radiology-based biopsy score risk assessment

Abstract

Computer-implemented methods for determining tumor heterogeneity are provided. Embodiments of the methods involve analyzing radiological images of a tumor to compute a value indicative of tumor activity at each of a plurality of locations in the tumor; determining whether each computed value is above a predetermined threshold; and computing a heterogeneity score for the tumor; wherein the heterogeneity score is computed based on a volume fraction of the tumor having a value indicative of tumor activity that is above the predetermined threshold. Associated systems and computer program products are also provided.

Description

TECHNICAL FIELD

[0001] Embodiments described herein relate generally to methods, systems, and computer program products for determining tumor heterogeneity. In particular aspects, the invention relates to computer-implemented methods for determining a representative pathology score (e.g., a score indicative of expression of Programmed Death 1 Receptor and its Ligand (PD-1 / PD-L1)) obtained from processing a biopsy sample of tumor tissue. The methods are particularly useful for determining whether a pathology score obtained from a potentially heterogeneous tumor can be inaccurate. The output of such methods is useful in clinical decision making, e.g., for determining further diagnostic and / or therapy steps in cancer treatment. BACKGROUND

[0002] With the introduction of checkpoint inhibition immunotherapy (CIT), treatments that modulate the immune system have become an important class of cancer therapies, with the FDA approving many cancer types even in first line treatment. These therapies can achieve durable responses in advanced patients. However, the percentage of responders to monotherapy is only 15-30% depending on the cancer type. The basic mechanism of CIT is to activate cytotoxic T cells that are able to recognize tumors as foreign. By blocking checkpoint molecules on target cells, effector functions can be used and tumor cells can be killed by infiltrating T cells.

[0003] The most important checkpoint clinically is the Programmed Death 1 Receptor and Ligand (PD-1 / PD-L1). Overexpression of the PD-L1 molecule on tumor cells is used as a biomarker for selecting patients in first line treatment of various cancers, such as non-small cell lung cancer (NSCLC).

[0004] Most lung cancer patients are diagnosed in an advanced stage and the diagnosis is based on small biopsies or cytological specimens. In the context of advanced cancer, surgery is usually not applicable. There are several techniques available for obtaining lung cancer biopsy samples, including bronchoscopy, ultrasound-guided bronchoscopy, mediastinoscopy, transthoracic needle aspiration, thoracentesis, and medical thoracoscopy. All of these have limitations in terms of tissue yield and their use depends to a large extent on the location of the lesion and its accessibility. This is a major challenge for lung cancer. In the best case, transthoracic needle aspiration is performed using a 20- to 22-gauge core needle, and each core biopsy typically contains approximately 500 cells. This material is then paraffin-embedded and sectioned for microscopic analysis.

[0005] Treatment options in CIT typically involve the use of such techniques to determine PD-L1 expression on tumor cells in a biopsy sample obtained from a patient. However, it is clear that PD-L1 expression is not a good predictor of patient response. One reason is the strong spatial heterogeneity of the biomarker expression limiting the validity of the assay on a small biopsy at a single time point.

[0006] Tumors consist of different cell and tissue types. Cancer cells at different locations within a tumor can differ in their genetic make-up, activity and environmentally related signaling behavior. It has been noted that, in particular, related PD-L1 expression can vary greatly within a single lesion and between lesions of the same patient. For the determination of PD-L1 expression by immunohistochemistry, usually only a biopsy is available. The score obtained from the biopsy can not reflect the average properties of the tumor. In this way, a negative result of the biopsy can turn into a false negative, as other parts of the tumor can be positive.

[0007] Currently, the likelihood of taking a non-representative biopsy but a potentially incorrect pathological evaluation cannot be estimated. PD-L1 scores are an important factor in treatment decisions. Therefore, there is a need for a method to determine tumor heterogeneity that can be used to predict how likely it is that a biopsy sample of a tumor is representative. In particular, there is a need for a method that can predict the likelihood of obtaining a biopsy specimen that is representative of a tumor with respect to, for example, PD-L1 expression for pathological evaluation.

[0008] US 2014 / 314292 discloses a method and system for integrating radiological (e.g., MR, CT, PET, and ultrasound) and pathological information for diagnosis, therapy selection, and disease monitoring. In this disclosure, a location corresponding to each of one or more biopsy samples is determined in at least one radiological image. An integrated display is used to display a histological image corresponding to each biopsy sample, a radiological image, and a location in the radiological image corresponding to each biopsy sample. SUMMARY

[0009] The objects of the present application are solved by the subject matter of the independent claims, wherein further embodiments are incorporated in the dependent claims. It should be noted that the following described aspects of the present application equally apply to the method, system and computer program product for determining tumor heterogeneity.

[0010] In embodiments of the application, the tumour heterogeneity score is based on radiological image analysis, for example using positron emission tomography (PET) to calculate. Radiological image analysis can be used to indicate the risk of obtaining a non-representative result from pathological analysis of a biopsy sample. In particular, using the present method, the heterogeneity score obtained from radiological image analysis can be compared to the pathology score obtained from analysis of a biopsy sample. By comparing the heterogeneity score to the pathology score, the representativeness of the pathology score to the whole tumour can be predicted. For example, in some embodiments, the method can predict that the (pathology) score based on the biopsy is non-representative and / or inaccurate due to tumour local variations at the biopsy site. This can have important implications for therapy, for example by identifying false negative results in the pathology score and indicating that further samples / tests are required.

[0011] For example, in one embodiment, FDG-PET / CT image analysis is used to indicate potential heterogeneity of PD-L1 expression in a biopsy sample, and therefore the probability of obtaining an incorrect / non-representative PD-L1 score in the biopsy. In another embodiment, radiological image analysis can be used to guide the collection of a biopsy sample.

[0012] Thus, in one embodiment, the present application provides a computer-implemented method for determining tumour heterogeneity, the method comprising: analysing a radiological image of a tumour to calculate a value indicative of tumour activity at each of a plurality of locations in the tumour; determining whether each calculated value is above a predetermined threshold; and calculating a heterogeneity score for the tumour; wherein the heterogeneity score is calculated based on a volume fraction of the tumour having a value indicative of tumour activity above the predetermined threshold.

[0013] In one embodiment, the method further comprises calculating a probability that a biopsy sample taken from the tumour has a tumour activity value above the predetermined threshold, wherein the probability is calculated based on the heterogeneity score and a volume of the biopsy sample.

[0014] In another embodiment, the method further comprises: analysing a microscopic image of a tumour biopsy sample to determine a value of a biological parameter associated with the tumour; and determining a likelihood value that the value of the biological parameter is representative of the tumour, wherein the likelihood value is calculated based on the heterogeneity score and / or the probability.

[0015] In some embodiments, the radiological image is obtained by positron emission tomography (PET), magnetic resonance imaging (MRI), computed tomography (CT) and / or single photon emission computed tomography (SPECT).

[0016] In one embodiment, the values indicative of tumor activity in the tissue are standard uptake values (SUVs) obtained, for example, by PET. For example, the values indicative of tumor activity (e.g. SUVs) can be obtained by PET imaging using (18-F)fluorodeoxyglucose (FDG).

[0017] In one embodiment, the biological parameter comprises a level of expression of programmed death ligand 1 (PD-L1) on tumor cells.

[0018] In one embodiment, the method further comprises computing a contour plot showing tumor regions having equal tumor activity values. Preferably, at least one contour line in the contour plot corresponds to a predetermined tumor activity threshold. In some embodiments, the contour plot can comprise a plurality of contour lines, each contour line corresponding to a different predetermined tumor activity threshold. In some embodiments, each contour line (and each predetermined tumor activity threshold) can be associated with a clinically relevant threshold of the biological parameter (e.g. a different percentage of tumor cells positive for PD-L1 expression (tumor proportion score or TPS), such as 1% or 50%).

[0019] In one embodiment, the method further comprises indicating to a user a tumor region having a tumor activity above a predetermined threshold, from which a biopsy sample is to be extracted.

[0020] In one embodiment, the volume fraction of the tumor is determined based on the smallest connected volume having values above a predetermined threshold.

[0021] In one embodiment, the values indicative of tumor activity are indicative of metabolic activity, hypoxia and / or proliferation in the tumor.

[0022] In one embodiment, each value indicative of tumor activity is determined for a volume element within the tumor, and the mean value of the volume elements is compared to a predetermined threshold. In another embodiment, each value indicative of tumor activity is determined for a volume element within the tumor, and the maximum value of the volume elements is compared to a predetermined threshold.

[0023] In one embodiment, the values of tumor activity (e.g. obtained from radiological images) are correlated with values of a biological parameter (e.g. obtained from microscopic images). Preferably, the predetermined threshold of tumor activity is correlated with a clinically relevant threshold of the biological parameter associated with the tumor. For example, in one embodiment, the biological parameter is the percentage of tumor cells positive for PD-L1 expression (tumor proportion score or TPS). The clinically relevant threshold can be, for example, a TPS value indicative of whether a checkpoint inhibitor treatment (or e.g. a particular modality thereof, such as a combination of checkpoint inhibitor treatment and chemotherapy) is suitable.

[0024] In one embodiment, the heterogeneity score is compared to a value of a biological parameter. For example, an inconsistency between the heterogeneity score and the value of the biological parameter can indicate that the value of the biological parameter is not representative. In particular, if the value of the biological parameter is below a clinically relevant threshold, the heterogeneity score can indicate that a significant volume fraction of the tumor has an activity above a corresponding predetermined threshold. In such an embodiment, it can be recommended to take further biopsy samples.

[0025] In another aspect, the present application provides a system for determining tumor heterogeneity, the system comprising: an interface for receiving a radiological image of a tumor; a memory; and a processor configured to execute instructions stored on the memory to: (a) analyze the radiological image of the tumor to calculate a value indicative of tumor activity at each of a plurality of locations in the tumor; (b) determine whether each calculated value is above a predetermined threshold; and (c) calculate a heterogeneity score for the tumor; wherein the heterogeneity score is calculated based on a volume fraction of the tumor having a value indicative of tumor activity above the predetermined threshold.

[0026] In another aspect, the present application provides a computer program product comprising a non-transitory computer readable medium having computer readable code embodied therein, the computer readable code being configured such that, on execution by a suitable computer or processor, the computer or processor is caused to perform the method as described above. BRIEF DESCRIPTION OF DRAWINGS

[0027] Exemplary embodiments of the present application will now be described with reference to the following drawings, in which:

[0028] Figure 1 A system block diagram of a computerized system for determining tumor heterogeneity is shown.

[0029] Figure 2 A flowchart of a method for determining tumor heterogeneity according to the present application is shown.

[0030] Figure 3 Pathology images from a biopsy sample of a tumor stained for PD-L1 expression are shown. (A) shows a low resolution image of the stained sample, indicating regions B and C. (B) shows a high resolution image of region B, which is PD-L1 positive (i.e., showing elevated PD-L1 expression). (C) shows a high resolution image of region C, which is PD-L1 negative.

[0031] Figure 4 A PET image of a tumor is shown, with a contour plot in which solid lines indicate tumor regions with equal metabolic activity (SUV) values. Regions of tumor volume with SUV above 10, SUV between 5 and 10, and SUV between 3 and 5 are thus indicated.

[0032] Figure 5 SUV histograms are shown for volume elements from a region of interest. Solid lines for each voxel, dashed lines for each analyzed volume (corresponding to biopsy volumes). Vertical lines indicate pre-selected thresholds corresponding to an interpretation of the disease. Cut-off points indicate fraction of volume elements with SUV below threshold. For larger volume elements (i.e. biopsy volumes), the fraction is smaller as indicated by dashed lines. DETAILED DESCRIPTION

[0033] Pathology information is critical for diagnosis, staging, and typing of cancer. Often only small core or needle biopsy samples are available. Since tumor properties show spatial heterogeneity, there is a significant risk of non-representative diagnostic results. Described herein in one embodiment is a method of using radiological image analysis to indicate the risk of having a non-representative result from a biopsy pathology analysis. As one embodiment, we describe using FDG-PET / CT image analysis to indicate potential heterogeneity of PD-L1 expression and the chance of obtaining an incorrect PD-L1 score in a biopsy. Another embodiment describes using such analysis to guide biopsies.

[0034] Various embodiments are described in greater detail below with reference to the accompanying drawings, which form a part of this disclosure. The accompanying drawings and descriptive text thereto are included to illustrate specific exemplary embodiments of the disclosure and to enable a person of ordinary skill in the art to make and use the disclosure. The concepts, technologies, and implementations of the disclosure, however, can be practiced with alterations and modifications to fit particular applications. Embodiments can be practiced as methods, systems, or devices. Accordingly, embodiments can take the form of a hardware implementation, a completely software implementation, or an implementation combining software and hardware aspects. The following detailed description is, therefore, not to be taken in a limiting sense.

[0035] Reference throughout the specification to "one embodiment" or "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one example implementation or technique according to the disclosure. The appearance of the phrases "in one embodiment" in various places in the specification are not necessarily all referring to the same embodiment.

[0036] Unless specifically stated otherwise as apparent from the following discussion, it is appreciated that throughout the description, discussions utilizing terms such as "processing" or "computing" or "calculating" or "determining" or "displaying" or the like, refer to the action and processes of a computer system, or similar electronic computing device, that manipulates and transforms data represented as physical (electronic) quantities within the computer system memories or registers or other such information storage, transmission or display devices. Portions of the present disclosure include processes and instructions that can be embodied in software, firmware or hardware, and when embodied in software, can be downloaded to reside on and be operated from different platforms used by a variety of operating systems. The present invention can be implemented in various components and in various arrangements of components, and in various process operations and arrangements of process operations.

[0037] The present disclosure also relates to an apparatus for performing the operations herein. This apparatus can be specially constructed for the required purposes, or it can comprise a general purpose computer selectively activated or reconfigured by a computer program stored in the computer. Such a computer program can be stored in a computer readable storage medium, such as, but not limited to, any type of disk including floppy disks, optical disks, CD-ROMs, magnetic-optical disks, read-only memories (ROMs), random access memories (RAMs), EPROMs, EEPROMs, magnetic or optical cards, special purpose integrated circuits (ASICs) or any type of media suitable for storing electronic instructions, and each coupled to a computer system bus. Furthermore, the computers described in this specification can include a single processor or can be architectures employing multiple processor designs for increased computing capability.

[0038] The processes and displays presented herein are not inherently related to any particular computer or other apparatus. Various general purpose systems can be used with programs in accordance with the teachings herein, or it can prove convenient to construct more specialized apparatus to perform one or more method steps. The structure for a variety of these systems will appear as follows. Additionally, any particular programming language can be used to implement the present disclosure in its technological and implementation aspects. A variety of programming languages can be used to implement the present disclosure as discussed herein.

[0039] Additionally, the language used in the specification is principally intended to be read in a manner most consistent with the assignment of claims to achieve the most broadly contemplated scope of the disclosure. Accordingly, the disclosure is intended to be illustrative, but not limiting, of the scope of the concepts discussed herein. Many additional advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description.

[0040] Figure 1 A system 100 for determining tumor heterogeneity is illustrated in accordance with one embodiment.

[0041] The system 100 can include a user input / output (I / O) device 102 and a processor 104 that executes instructions stored on a memory 106. The processor 104 can be in communication with or otherwise include an interface 110 for receiving imaging data from image data sources 112 and 114. For example, the image data source 112 can include a radiological imaging system such as a PET scanner, while the image data source 114 can include a microscopic imaging system such as a digital microscope.

[0042] The I / O device 102 can be any suitable device that can receive commands from an operator and output radiological and imaging data and associated scores therein. The I / O device 102 can be configured as, for example, but not limited to, a personal computer, a tablet computer, a laptop computer, a mobile device, or the like.

[0043] The processor 104 can be any specifically configured processor or hardware device capable of executing instructions stored on the memory 106 to process radiological and optionally microscopic imaging data to determine quantitative values therefrom. The processor 104 can include a microprocessor, a field-programmable gate array (FPGA), an application-specific integrated circuit (ASIC), or other similar device. In some embodiments, such as embodiments relying on one or more ASICs, functionality described as being provided in part via software can instead be hardwired into the operation of the ASIC, and thus, any associated software can be omitted.

[0044] The memory 106 can be an LI, L2, L3 cache or RAM memory configuration. As noted above, the memory 106 can include non-volatile memory such as flash memory, EPROM, EEPROM, ROM, and PROM, or volatile memory such as static or dynamic RAM. The exact configuration / type of the memory 106 can of course vary so long as instructions for analyzing radiological and optionally microscopic imaging data can be executed by the processor 104.

[0045] The interface 110 can receive radiological imaging data from the data source 112 and optionally microscopic imaging data from the data source 114. The interface 110 can then communicate the received data to the processor 104 for analysis. The radiological and microscopic imaging data is typically in the form of digital images of tissue of interest within a subject. The radiological images typically show tumor activity at different locations within a tumor. The radiological images can be obtained, for example, by PET analysis and standard data processing techniques can be used to convert the raw data into images indicative of tumor activity. Additionally, corrections can be employed to account for, for example, contrast agent concentration and tracer attenuation. The microscopic imaging data can show, for example, expression of tumor markers of other proteins of interest within a tumor.

[0046] The processor 104 is configured to compute a value indicative of tumor activity at each of a plurality of locations in the radiological image and determine whether each computed value is above a predetermined threshold. Typically, the predetermined threshold is derived from a correlation of tumor activity to a biological parameter of interest. For example, the predetermined threshold (for tumor activity) can correspond to a particular clinical threshold of the biological parameter that is indicative of a diagnostic or therapeutic decision. The processor 104 then computes a heterogeneity score of the tumor based on a fraction of the tumor volume having a tumor activity value above the predetermined threshold.

[0047] In some embodiments, the processor 104 can be further configured to compute a representative score of a biopsy sample obtained from the tumor. The representative score can correspond to a probability that a biopsy sample obtained from the tumor has a value of the biological parameter above a predetermined threshold. Since the tumor activity can be correlated to the biological parameter of interest, the representative score can also correspond to a probability that a biopsy sample obtained from the tumor has a value of the biological parameter above a clinically relevant threshold. The processor can be configured to compute the probability based on the heterogeneity score and a volume of the biopsy sample.

[0048] In some embodiments, the processor 104 can analyze the microscopic image to determine a biological parameter associated with the tumor, e.g., an expression level of a protein of interest in the tissue. The processor can then determine a likelihood value or probability that a value of the biological parameter represents the tumor based on the heterogeneity score, the predetermined threshold, and / or the probability. In alternative embodiments, the analysis of the microscopic image can be performed manually, e.g., by a pathologist or other clinician. The clinician can also manually determine the probability that a value of the biological parameter represents the tumor.

[0049] After analyzing the received data, the processor 104 can output, e.g., the heterogeneity score, the representative score, and / or the probability to the I / O device 102 or another display unit. In some embodiments, the output can include a diagnostic and / or therapeutic recommendation, e.g., an indication that a further biopsy sample should be taken and analyzed.

[0050] Figure 2 A flowchart of a method 200 for determining tumor heterogeneity using a system for using Figure 1 radiological images according to one embodiment is depicted. Step 202 involves analyzing a radiological image of a tumor to compute a value indicative of tumor activity at each of a plurality of locations in the tumor. As used herein, “image” is to be interpreted broadly as including not only the entire array of image signals, but also, at the extreme case, a single image signal for a single voxel location or a selection of such image signals for a subset of voxel locations in the entire field of view.

[0051] The values indicative of tumor activity can be, for example, standard uptake values (SUV). For example, SUVs can be obtained by PET imaging using (18-F)fluorodeoxyglucose (FDG). In some embodiments, the SUVs can be calculated as an average or mean over a defined volume at each of a plurality of locations in the tissue. In such embodiments, each defined tissue volume in the sample is represented by a single value indicative of metabolic activity. In alternative embodiments, the radiological images can be analyzed to calculate SUV values at each individual voxel location in the tissue. In these embodiments, the SUV values at different locations in the tissue can be represented visually by images showing the distribution of values (e.g. SUVs) throughout the tissue.

[0052] The radiological images can be received by the interface 110 from the radiological imaging system 112. For example, the interface 110 can receive the images from the imaging system 112 via a network connection, such as a wired or wireless connection. In alternative embodiments, the interface 110 can receive the images from the imaging system 112 via a removable storage medium, such as a USB stick or a DVD. Figure 1 The processor of the processor 104 can receive these images from the interface 110. In alternative embodiments, the radiological images can be transferred to the interface 110 and / or the processor 104 by the I / O device 102, for example, after the imaging system has acquired the images, the images are stored in another location. Figure 1 The processor of the processor 104 can receive these images from the interface 110. In alternative embodiments, the radiological images can be transferred to the interface 110 and / or the processor 104 by the I / O device 102, for example, after the imaging system has acquired the images, the images are stored in another location.

[0053] Step 204 involves determining whether each calculated value (e.g. SUV at each location) is above a predetermined threshold value.

[0054] Step 206 involves calculating a heterogeneity score for the tumor. The heterogeneity score generally corresponds to the fraction of tumor volume having tumor activity values above the predetermined threshold value. The heterogeneity score can thus be obtained, for example, by dividing the number of calculated values (SUVs) determined to be above the threshold value by the total number of calculated values. In the case where SUVs are calculated for each voxel, this corresponds to the number of voxels above the threshold value divided by the total number of voxels within the depicted lesion volume.

[0055] In some embodiments, an additional step 208 can comprise calculating a representative score for a biopsy sample to be taken from the tumor. The representative score can correspond to the probability that the biopsy sample has a tumor activity value (e.g. mean) above a predetermined threshold value and / or the probability that the biopsy sample has a value of a biological parameter above a clinically relevant threshold value. This probability can be calculated based on the heterogeneity score and the known volume of the biopsy sample. For example, if the volume of the biopsy sample is relatively small compared to the whole tumor, the representative score will be similar to the heterogeneity score. As the volume of the biopsy sample increases, the representative score decreases, as the risk of including volume elements below the threshold value increases, which will cause the average of the tumor activity to decrease below the threshold value.

[0056] The method can also optionally include step 210, in which the microscopic images of the biopsy sample are analyzed to determine a value of a biological parameter associated with the tumor. For example, the processor 104 can use an algorithm suitable for analyzing digital microscopic images to identify stained objects and / or overall staining levels within the tissue, e.g., to identify expression levels of a protein that has been stained in the images by immunohistochemical processing of the biopsy sample. In one embodiment, the processor analyzes the images to determine the level of PD-L1 expression on tumor cells in the tissue. The output of this analysis can be a continuous variable (e.g., the percentage of PD-L1 positive tumor cells) or a binary result (e.g., a positive / negative score for PD-L1 based on expression above or below a defined threshold level).

[0057] The method can also optionally include step 212, in which the processor determines a likelihood value or probability that the value of the biological parameter represents the tumor. This likelihood value can be determined, for example, from the representative score, as the probability that the biopsy sample was obtained from a tumor region with tumor activity above a threshold is directly related to the likelihood that the value of the biological parameter represents the entire tumor. Alternatively, the likelihood value can be determined from the heterogeneity score, as this is associated with the representative score related to the volume of the biopsy sample.

[0058] The output of the method can be, for example, the heterogeneity score, the representative score, and / or an indication of the likelihood that the value of the biological parameter represents the tumor. One or more of these results can be output via the I / O device 102 or display.

[0059] In one embodiment, one or more of the radiology-derived features can be compared to one or more of the pathology-derived features to check for consistency. For example, the value of the biological parameter (obtained from analysis of the microscopic images) can be compared to the heterogeneity score (obtained from analysis of the radiological images). If a significant volume of the tumor is found to have tumor activity above a threshold (as indicated by the heterogeneity score), it can be expected that the value of the biological parameter (e.g., PD-L1 expression) would be above a clinically relevant threshold. If this is not the case, it can be flagged as an unexpected (or inconsistent) result and indicated to the clinician.

[0060] The method as a whole is directed to determining tumor heterogeneity and involves calculating a heterogeneity score for the tumor. A "heterogeneity score" in terms of, for example, functional activity and / or expression of one or more tumor markers refers to a score that correlates with spatial variability in tumor properties. In embodiments of the application, the heterogeneity score is represented by a volume fraction of the tumor having values indicative of tumor activity above a predetermined threshold. Thus, a high value of the score can indicate relatively low heterogeneity (i.e., tumor activity is relatively high and uniform throughout the tumor volume). Conversely, a low value of the score can indicate relatively high heterogeneity (i.e., a significant portion of the tumor volume shows relatively low activity compared to other high activity regions). Thus, the heterogeneity score can, for example, be inversely proportional to variability in tumor activity. In some embodiments, the heterogeneity score can thus alternatively be referred to as, for example, a tumor activity score, a tumor activity uniformity score, or a volume fraction tumor activity score above a threshold.

[0061] The radiological images (e.g., PET images) analyzed in step 202 shown in FIG. 2 can be obtained by any suitable radiological imaging method, for example, by positron emission tomography (PET), computed tomography (CT), magnetic resonance imaging (MRI), and / or single photon emission computed tomography (SPECT). Typically, the images are PET images. Figure 2

[0062] The values indicative of tumor activity in the tissue (e.g., determined by PET imaging) can, for example, be an indicator of metabolic activity of the tumor. "Metabolic activity" as used herein includes the chemical transformation (or processing) of certain compounds ("metabolites") by cells of the human or animal body. One important, but non-limiting, example is the transformation of glucose for energy production.

[0063] The TCA cycle is a key metabolic pathway that allows mammalian cells to utilize glucose, amino acids, and fatty acids. These fuels into the cycle are carefully regulated to efficiently meet the cell's bioenergetic, biosynthetic, and redox balance requirements. Cancer cells exhibit unique pathophysiology compared to normal cells, including: (a) autonomous mechanisms of cell growth, (b) divergence from growth inhibitory related factors, (c) evasion of anabiosis apoptosis, immune surveillance, and apoptosis, (d) evolutionary regulation of growth, (e) invasive and metastatic colonization.

[0064] ​In order to undergo replication and division, a cell must copy its genome, proteins, and lipids and assemble these elements into daughter cells. The increased rate of cell division in cancer requires the redesign of metabolic pathways, resulting in tumor cell metabolism. Reprogramming of glucose metabolism is a key event in tumorigenesis. Tumor cells undergo a metabolic shift from oxidative phosphorylation (OXPHOS) to glycolysis, in which a glucose molecule is degraded into two pyruvate molecules. Depending on the oxygen supply to the cell, pyruvate is either reduced to lactate via the anaerobic glycolysis pathway in the absence of oxygen, or oxidized in the presence of oxygen to produce acetyl-CoA, which is then completely oxidized to CO2 and H2O via the citric acid cycle. The growth and survival of most tumor cells depend on a high rate of glycolysis, even in the presence of sufficient oxygen. This aerobic glycolysis is known as the Warburg effect.

[0065] To support this sustained cell proliferation, the biosynthetic capacity of tumor cells is increased, including the synthesis of fatty acids and nucleotides. Conversely, β-oxidation of fatty acids is inhibited and futile cycles are minimized. These changes increase the metabolic autonomy of transformed cells, enabling them to acquire an enhanced anabolic phenotype. Therefore, metabolic activity in tissues can be used as a marker of tumor activity.

[0066] PET scans using the tracer fluorine-18 (18-F) fluorodeoxyglucose (FDG), known as FDG-PET, are widely used in clinical oncology. This tracer is a glucose analog that is taken up by glucose-using cells and phosphorylated by hexokinase (the mitochondrial form of which is greatly elevated in fast-growing malignancies). 18 F-FDG activity ranges from 200 to 450 MBq. Because the oxygen atom replaced by F-18 to generate FDG is required for the next step in glucose metabolism in all cells, no further response occurs in FDG. In addition, most tissues (except the liver and kidneys) cannot remove the phosphate added by hexokinase. This means that FDG is trapped in any cell that takes it up until it decays, because the phosphorylated sugars cannot be excreted from the cell due to their charge. This results in strong radiolabeling of tissues with high glucose uptake, such as the brain, myocardium, liver, and most cancers. Therefore, FDG-PET can be used for the diagnosis, staging, treatment planning, and monitoring of treatment of cancer, particularly Hodgkin's lymphoma, non-Hodgkin's lymphoma, breast cancer, melanoma, and lung cancer. Analysis of PET images is well established, and signal intensity, as shown by standardized uptake values ​​(SUV, SUVmax), is used for staging and prognosis.

[0067] Thus, in one embodiment, the value indicative of tumor activity is the SUV, e.g., the radiological image is obtained by PET imaging using 18F-FDG. Other suitable metrics derived from PET images that can be used to indicate tumor activity include, for example, SUVmax, SUVmean, metabolic tumor volume (MTV), total lesion glycolysis (TLG), or other higher-order so-called radiomic features. See, e.g., Paul E Kinahan et al in“PET / CT Standardized Uptake Values (SUVs) in Clinical Practice and Assessing Response to Therapy”, Semin Ultrasound CT MR, December 2010 December, vol 31(6), pp 496-505.

[0068] In other embodiments, additional biological parameters can be used to indicate tumor activity as an alternative to the parameter indicative of metabolic activity. In general, various functional and / or anatomic parameters can be assessed using radiological imaging methods to determine tumor activity. For example, specific PET tracers can be used to determine tumor proliferation and / or hypoxia as markers of tumor activity. In one embodiment, a value indicative of tumor cell proliferation in the tissue can be determined from images obtained using 18F-FLT PET imaging. In another embodiment, a value indicative of tumor cell hypoxia can be computed from images obtained using an appropriate tracer, e.g., using 18F-FMISO PET imaging.

[0069] In further embodiments, a marker indicative of tumor activity can be, e.g., PD-L1 expression determined in the tumor using radiological imaging. For example, PD-L1 expression in the tissue can be determined from PET images obtained using an anti-PD-L1 antibody PET tracer (e.g., 89Zr-atezolizumab, 18F-BMS-986192). In some such embodiments, the heterogeneity score can additionally be compared to a pathology score of expression of a biological parameter, such as PD-L1, measured in a biopsy sample from the tumor. In other words, the radiological (e.g., PET) score of PD-L1 expression is compared to a pathology score of PD-L1 expression. A significant difference between the two values indicates that the pathology-derived value can be incorrect / representative, and a further biopsy should be taken and analyzed.

[0070] In one embodiment, the value indicative of tumor activity can be computed by analyzing a combination of images obtained by PET (e.g. providing metabolic activity) and contrast-enhanced perfusion CT (e.g. providing information about angiogenesis and blood flow characteristics in the tumor). In a specific embodiment, CT perfusion data (e.g. blood flow (BF), blood volume (BV) and mean transit time (MTT)) are obtained by analysis of time-contrast enhanced curves. In one embodiment, the perfusion status of the tissue can also be derived from MR images.

[0071] The methods described herein in some embodiments (e.g. in step 210 shown in Figure 2 The methods described herein in some embodiments (e.g. in step 210 shown in

[0072] Tissue specimens (e.g. obtained by biopsy or resection from a subject) can be prepared using known techniques for optical microscopic analysis and imaging. For example, hematoxylin and eosin (H&E) staining of paraffin-embedded sections is the default technique for visualizing tissue on a glass slide for pathological analysis. Immunohistochemistry (IHC) staining is a well-known method for identifying overexpression of proteins on cells in pathological tissue sections. The staining results in a typical brown appearance of the tissue where the target protein is overexpressed compared to normal. For example, overexpression of programmed death ligand 1 (PD-L1) can be detected by using antibodies against it. The result is typically expressed as a so-called proportion score, i.e. the percentage of tumor cells that are designated positive (above a threshold value).

[0073] 3D images can be obtained from biopsy or resection tissue material, with or without staining, to provide an intermediate 3D representation of the biopsy morphology before treatment with paraffin and microscopic sectioning into thin microscope sections for pathological image acquisition. This can be done, for example, by optical coherence tomography, X-ray and / or multi-focal microscopy after clearing of the tissue with a refractive index matching substance.

[0074] Detection of features in the pathological sections can be performed using known computer algorithms that can analyze digital images of the sections. For example, a convolutional neural network can be trained by providing an annotated dataset of pathological images in which objects of interest have been manually annotated by a pathologist. Such objects can be, for example, cell nuclei. It can also be possible to successfully train a deep learning computer algorithm to classify cell nuclei as, for example, tumor cells or immune cells. Thus, in a particular embodiment, a deep learning computer algorithm can be trained and / or used to detect cell nuclei on a digital image of tissue, even to distinguish between tumor and non-tumor tissue.

[0075] Computer-based detection algorithms can also be used in combination with IHC to automatically detect cells of interest (e.g. overexpressing a particular protein). For example, the presence and abundance of cells classified as positive for overexpression of PD-L1 can be determined in IHC images. Using computer-based detection enables a true quantification of the number and density of objects in a region of interest (e.g. a tumor lesion).

[0076] In a preferred embodiment, the biological parameter detected in the biopsy sample is PD-L1 expression. PD-L1 is a cell membrane protein, one of the so-called checkpoint molecules, important for communication between tumor and immune cells. It can also be considered a marker for increased metabolic activity of tumor cells. PD-L1 expression on tumor cells can be constitutive (mutation driven) or induced (via immune attack). PD-L1 expression is typically determined using immunohistochemical staining assays on tissue sections. Several commercial assays are available for use (see e.g. F.R. Hirsch et al., J. Thorac. Onc., 12, 208). Overexpression of PD-L1 on tumor and in some cases immune cells is used as a diagnostic test to guide therapy selection. For example, patients with advanced lung cancer with PD-L1 positive tumor cells scoring more than 50% are eligible for checkpoint inhibition monotherapy, while patients with lower scores will be eligible for a combination of checkpoint inhibition and chemotherapy. It is known that expression of PD-L1 can be heterogeneous within and between lesions in a patient. Figure 3 An example of PD-L1 staining of a tumor biopsy tissue section is shown, indicating regions of very different PD-L1 expression.

[0077] In an embodiment, described herein is a method to obtain a representative score of a biopsy of a tumor based on analysis of radiological images, in particular FDG-PET / CT. The method can comprise the following steps:

[0078] 1. Acquire (FDG-)PET / CT images of a tumor lesion;

[0079] 2. Analyze the PET images:

[0080] a. Define a region of interest (select the contour line of the lesion);

[0081] b. Determine a parameter related to metabolic activity (e.g. SUV) of the region of interest;

[0082] c. Calculate a contour plot of iso-parametric volumes;

[0083] d. Calculate the percentage of lesion volume with parameter above a threshold;

[0084] 3. Calculate a score reflecting the probability of obtaining a biopsy that would represent a lesion fraction above a threshold value;

[0085] 4. Report the score to the radiologist and / or pathologist and / or treating physician;

[0086] 5. Optionally indicate if the score is below a predefined value.

[0087] In one embodiment, the volume percentage (step 2.d) can be calculated based on the minimum connected volume having a value above a threshold value. The volume can be defined relative to the volume of the core biopsy sample. The threshold value can be related to the volume average or the maximum value in a specific volume element.

[0088] In another embodiment, no threshold value is used in step (2.d) but the probability of a biopsy volume being removed that is equal or higher than the volume average of the total lesion is calculated (step 3).

[0089] In another embodiment, the local activity parameter derived from the local PET SUV value can be corrected by a local parameter reflecting tissue density obtained from a CT image of the same location. This correction can be performed by e.g. dividing the SUV value by the density parameter. In another embodiment, the local parameter reflecting tissue density can be determined by a suitable MR sequence.

[0090] If the biopsy location is known and can be registered to the radiological image, the probability of the biopsy being representative can be calculated with higher precision by calculating the volume fraction having a value that is the same or higher than the average of the local biopsy volume elements. If the biopsy location can be perfectly registered to the radiological image, assuming a known correlation of the biological parameter and the tumor activity, the value of the biological parameter obtained from the biopsy should correspond to the tumor activity value at that location. Typically, the biopsy volume is of the same order of magnitude as the resolution of the radiological (e.g. PET) image. Thus, if the heterogeneity score indicates that a high volume fraction of the tumor exceeds a predetermined threshold value of tumor activity, it is expected that a biopsy sample taken from most locations within the tumor shows a value of the biological parameter that is above its clinically relevant threshold value. However, if the value of the biological parameter obtained from the biopsy is negative (i.e. below the clinically relevant threshold value), despite a high volume fraction of the tumor being above the predetermined tumor activity threshold value obtained by radiology, this inconsistency indicates that the biological parameter can not be representative of the tumor.

[0091] The contour map as well as the probability of the biopsy being representative can be provided to the radiologist or the nuclear medicine physician, e.g. on a display. The same information can also be reported to the pathologist to be taken into account when making the pathology score, and to the treating physician. For example, a risk score derived from the probability can be displayed to the radiologist and / or pathologist and / or treating physician.

[0092] In a preferred embodiment, the contour plot can be used to guide the biopsy to the location of the largest connected volume with a parameter value larger than a threshold value.

[0093] In another embodiment, the local heterogeneity of a parameter can be computed for a lesion and optionally displayed in the enhanced radiological image. The local heterogeneity can be correlated with a histopathology parameter. This can be performed for all lesions of a patient with multiple lesions. In this way, an inter-lesion heterogeneity parameter can be defined.

[0094] The threshold value (2.d) can be pre-determined, e.g. from a cohort study, and optionally adjusted to the specific patient image acquisition conditions.

[0095] In one embodiment, FDG-PET and SUV are used in relation to PD-L1 expression in pathology. The steps of one such non-limiting embodiment are described below.

[0096] FDG-PET / CT images are acquired according to well-known clinical procedures and instruments. In a first step after image acquisition and standard image correction and post-processing, the lesion volume is defined as a contour in the 3D image. The SUV is computed for each voxel in the lesion image. For a predetermined (cut-off) SUV value, the volume elements of equal SUV value are determined and indicated by an iso-SUV contour in the image. A contour plot showing the volume elements of equal SUV value is shown, e.g., in Figure 4 A restriction can be applied to the minimum size of the volume elements indicating such created sub-volumes (minimum island size, granularity). Additionally, a histogram of the SUV distribution can be generated and from it a heterogeneity score can be computed, e.g., as shown in Figure 5

[0097] The SUV predetermined threshold value in relation to the correlation with a pathology parameter, e.g. PD-L1 expression on tumor cells, can be selected and indicated in the histogram and optionally, all volume elements with SUV equal or higher than the threshold value can be highlighted in the image. The volume fraction of the total lesion with SUV values higher than the threshold value can be represented as a score. This score is related to the unbiased probability of extracting a sample by a biopsy procedure featuring SUV values higher than the threshold value.

[0098] The volume fraction of the lesion higher than the threshold value and the location of the volume elements higher than the threshold value are reported.

[0099] This information can be used to instruct the clinician to guide the biopsy needle to the region with the highest SUV value and a connected volume significantly larger than the biopsy volume to be removed, if this region is reachable. If not, the best alternative region to guide the biopsy can be selected. At the same time, the volume fraction expected to be higher than the threshold value in the biopsy sample is computed. ​

[0100] In the case of an unblinded biopsy, information about the volume fraction above the threshold (i.e. the heterogeneity score) is reported to the pathologist. The pathologist (optionally using an automated image analysis system) can have determined a score for PD-L1 expression on tumor cells as a percentage of positive tumor cells (tumor proportion score or TPS). If this score is negative or below 50%, it can have an impact on the care path selected for the patient. If the TPS is below the threshold and the volume fraction above the SUV threshold in the PET image is relatively high (e.g. above 20%, the discordance between TPS and SUV indicates a risk that there is a non-representative result), the TPS score can be flagged as potentially false negative (or false below the 50% threshold). A re-biopsy can be recommended based on such a result. If this is not acceptable, the risk of a false negative PD-L1 score can be reported to the clinician and taken into account when selecting a therapy.

[0101] The method uses a predetermined threshold for the tumor activity value obtained from the radiological image to determine the heterogeneity score. Since the tumor activity parameter (e.g. based on SUV) is related to the biological parameter (e.g. PD-L1 score), a clinically relevant threshold for the biological parameter can be used to calculate the predetermined threshold for the tumor activity. As long as the relationship between the clinically relevant threshold for the biological parameter and the predetermined threshold for the tumor activity is known (e.g. whether the correlation is linear or non-linear), a threshold of e.g. 1% and 50% TPS can be converted to a corresponding SUV threshold, e.g. by a simple correlation equation that can be obtained from a one-off experimental correlation curve. The relationship between SUVs obtained e.g. by FDG-PET and PD-L1 expression is known (see e.g. Zhou et al. (2018), Eur J Nucl Med Mol Imaging 45 (Suppl 1): S441, EP-0367).

[0102] In a preferred embodiment, a correlation function is used to relate the predetermined threshold for the heterogeneity determination to the clinically relevant threshold for the biological parameter. In other words, the correlation function allows to select a suitable predetermined threshold for the tumor activity value (e.g. SUV) based on a known cut-off value of the biological parameter that provides a diagnostic or clinical indication (e.g. TPS indicating suitability for checkpoint inhibitor therapy).

[0103] The correlation function relates a local biological parameter from pathology to a computed local radiological signal. In one example, the biological parameter is the percentage of PD-L1 positive tumor cells in a tissue region and the tumor activity is determined as a local SUV from a FDG-PET image. For the clinical question, a specific percentage of PD-L1 positive tumor cells (the so-called tumor proportion score TPS) is relevant. From the correlation function, a corresponding SUV value can be derived at the selection of the clinical cut-off value.

[0104] For patients with advanced lung cancer, pembrolizumab checkpoint immunotherapy with a 50% TPS cutoff was used. Therefore, the SUV value corresponding to 50% TPS was selected as the threshold for calculating tumor heterogeneity. For the combination of pembrolizumab and chemotherapy, a 1% TPS cutoff was used. In this case, a different threshold for SUV was selected for calculating tumor heterogeneity. Therefore, the reported tumor heterogeneity and the corresponding risk with an unrelated biopsy result depend on the clinical threshold of the biological parameter involved. For the same patient, the heterogeneity score is different for 1% TPS and 50% TPS. In general, the higher the TPS threshold, the higher the corresponding predetermined threshold for SUV, and therefore the smaller the tumor lesion volume fraction with a SUV value equal to or higher than the predetermined threshold. The heterogeneity score determined using the two different thresholds will vary accordingly.

[0105] The methods, systems, and devices discussed above are examples. Various configurations can omit, substitute, or add various procedures or components. For instance, in alternative configurations, the methods can be performed in an order different from that described, and / or various steps can be added, omitted, or combined. Also, features described with respect to certain configurations can be combined in various other configurations. Different aspects and elements of the configurations can be combined in a similar manner. Also, technology evolves and, thus, many of the elements are examples and do not limit the scope of the disclosure or claims.

[0106] For example, embodiments of the present disclosure are described above with reference to block diagrams and / or operational illustrations of methods, systems, and computer program products according to embodiments of the present disclosure. The functions / acts noted in the blocks can occur out of the order noted in any flowchart. For example, two blocks shown in succession can in fact be executed substantially concurrently or the blocks can sometimes be executed in the reverse order, depending upon the functionality / acts involved. Additionally, or alternatively, not all of the blocks shown in any flowchart need be executed and / or performed to practice the present disclosure. For example, if a flowchart has five blocks, only three of the blocks can be executed and / or performed in some implementations. In this example, the three blocks executed and / or performed can be any three of the five blocks.

[0107] Specific details are given in the description to provide a thorough understanding of the example configurations (including implementations). The configurations can be practiced without these specific details. For example, well-known circuits, processes, algorithms, structures, and techniques have been shown without unnecessary detail in order to avoid obscuring the configurations. This description provides example configurations only, and does not limit the scope, applicability, or configurations of the claims. Rather, the preceding description of the configurations will provide those skilled in the art with an enabling description for implementing described techniques. Various changes can be made in the function and arrangement of elements without departing from the spirit or scope of the disclosure.

[0108] Having described several example configurations, modifications, alternative constructions, and equivalents can be used without departing from the spirit of the disclosure. For example, the above elements can be components of a larger system, wherein other rules can take precedence over or otherwise modify the application of the various implementations or techniques of the disclosure. Also, a number of steps can be undertaken before, during, or after the above elements are considered.

[0109] Having provided the description and drawings of this application, those skilled in the art can devise variations, modifications, and alternative embodiments of the general inventive concepts discussed herein without departing from the scope of the appended claims.

Claims

1. A computer-implemented method (200) for determining the representativeness of a pathology score for a tumor, the method comprising: analyzing a radiographic image of a tumor to calculate a value indicative of tumor activity at each of a plurality of locations in the tumor; determining whether each calculated value at each of the plurality of locations in the tumor is above a predetermined threshold to identify a total number of calculated values ​​above the predetermined threshold; as well as calculating a heterogeneity score for the tumor; wherein the heterogeneity score is calculated based on the volume fraction of the tumor having values ​​indicative of tumor activity above the predetermined threshold by dividing the total number of calculated values ​​above the predetermined threshold by the total number of locations in the tumor at which the calculated values ​​were determined; and A probability is calculated that a biopsy sample obtained from the tumor has a value for tumor activity above the predetermined threshold, wherein the probability is calculated based on the heterogeneity score and a volume of the biopsy sample. 2 . The method of claim 1 , wherein the predetermined threshold value of tumor activity is related to a clinically relevant threshold value of a biological parameter associated with the tumor. 3 . The method of claim 2 , wherein the biological parameter comprises the expression level of programmed death ligand 1 (PD-L1) on tumor cells.

4. The method of claim 2 or claim 3, wherein the clinically relevant threshold value of the biological parameter is the percentage of tumor cells positive for PD-L1 expression, indicating suitability for checkpoint inhibition therapy.

5. The method according to any one of claims 1 to 3, further comprising: comparing the heterogeneity score to a value of the biological parameter determined from a biopsy sample obtained from the tumor; wherein a discordance between the heterogeneity score and the value of the biological parameter indicates that the value of the biological parameter is not representative of the tumor.

6. The method of claim 5, wherein if the heterogeneity score indicates that a significant volume fraction of the tumor has a value indicative of tumor activity above the predetermined threshold, and the value of the biological parameter is below a clinically relevant threshold, the method further comprises indicating the value of the biological parameter as a potential false negative result.

7. The method according to any one of claims 1, 2, 3 and 6, wherein the radiological image is obtained by positron emission tomography (PET), magnetic resonance imaging (MRI), computed tomography (CT) and / or single photon emission computed tomography (SPECT) or any combination thereof.

8. The method of claim 7, wherein the value indicative of tumor activity in the tumor is a standardized uptake value (SUV) obtained from PET imaging.

9. The method according to claim 8, wherein the value indicative of tumor activity is obtained by PET imaging using fluorine-18 labeled fluorodeoxyglucose.

10. The method according to any one of claims 1, 2, 3, 6, 8 and 9, further comprising: A contour map is calculated, the contour map showing regions of the tumor having equal values ​​of tumor activity, wherein contour lines in the contour map correspond to the predetermined threshold value of tumor activity.

11. The method according to any one of claims 1, 2, 3, 6, 8 and 9, further comprising: Regions of the tumor from which a biopsy sample is to be extracted that have tumor activity above a predetermined threshold are indicated to the user.

12. The method according to any one of claims 1, 2, 3, 6, 8 and 9, wherein the volume fraction of the tumor is determined based on a minimum connected volume having a value above the predetermined threshold.

13. The method of any one of claims 1, 2, 3, 6, 8 and 9, wherein the value indicative of tumor activity is indicative of metabolic activity, hypoxia, tumor activation of a signaling pathway and / or proliferation in a tumor.

14. A system (100) for determining the representativeness of a pathology score for a tumor, the system comprising: an interface (110) for receiving a radiological image of a tumor; Memory (106); as well as A processor (104) configured to execute instructions stored on the memory to: (a) analyzing the radiographic image of the tumor to calculate a value indicative of tumor activity at each of a plurality of locations in the tumor; (b) determining whether each calculated value at each of the plurality of locations in the tumor is above a predetermined threshold to identify a total number of calculated values ​​above the predetermined threshold; as well as (c) calculating a heterogeneity score for the tumor; wherein the heterogeneity score is calculated based on the volume fraction of the tumor having values ​​indicative of tumor activity above the predetermined threshold by dividing the total number of calculated values ​​above the predetermined threshold by the total number of locations in the tumor at which the calculated values ​​were determined; and (d) calculating a probability that a biopsy sample obtained from the tumor has a value of tumor activity above the predetermined threshold, wherein the probability is calculated based on the heterogeneity score and the volume of the biopsy sample.

15. A computer program product comprising a non-transitory computer readable medium having computer readable code embodied therein, the computer readable code being configured such that, when executed by a suitable computer or processor, the computer or processor is caused to perform the method according to any one of claims 1 to 13.

Images