model, an estimate of the likelihood of clinically significant cancer at each voxel of a three-dimensional image.

[0010] In one example, the method further includes inputting prostate-specific antigen (PSA) data elements into the machine learning model. In one example, the machine learning model is trained on a population dataset that includes the data elements. In one example, the output includes a visual representation of the three-dimensional image with a color-coded heat map, where the color-coded heat map represents the likelihood of clinically significant cancer at each voxel.

[0011] In at least one example of the present disclosure, a method for mapping cancer includes inputting data elements from a medical image into a machine learning model that estimates the likelihood that a patient has a clinically significant cancer and presenting a visual representation that shows an estimate of the likelihood of clinically significant cancer at each voxel of a three-dimensional image. The visual representation may include a Cancer Estimation Map (CEM), which shows a color-coded heat map overlaid on the image, where the color-coded heat map represents the likelihood of a clinically significant cancer lesion. The CEM includes a lesion contour representing the size of the cancer lesion and a curve representing the relationship between an encapsulation confidence score and the lesion size, the curve including points representing the lesion size and the encapsulation confidence score. In such an example, the points are configured to be visually manipulated along the curve to change the lesion size and the encapsulation confidence score represented by the points, and manipulating the points changes the lesion contour.

[0012] In one example, the method further includes presenting an intervention device located at a position relative to the image. In one example, the position of the intervention device is configured to change relative to the image. In one example, the method further includes presenting the location of a biopsy core overlaid on the image. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] The present disclosure will be readily understood by the following detailed description in conjunction with the accompanying drawings, in which like reference numerals refer to like structural elements, wherein:

[0014] Figure 1 shows a framework of a machine learning model for estimating the likelihood of clinically significant cancer;

[0015] Figure 2 shows exemplary inputs and outputs of a machine learning model for estimating the likelihood of clinically significant cancer;

[0016] Figure 3 shows MRI data elements;

[0017] Figures 4A to 4B shows additional data elements including prostate segmentation and region of interest (ROI) as outputs of the machine learning model;

[0018] Figures 5A to 5B shows csPCa positive and csPCa negative biopsy core data elements;

[0019] Figure 6 shows biopsy data elements from biopsy locations merged with MRI data elements;

[0020] Figure 7 shows a Cancer Evaluation Map (CEM);

[0021] Figure 8A graphical user interface showing a patient-specific chart;

[0022] Figures 9A to 10B Shows lesions with CEM and various lesion contours selected from individual points on the envelope confidence score versus lesion size curve;

[0023] Figure 11 Shows a 3D reconstruction of the pathological region;

[0024] Figure 12 Shows the selection of an intervention tool;

[0025] Figures 13 to 16B Shows the placement of an intervention tool in a three-dimensional representation of prostate segmentation;

[0026] Figure 17 Shows an exemplary "half-gland" margin;

[0027] Figure 18 Shows an isotropic expansion technique; and

[0028] Figures 19 to 28 Illustrates the visualization and implementation of software and user interfaces that use machine learning algorithms to map cancer. Detailed Description

[0029] Reference will now be made in detail to representative embodiments shown in the accompanying drawings. It should be understood that the following description is not intended to limit the embodiments to one preferred embodiment. Instead, it is intended to cover alternatives, modifications, and equivalents that may be included within the spirit and scope of the described embodiments as defined by the appended claims.

[0030] The following disclosure generally relates to systems and methods for detecting and characterizing cancerous lesions. More specifically, the present disclosure relates to systems and methods for identifying and characterizing cancerous lesions using machine learning algorithms.

[0031] Traditional MRI and biopsy techniques for detecting cancerous lesions provide a general location of the cancerous lesion, indicating the presence of a cancerous lesion. After cancer is detected, radiotherapy, chemotherapy, hormone therapy, surgery, ablation therapy, and / or other cancer treatment methods are applied. Such treatment methods may cause unnecessary or excessive exposure of the tissue or organ of the cancerous lesion to radiation or chemotherapy, or unnecessary removal of healthy tissue surrounding the cancer cells.

[0032] Although the methods and systems for detecting and mapping a patient's cancer described herein can be applied to multiple or all forms of cancer, the detection and mapping of prostate cancer is one example. Traditional methods may result in excessive exposure of the prostate to radiation or ablation therapy, or removal of healthy prostate tissue, which may have a negative impact on urinary, sexual, and / or bowel function, resulting in a decrease in the patient's quality of life.

[0033] The methods and systems for detecting and mapping cancer described herein can more accurately identify cancerous lesions to minimize negative impacts on surrounding healthy tissue, including methods for increasing confidence in lesion mapping thresholds. In one example of the method described herein, a machine learning algorithm can receive MRI and biopsy data (including biopsy pathology labels) as input to determine lesion thresholds and the likelihood of cancer encapsulation, as well as associated encapsulation confidence scores to minimize the risk of overexposure and overresection during ablation, radiation, and / or surgery. The machine learning algorithm described herein can be trained on a large number of data sets to improve the accuracy of cancer mapping, confidence scores, and threshold boundaries. In at least one example, the algorithm can output an estimate of the likelihood of clinically significant cancer at each voxel of a three-dimensional image.

[0034] These lesion encapsulation boundaries and associated confidence scores can be intuitively presented to the physician to convey important information for determining the best intervention strategy. These visualization outputs can include a cancer estimation map (CEM) and a three-dimensional cancer lesion contour (CLC) surrounding an area of ​​increased cancer likelihood, the cancer estimation map showing a heat map of the lesion location overlaid on a medical image of the patient's anatomy. As described herein, when mapping cancer, the CEM can indicate the spatial likelihood of the presence of a tumor, and the CLC can indicate the estimated tumor range. Although the examples of the systems described herein include CEMs, the systems described herein can be applied to CEM mapping and / or CLC generation. In addition, the systems described herein can intuitively output a graph or curve of encapsulation confidence relative to lesion size. The physician can change these outputs during the analysis process as the physician deems appropriate, thereby seeking a balance between the risk of removing or affecting healthy tissue and the risk of missing cancer cells during treatment, as shown by the confidence score curve.

[0035] In at least one example described herein, CEM can be used alone or in combination with other factors to assess the stage of a cancer. In at least one example, the cancer is prostate cancer, and the stage assessment includes an estimate of the likelihood and / or location of extra-prostatic spread. In at least one example, CEM can be used alone or in combination with other factors to assess whether a patient is suitable for a course of treatment.

[0036] In some examples, a machine learning algorithm can utilize inputs from at least one, two, or more data elements such as medical imaging (including MRI imaging, X-rays, and ultrasound imaging), other relevant medical imaging, tracked biopsies, biopsy pathology, biopsy core locations, fusion-based biopsy data, biomarkers (such as PSA), patient demographics (such as age, genomic markers), and / or other inputs. The algorithm can output an estimate of clinically significant cancer at each voxel of a three-dimensional image to create the aforementioned CEM, CLC, and encapsulation confidence scores. The estimate of clinically significant cancer can be used to identify and narrow treatment therapies (such as chemotherapy, radiation therapy, surgery, etc.). In one example, the encapsulation confidence score represents the estimated likelihood that the lesion contour contains all of the csPCa.

[0037] The machine learning algorithm of the present disclosure can be trained on a large population dataset to hone a single CEM for a specific patient. The thresholding of the CEM can be based on that population analysis. The training dataset can be used as the ground truth for training the algorithm and can include some or all of the aforementioned inputs for a large population, as well as other inputs specific to certain types of cancer or other inputs including post-care data and outcomes. The algorithm can then distinguish the cancer probability at any point within a patient's anatomy. Additionally, a smaller subset of surgical data can be used as the "tuning" dataset for the algorithm, enabling the estimation of the tumor encapsulation probability for a specific CLC and patient.

[0038] While the methods and systems described are applicable to many or all cancerous lesions, prostate cancer is used herein as an example for illustration and description. The machine learning algorithm described herein can include other inputs in addition to those described above, such as inputs specific to certain types of cancer. In the case of prostate cancer, the aforementioned inputs can be combined with other data inputs (such as prostate-specific antigen (PSA) levels) and processed through the machine learning algorithm to provide specific regions within the prostate that are affected, enabling more effective treatment and reducing the impact of urinary, sexual, and / or bowel complications.

[0039] In another example, analyzing inputs from one, two, or more data elements yields an output that includes a cancer estimation map (CEM), also known as a cancer probability map (CPM). It should be understood that in the context of the present disclosure, the terms cancer estimation map (CEM) and cancer probability map (CPM) can be used interchangeably and define the likelihood of clinically significant prostate cancer at each voxel of a three-dimensional image.

[0040] In one example, the method further includes displaying metadata and predictive statistics derived from a cancer estimation map, an encapsulation confidence score, or a tertiary statistical model. In one example, the metadata includes an encapsulation confidence score, which represents the probability that all clinically significant cancers are included within a specified lesion contour. In one example, the metadata includes the lesion contour volume. In one example, the predictive statistics include an estimate of the tumor volume. In one example, the predictive statistics include an estimate of the cancer stage. In one example, the predictive statistics include an estimated likelihood of extracapsular extension, and the possible locations with or without extracapsular extension. In one example, the predictive statistics include an estimate of whether the patient is suitable for a treatment process, such as ablation therapy, radiation, radical prostatectomy, or active surveillance. In one example, the predictive statistics include the estimated outcome of a treatment process, such as the probability of requiring additional treatment, biochemical recurrence or metastasis, the probability of treatment-related side effects, or the probability of death.

[0041] In another example, the output of the machine learning algorithm is further processed, or a second machine learning algorithm is used to convey additional metadata or information beyond the visual representations of CEM, CLC, and ECS. For example, the automatic estimated suitability of the patient for a particular treatment (radiation, surgery, ablation therapy, etc.) can be displayed, along with an estimate of the probability of success with or without that treatment. In another example, the estimated cancer stage can be displayed, along with the localization and quantification of potential sites of invasive cancer on or outside the organ of interest. Such information can be used to further assist in determining and narrowing down treatment options (such as chemotherapy, radiation therapy, surgery, etc.).

[0042] These and other embodiments will be discussed below with reference to Figures 1 to 28 However, those skilled in the art will readily understand that the detailed descriptions given herein for these figures are for explanatory purposes only and should not be construed as limiting. Additionally, as used herein, a system, method, article, component, feature, or sub-feature that includes at least one of a first option, a second option, or a third option should be understood to refer to a system, method, article, component, feature, or sub-feature that can include one of each of the listed options (e.g., only one first option, only one second option, or only one third option), multiple individual listed options (e.g., two or more first options), two options simultaneously (e.g., one first option and one second option), or a combination thereof (e.g., two first options and one second option).

[0043] While the embodiments described herein may be subject to various modifications and alternative forms, specific embodiments have been shown by way of example in the drawings and will be described in detail herein. However, the exemplary embodiments described herein are not intended to be limited to the particular forms disclosed. On the contrary, this disclosure covers all modifications, equivalents, and alternatives falling within the scope of the appended claims.

[0044] Figure 1 FIG. shows a high-level block diagram of a computer system 100 that can be used to implement embodiments of the present disclosure. In various embodiments, the computer system 100 may include Figure 1 various collections and subsets of the components shown therein. Thus, Figure 1 FIG. shows that various components can be based on the system 100, and these components can be included in various combinations and subsets based on the operations and functions performed in different embodiments. It should be noted that when described or recited herein, the use of articles such as "a" or "an" should not be construed as limited to one, but is intended to mean one or more, unless expressly stated otherwise herein.

[0045] The computer system 100 may include a central processing unit (CPU) or processor 102, which is connected via a bus 104 for electrical communication with a memory device 106, a power supply 108, an electronic storage device 110, a network interface 112, an input device adapter 116, and an output device adapter 120. For example, one or more of these components may be interconnected via a substrate (e.g., a printed circuit board or other substrate) that supports the bus 104 and other electrical connectors that provide electrical communication between the components. The bus 104 may include a communication mechanism for transferring information between parts of the system 100.

[0046] The processor 102 may be a microprocessor or similar device configured to receive and execute a set of instructions 124 stored in the memory 106. The memory 106 may be referred to as main memory, such as random access memory (RAM) or another dynamic electronic storage device, for storing information and instructions for the processor 102 to execute. The memory 106 may also be used to store temporary variables or other intermediate information during the execution of instructions by the processor 102. The power supply 108 may include a power source capable of providing power to the processor 102 and other components connected to the bus 104 (e.g., a connection to an electrical utility grid or a battery system).

[0047] The storage device 110 may include a read-only memory (ROM) or another type of static storage device coupled to the bus 104 for storing static or long-term (i.e., non-dynamic) information and instructions for the processor 102. For example, the storage device 110 may include a magnetic or optical disk (e.g., a hard disk drive (HDD)), a solid-state memory (e.g., a solid-state disk (SSD)), or a similar device. The instructions 124 may include information for performing processes and methods using the components of the system 100.

[0048] The network interface 112 may include an adapter for connecting the system 100 to external devices via a wired or wireless connection. For example, the network interface 112 may provide a connection to a computer network (e.g., a cellular network, the Internet, a local area network (LAN)), a separate device capable of wireless communication with the network interface 112, other external devices or network locations, and combinations thereof. In one example implementation, the network interface 112 is a wireless network adapter configured to connect to another device having interface capabilities using the same protocol via WI-FI(R), BLUETOOTH(R), BLE, Bluetooth Mesh, or a related wireless communication protocol. In some implementations, network devices or groups of network devices in the network 126 may be considered part of the system 100. In certain cases, network devices may be considered connected to the system 100 but not part of the system 100.

[0049] The input device adapter 116 may be configured to provide the system 100 with connections to various input devices, such as a touch input device 113 (e.g., a display or display component), a keyboard 114, or other peripheral input devices, one or more sensors 128, related devices, and combinations thereof. In some configurations, the input device adapter 116 may include the aforementioned touch controller or a similar interface controller. The sensors 128 may be used to detect physical phenomena (e.g., light, sound waves, electric fields, forces, vibrations, etc.) near the computing system 100 and convert these phenomena into electrical signals. The keyboard 114 or another input device (e.g., a button or switch) may be used to provide user input, such as input regarding the settings of the system 100.

[0050] The output device adapter 120 may be configured to enable the system 100 to output information to a user, such as by providing visual output using one or more displays 132, providing auditory output using one or more speakers 135, or providing touch-perceived tactile feedback using one or more tactile feedback devices 137. Other output devices may also be used. The processor 102 may be configured to control the output device adapter 120 to provide information to the user via the output devices connected to the adapter 120.

[0051] Figure 1Any features, components, and / or parts (including their arrangements and configurations) shown in [Figure 0] can be included, either individually or in any combination, in examples of any devices, features, components, and parts shown in other figures. For example, computing system 100 can be used to run the algorithms described herein and display the visual representations described herein and shown in other figures. Similarly, any devices, features, components, and parts Figure 1 shown in [Figure 1], including their arrangements and configurations, can be included, either individually or in any combination, in Figure 1 examples of the features, components, and / or parts shown in [Figure 2].

[0052] Figure 2 An exemplary data flow diagram 200 of a machine learning model is shown that is used to estimate the likelihood of clinically significant cancer. As an example, the systems described herein are described with reference to prostate cancer. However, the systems and methods described herein can also be applied to other types of cancer. In at least one example, a software program can utilize at least one, two, or more data elements as input 202, where input 202 includes MRI data element 204, biopsy pathology data element 206, prostate-specific antigen (PSA) data element 208, and / or other data elements for determining the cancer probability. The detection of prostate cancer, particularly using PSA as an input to the systems described herein, is merely exemplary and is not intended to be limiting. Instead, as described above, the systems and methods described herein can be applied to other types of cancer using other types of antigens, imaging modalities, genetic information, demographic data, or biomarkers indicative of other types of cancer as inputs to the system. In one example, input 202 can exclude PSA data element 208.

[0053] In at least one example, data elements 204, 206, 208 are used as inputs 202 to a machine learning model 210. The machine learning model can be a single model, multiple models operating in series or in parallel, and / or one or more models with added post-processing analysis. In one example, the machine learning model 210 can then estimate the likelihood of clinically significant cancer. In one example, in the case of prostate cancer, clinically significant cancer can be defined as Gleason grade group 2 or higher disease. In at least one example, the machine learning model 210 can then provide an output 216 that includes an estimate of the likelihood of clinically significant cancer at each voxel of a 3D image defined herein as a Cancer Estimation Map (CEM) 212. The final output can be a lesion contour 214. The lesion contour can be a cancer probability threshold generated by a 3D surface. The output can include additional data related to or derived from the machine learning model, such as an estimated probability of tumor encapsulation, an estimated tumor stage, an estimate of the patient's suitability for a particular therapy (surgery, radiation, ablation therapy, etc.), segmentation of the tumor, and / or segmentation of anatomical structures (prostate, urethra, bladder, seminal vesicles, prostate zone, vas deferens, rectum, pelvic bones, etc.).

[0054] Figure 2 Any features, components, and / or parts (including their arrangements and configurations) shown in can be included, either individually or in any combination, in examples of any devices, features, components, and parts shown in other figures. Similarly, any devices, features, components, and parts Figure 2 shown in the devices, features, components, and parts (including their arrangements and configurations) can be included, either individually or in any combination, in Figure 2 examples of the features, components, and / or parts shown in.

[0055] Figures 3 to 28 shows Figure 2 an example of the output 216 of the machine learning algorithm shown in, and additional data related to or derived from the machine learning model. When a software program that executes the machine learning algorithm or the output of the machine learning algorithm is run by a computing device, Figures 3 to 28 the output shown in can be visually presented to a doctor or other user on a display screen. In one example, as Figure 3 shown, the input 302 can include a magnetic resonance image (MRI) 300 of a patient, which can be used as an MRI data element 304 for display. The MRI data element 204 can contain information related to the MRI coordinate space and can be one of the inputs 202 to the machine learning model 210 as described in Figure 2 . The MRI data element 204 can be one or more MRI sequences (T2-weighted, diffusion-weighted, perfusion-weighted, etc.) obtained from the same patient.

[0056] Figure 4A shows an exemplary prostate MRI 300, as previously discussed in Figure 3 and includes an exemplary prostate segmentation 402 identified by a machine learning model 210. Figure 4B shows an exemplary prostate MRI 300 of a patient, including an exemplary prostate segmentation 402 and a region of interest (ROI) 404 identified by a machine learning model 210 as outputs.

[0057] Figure 5A shows csPCa-positive biopsy cores 502 that are from a biopsy system, e.g., a biopsy system that registers or fuses biopsy locations with MRI data elements. For example, a patient can receive a biopsy from a biopsy system with identified csPCa-positive cores, and the core locations on the prostate can be digitally transcribed onto the MRI image relative to the Figure 4A prostate segmentation 402 as described. Similarly, Figure 5B shows csPCa-negative biopsy cores 504 identified using a biopsy system capable of tracking core locations within the prostate. The location data of the csPCa-negative biopsy cores 504 are received as input to the machine learning model 210 and are represented in a color different from that of the csPCa-positive biopsy cores 502. For example, the csPCa-positive biopsy cores 502 can be represented in red, while the csPCa-negative biopsy cores 504 can be represented in blue, thereby creating a visually distinguishable difference between the positive biopsy cores 502 and the negative biopsy cores 504. Other cores suitable for other classes can be represented in one or more third colors. For example, biopsy cores containing clinically insignificant cancer can be represented in orange.

[0058] As Figure 6 shown, each biopsy core 502, 504 is labeled with one, two, or more attributes 602, such as Gleason score, cancer percentage, cancer length, and core length determined and recorded by a pathologist or a pathology analysis algorithm in a pathology report.

[0059] Figure 7 provides a visual representation of a cancer evaluation map (CEM) 702 as an exemplary output of the machine learning model 210. The CEM 702 identifies the probability of cancer locations according to a color gradient (e.g., a heat map), which provides a visual representation of the probability of cancerous tissue at a specific location relative to the prostate segmentation 402. For example, as Figure 2As described in the foregoing, the machine learning model 201 can receive one, two, or more inputs (e.g., tracking biopsies, biopsy pathology, and prostate specific antigen (PSA)) and estimate prostate segmentation 402, regions of interest 404, and / or cancer probability maps 702 as outputs 216. The outputs 402, 404, 702 improve the accuracy and effectiveness of prostate cancer treatment and / or reduce the amount of prostate tissue removed or damaged compared to conventional prostate treatment methods.

[0060] The clinician is also presented with a patient-specific graph 800 showing an envelope confidence curve 806, also known as a Marks confidence curve, such as Figure 8 800. The x-axis 802 of the patient-specific chart 800 represents the percentage of prostate voxels encapsulated by the iterative thresholding CEM 702. The y-axis 804 of the patient-specific chart 800 represents the encapsulation confidence score, i.e., the confidence that all cancer cells in the lesion are encapsulated using each CEM threshold, ranging from zero to one hundred, expressed as a percentage. The encapsulation confidence score is based on a lookup table that relates the probability of csPCa encapsulation to the CEM threshold. The lookup table provides the clinician with data from a retrospective review of the overall pathology data.

[0061] like Figure 9A As shown, the machine learning model 210 generates a default point 902 within the patient-specific chart 800. The default lesion contour is selected to maximize the encapsulation confidence score 804 represented on the y-axis while minimizing the lesion size 802 represented on the x-axis.

[0062] Figure 9B The lesion outline shown in is generated after selecting a point on the patient-specific chart. For example, the user-selected point 904 on the patient-specific chart 800 selects the smaller lesion size 802 represented on the x-axis (e.g., Figure 8 ), thereby reducing the encapsulation confidence score 804 represented on the y-axis (also shown Figure 8 904). The user-selected point 904 results in a lower confidence level of encapsulation than the default point 902. The resulting lesion contour size is determined by the lesion contour 906 (as shown in FIG. Figure 9B ), which is superimposed on CEM 702, thereby creating a visual representation of the selected lesion contour size relative to CEM 702, allowing the clinician to effectively "adjust" the lesion contour size represented by lesion contour 906, thereby increasing the likelihood of treatment success or reducing the amount of treatment.

[0063] Similarly, Figures 10A to 10B The encapsulation confidence score 804 of the user selected point 904 is shown to be greater than the default point 902, thereby increasing the lesion contour size on the x-axis, as represented by the lesion contour 906 on the CEM 702 above the default point 902 shown on the curve.

[0064] The user can freely adjust the lesion contour 906 representing the lesion contour size in the CEM 702 by selecting any point on the encapsulation confidence curve 806. The encapsulation confidence curve 806 helps to balance the likelihood of csPCa encapsulation and the lesion contour size. For example, as Figures 9A to 9B shown, the user can reduce the lesion contour size by selecting a lower CEM 702 threshold and then update the encapsulation confidence score 804. Conversely, as Figures 10A to 10B shown, the user can increase the lesion contour size 96 by selecting to increase the encapsulation confidence score 804. The lesion contour size selected by the user may depend on the patient's anatomy, the doctor's expertise, the type of planned intervention, and other factors.

[0065] As Figure 11 shown, the cancer probability map 702 and the lesion contour are evaluated using whole pathology. The pathologically tumor regions with MRI registration and 3D reconstruction are used to define the ground truth csPCa, enabling the precise and objective evaluation of key software features such as the encapsulation confidence curve 806.

[0066] As Figure 12 shown, once the cancer lesion contour size is confirmed, the user can place virtual intervention devices with customizable sizes. In one example, these devices may represent interstitial catheters, and the intervention may be thermal ablation of cancerous tissue. The user can select the tools suitable for the desired ablation size. For example, if a smaller cancer lesion contour size is determined, a smaller ablation size will be applicable. If a larger cancer lesion contour size is determined, a larger ablation size will be applicable. The ablation size, position, and orientation are selectable by the user, and the placement of the intervention device is customizable. The user can choose to place the intervention device manually or semi - automatically.

[0067] As Figures 13 to 16B shown, the user can view the prostate segmentation 1302. In another example, the user can view the prostate segmentation 1302, the cancer lesion contour 1305, and the intervention device 1304. The position and number of the intervention device 1304 are identified and presented digitally to the user. In one example, the intervention device 1304 represents a probe for inducing tissue ablation, and the ablation volume associated with each probe is displayed relative to the image. In one example, the position of the intervention device 1304 and / or the ablation volume is compared with the position of other anatomical structures. In one example, the configuration of the intervention device 1304 and other anatomical structures can be identified as potential causes of safety or efficacy issues.

[0068] Figure 16A and Figure 16BSpecifically shown is CEM 1301 and corresponding medical image 1307, which shows prostate segmentation 1302 and virtual intervention device 1304. Figure 16A A plan view of medical image 1307 is shown, Figure 16B and a side view thereof is shown to illustrate the three-dimensional information and properties of the shown prostate segmentation 1302 and intervention device 1304. Medical image 1307 and CEM 1301 can be shown side by side to provide a visual context to the practitioner. Prostate segmentation 1302 can be represented in three-dimensional space, and intervention device 1304 can be placed within the virtual three-dimensional space of prostate segmentation 1302. The position of intervention device 1304 superimposed on or within three-dimensional segmentation 1302 can correspond to the recommended position of the actual device used during the intervention. CEM 1301 can include prostate segmentation 1302, lesion contour 1305, and region of interest 1309, as described elsewhere herein.

[0069] Figure 17 The most conventional method of prostate cancer treatment is shown, i.e., the "hemi-gland" margin. Studies have shown that nearly half of cases that appear to be unilateral cancer actually have bilateral cancer. In this example, the hemi-gland margins of the right or anterior hemispheres would both have failed, indicating the need for a more comprehensive, patient-specific approach to cancer detection and treatment.

[0070] Figure 18 Another conventional method of prostate cancer detection and treatment is shown, i.e., the isotropic region of interest (ROI) expansion method. This method defines a uniform or isotropic margin around the ROI. However, this method fails to account for the fact that MIR-invisible tumor expansion typically grows in an unpredictable and asymmetric manner, indicating the need for a more comprehensive, patient-specific approach to cancer detection and treatment.

[0071] Figures 19 to 28 Various visual representations and implementations of software and user interfaces that map cancer using machine learning algorithms are shown, as described above.

[0072] The articles "a", "an", and "the" are intended to denote the presence of one or more of the elements recited in the foregoing description. The terms "comprising", "including", and "having" are intended to be inclusive and to denote the possible presence of elements other than those recited. Further, it should be understood that references to "one embodiment" or "an embodiment" of the present disclosure are not intended to be construed as excluding the existence of other embodiments that also incorporate the recited features. The numbers, percentages, ratios, or other values recited herein are intended to include that value, as well as other values that are "about" or "approximate" the recited value, as will be understood by one of ordinary skill in the art, and such values are incorporated in the embodiments of the present disclosure. Accordingly, the values should be interpreted broadly as being at least close enough to the recited value to perform the desired function or achieve the desired result. The values include at least the variations that are expected in a suitable manufacturing or production process, and may include values within 5%, within 1%, within 0.1%, or within 0.01% of the recited value.

[0073] Given the present disclosure, one of ordinary skill in the art should recognize that equivalent structures do not depart from the spirit and scope of the present disclosure, and that various changes, substitutions, and alterations can be made to the embodiments disclosed herein without departing from the spirit and scope of the present disclosure. Equivalent structures, including functional "means-plus-function" clauses, are intended to cover structures that perform the recited function as described herein, including structural equivalents that operate in the same manner and equivalent structures that provide the same function. The applicant expressly disclaims reliance on means-plus-function or other functional claim limitations, except in those claims in which the term "means" appears in conjunction with the relevant function. Every addition, deletion, and modification to an embodiment that falls within the meaning and scope of the claims should be included in the claims.

[0074] The terms "about", "substantially", and "essentially" as used herein denote a quantity that is close to the recited quantity, but still capable of performing the desired function or achieving the desired result. For example, the terms "about", "substantially", and "essentially" can refer to a quantity that is within 5% less than the recited quantity, within 1% less than the recited quantity, within 0.1% less than the recited quantity, and within 0.01% less than the recited quantity. Further, it should be understood that any direction or reference frame in the foregoing description is merely a relative direction or motion. For example, any reference to "up" and "down" or "above" or "below" is merely a description of the relative position or motion of the relevant elements.

[0075] The present disclosure may be embodied in other specific forms without departing from its spirit or essential characteristics. The embodiments are to be considered in all respects only as illustrative and not restrictive. Thus, the scope of the present disclosure is indicated by the appended claims rather than by the foregoing description. All changes that come within the meaning and range of equivalency of the claims are to be embraced within their scope.

Claims

1. An apparatus for mapping cancer, comprising: a processor electrically connected to a memory component that stores electronic instructions which, when executed by the processor, cause the apparatus to perform a machine learning algorithm configured to receive an input and generate an output based on the input, where the input includes data elements from a medical image; and the output includes an estimate of the likelihood of clinically significant cancer at each voxel of a three-dimensional image.

2. The apparatus according to claim 1, wherein the input further includes prostate-specific antigen (PSA); and the likelihood of clinically significant cancer includes the likelihood of clinically significant prostate cancer (csPCa).

3. The apparatus according to claim 1, wherein the output further includes a cancer estimation map (CEM).

4. The apparatus according to claim 3, wherein the cancer estimation map shows a color-coded heat map representing the likelihood of cancer at each voxel of the three-dimensional image.

5. The apparatus according to claim 4, wherein the medical image is an MRI image of a patient's anatomy.

6. The apparatus according to claim 5, wherein the anatomy includes the prostate.

7. The apparatus according to claim 4, wherein the cancer estimation map includes a lesion contour representing the lesion size of a cancer lesion shown in the three-dimensional image; and the lesion contour includes an encapsulation confidence score.

8. The apparatus according to claim 7, wherein the output further includes a visualization curve representing the relationship between the encapsulation confidence score and the lesion size.

9. The apparatus according to claim 8, wherein the visualization curve includes points representing a specific lesion size and a specific encapsulation confidence score.

10. The apparatus according to claim 9, wherein the points are configured to be visually manipulated along the visualization curve to change the specific lesion size and specific encapsulation confidence score represented by the points.

11. The apparatus according to claim 10, wherein manipulating the points changes the lesion contour.

12. The apparatus according to claim 1, wherein the medical image includes an MRI image.

13. A method for mapping cancer, comprising: inputting data elements from a medical image into a machine learning model that estimates the likelihood of a patient having clinically significant cancer; and generating an output by the machine learning model, the output including an estimate of the likelihood of clinically significant cancer at each voxel of a three-dimensional image.

14. The method according to claim 13, further comprising inputting data elements from a biopsy and biopsy pathology labels into the machine learning model.

15. The method according to claim 13, wherein the machine learning model is trained on a population dataset including the data elements.

16. The method according to claim 13, wherein The output includes a visual representation of the three-dimensional image with a color-coded heat map that represents the likelihood of clinically significant cancer at each voxel.

17. A method for mapping cancer, comprising: inputting data elements from a medical image into a machine learning model that estimates the likelihood of a patient having clinically significant cancer; displaying a visual representation of the likelihood at each voxel of a three-dimensional image, the visual representation including: a Cancer Estimation Map (CEM) showing a color-coded heat map overlying the three-dimensional image, the color-coded heat map representing the likelihood of clinically significant cancer, the Cancer Estimation Map including a lesion contour representing the size of the cancer lesion; and a curve representing the relationship between an encapsulation confidence score and the size of the cancer lesion, the curve including points that represent the cancer lesion size and the encapsulation confidence score; wherein: the points are configured to be visually manipulated along the curve to change the cancer lesion size and the encapsulation confidence score represented by the points; and manipulation of the points changes the lesion contour.

18. The method according to claim 17, further comprising displaying an intervention device in a position relative to the three-dimensional image.

19. The method according to claim 18, wherein, the position of the intervention device is configured to be changed relative to the three-dimensional image.

20. The method according to claim 18, further comprising displaying the position of a biopsy core overlying the three-dimensional image.

Images