Use of artificial intelligence derived features of collagen architecture to make medical prognosis in cancer patients

Abstract

In some embodiments, the present disclosure relates to a method. The method includes accessing digitized pathology imaging data from a cancer patient. The digitized pathology imaging data includes segmented imaging data that identifies one or more regions of interest (ROI) including tumor-stroma regions and / or collagen fibers. A plurality of collagen fiber features are extracted from the one or more ROI. The plurality of collagen fiber features are provided to a machine learning model that is trained to generate a medical prognosis relating to the cancer patient.

Description

USE OF ARTIFICIAL INTELLIGENCE DERIVED FEATURES OF COLLAGEN ARCHITECTURE TO MAKE MEDICAL PROGNOSIS IN CANCER PATIENTSREFERENCE TO RELATED APPLICATION

[0001] This Application claims the benefit of U.S. Provisional Application No. 63 / 565,556, filed on March 15, 2024, the contents of which are incorporated by reference in their entirety.FEDERAL FUNDING INFORMATION

[0002] This invention was made with government support under R01 CA249992 and 1 R01 CA149992-01 A1 awarded by the National Institutes of Health / National Cancer Institute, under U01 CA269181 awarded by the National Cancer Institute, and under W81 XWH-22-1 -0236 and W81 XWH-21 -1 -0160 awarded by the Department of Defense. The government has certain rights in the invention.BACKGROUND

[0003] In recent years, there has been significant interest in developing machine vision tools for interrogating medical images. Machine vision tools are computer systems that utilize artificial intelligence to analyze medical images. Such systems have the potential to improve health care for patients in a wide range of medical fields.BRIEF DESCRIPTION OF THE DRAWINGS

[0004] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate various example operations, apparatus, methods, and other example embodiments of various aspects discussed herein. It will be appreciated that the illustrated element boundaries (e.g., boxes, groups of boxes, or other shapes) in the figures represent one example of the boundaries. One of ordinary skill in the art will appreciate that, in some examples, one element can be designed as multiple elements or that multiple elements can be designed as one element. In some examples, an element shown as an internal component of another element may be implemented as an external component and vice versa. Furthermore, elements may not be drawn to scale.

[0005] Fig. 1 illustrates some embodiments of a block diagram of a cancer assessment system configured to utilize collagen fiber features extracted from a digitized pathology image to make a medical prognosis.

[0006] Fig. 2 illustrates some additional embodiments of a block diagram of a cancer assessment system configured to utilize collagen fiber features extracted from a digitized pathology image to make a medical prognosis.

[0007] Fig. 3 illustrates a table showing exemplary multivariable survival analysis of the disclosed cancer assessment system applied to cancer patients having oral cancer.

[0008] Fig. 4 illustrates some additional embodiments of a block diagram of a cancer assessment system configured to utilize collagen fiber features extracted from a digitized pathology image to make a medical prognosis.

[0009] Fig. 5 illustrates a heat map showing an association between collagen fiber features and demographic variables, clinical variables, molecular subtypes, and genetic variables.

[0010] Fig. 6 illustrates a flow diagram showing some embodiments of a method of generating a medical prognosis using collagen fiber features extracted from a digitized pathology image.

[0011] Fig. 7 illustrates a workflow diagram corresponding to an exemplary method of calculating collagen fiber features within a digitized pathology image.

[0012] Fig. 8 illustrates a block diagram of some additional embodiments of a cancer assessment apparatus configured to make a medical prognosis using collagen fiber features extracted from a digitized pathology image.

[0013] Fig. 9 illustrates some embodiments of a block diagram of an apparatus configured to make a medical prognosis relating to a cancer patient using collagen fiber features extracted from a digitized pathology image.DETAILED DESCRIPTION

[0014] The description herein is made with reference to the drawings, wherein like reference numerals are generally utilized to refer to like elements throughout, and wherein the various structures are not necessarily drawn to scale. In the following description, for purposes of explanation, numerous specific details are set forth in order to facilitate understanding. It may be evident, however, to one ofordinary skill in the art, that one or more aspects described herein may be practiced with a lesser degree of these specific details. In other instances, known structures and devices are shown in block diagram form to facilitate understanding.

[0015] Cancer is a collection of diseases that arise when processes that control normal cell growth, division, and / or life span malfunction. Normally, human cells grow and multiply to form new cells as the body needs them. However, cancer causes this process to break down, so that abnormal or damaged cells grow and multiply when they shouldn’t. The unwanted growth of cancer cells can affect the normal functioning of surrounding cells and ultimately lead to death if not successfully treated.

[0016] It has been appreciated that collagen fibers in a tumor microenvironment play an important role in the development and progression of different types of cancers. For example, dysregulation of the extracellular matrix, a part of the tumor microenvironment including collagen fibers, is present in human papilloma virus (HPV) related oral cancers such as HPV induced oropharyngeal squamous cell carcinoma (OPSCC). Similarly, the inhibition of COL1A1 (collagen type I alpha 1 ) may significantly reduce cell proliferation and invasion in Kirsten rat sarcoma virus (KRAS)-mutated colon cancer cells, while higher COL1 A1 expression is linked to serosal invasion and metastasis. Because of the important role that collagen fibers play within a tumor microenvironment, computer extracted features describing an architecture of collagen fibers within a tumor microenvironment may be able to improve medical prognoses for cancer patients.

[0017] The present disclosure relates to a method and apparatus configured to generate a medical prognosis using collagen fiber features extracted from a digitized pathology image of a cancer patient (e.g., an oral cancer patient, a colorectal cancer patient, etc.). In some embodiments, the method may include accessing digitized pathology imaging data from a cancer patient. The digitized pathology imaging data may include segmented data that identifies one or more regions of interest (ROI) including tumor-stroma regions and / or collagen fibers. A plurality of collagen fiber features are extracted from the one or more ROI. The plurality of collagen fiber features are provided to a machine learning model that is trained to generate a medical prognosis relating to the cancer patient. Because collagen fibers play an important role in the development and procession of cancer, the plurality of collagenfiber features are able to be used by the trained machine learning model to generate a medical prognosis that has a high degree of accuracy {e.g., to generate a medical prognosis that is prognostic of survival with a hazard ratio of greater than or equal to 1.85).

[0018] Fig. 1 illustrates some embodiments of a block diagram of a cancer assessment system 100 configured to utilize collagen fiber features extracted from a digitized pathology image to make a medical prognosis.

[0019] The cancer assessment system 100 comprises a memory 101 configured to store digitized pathology imaging data 102. The digitized pathology imaging data 102 includes imaging data from a cancer patient. In various embodiments, the cancer patient may have colon cancer, oral cancer {e.g., head and neck carcinoma, oral cavity squamous cell carcinoma (OCSCC), human papilloma virus (HPV) induced oropharyngeal squamous cell carcinoma (OPSCC), and / or the like), breast cancer, cervical cancer, and / or the like.

[0020] In some embodiments, the digitized pathology imaging data 102 comprises one or more digitized pathology images {e.g., one or more digitized biopsy slides, one or more patches obtained from a digitized biopsy slide, etc.) obtained from a pathological tissue sample taken from the cancer patient and stored in the memory 101. In some embodiments, the digitized pathology imaging data 102 may include segmented imaging data that identifies one or more regions of interest (ROI) 104. For example, the digitized pathology imaging data 102 may include segmented imaging data that has been segmented to identify one or more ROI 104 including tumor-stroma regions 108 and / or collagen fibers 110 within a tumor microenvironment.

[0021] In some embodiments, the cancer assessment system 100 may include a segmentation tool 106 configured to access the digitized pathology imaging data 102. The segmentation tool 106 is further configured to identify tumor-stroma regions 108 and / or the collagen fibers 110 within a tumor microenvironment. The tumor-stroma regions 108 include non-cancer cell and non-immune cell components of tumors within the digitized pathology imaging data 102. In some embodiments, the tumor-stroma regions 108 may include structural components holding tumor tissue together, such as an extracellular matrix and specialized connective tissue cells, comprising fibroblasts and mesenchymal stromal cells. In some embodiments,the segmentation tool 106 may be configured to identify the collagen fibers 110 within the tumor-stroma regions 108. For example, the segmentation tool 106 may identify the tumor-stroma regions 108 within a tumor microenvironment and then subsequently identify the collagen fibers 110 within the tumor-stroma regions 108.

[0022] A feature extraction tool 1 12 is configured to extract a plurality of collagen fiber features 114 from the one or more ROI 104 (e.g., from the tumor-stroma regions 108 and / or the collagen fibers 110) of the digitized pathology imaging data 102. In some embodiments, the plurality of collagen fiber features 114 may describe an architecture (e.g., an arrangement and / or microscopic characteristics) of the collagen fibers 110 within the tumor-stroma regions 108.

[0023] The plurality of collagen fiber features 114 are provided to a machine learning model 116. The machine learning model 116 has been trained (e.g., using test data and / or validation data) to generate a medical prognosis 118 for the cancer patient using the plurality of collagen fiber features 114. In some embodiments, the medical prognosis 1 18 may comprise a classification of the cancer patient as being a low risk 120 or a high risk 122. The low risk 120 classification is associated with significantly better overall survival than the high risk 122 classification. In other embodiments, the medical prognosis 118 may relate to a treatment response of the cancer patient (e.g., a predicted response to chemotherapy, adjuvant chemotherapy, immunotherapy, MEK and SRC inhibitor therapies, and / or the like).

[0024] The medical prognosis 118 generated by the machine learning model 1 16 can be utilized by health care professionals to make a more informed decision relating to a treatment of the cancer patient, thereby allowing for the cancer patient to have an improved quality of life and / or a lower risk of death. For example, the medical prognosis may allow for health care professionals to make decisions that improve a likelihood of success for a treatment, that avoid negative side effects of drugs that are not likely to produce positive outcomes, that would more accurately decide if the cancer patient would benefit from post operative therapy, etc.

[0025] Fig. 2 illustrates some additional embodiments of a block diagram of a cancer assessment system 200 configured to utilize collagen fiber features extracted from a digitized pathology image to make a medical prognosis.

[0026] The cancer assessment system 200 comprises a memory 101 configured to store digitized pathology imaging data 102 for a cancer patient 202. In variousembodiments, the cancer patient 202 may have colon cancer, oral cancer (e.g., head and neck carcinoma, OCSCC, HPV induced OPSCC, and / or the like), breast cancer, cervical cancer, and / or the like. In some embodiments, the digitized pathology imaging data 102 comprises one or more digitized pathology images (e.g., one or more digitized biopsy slides) obtained from a pathological tissue sample taken from the cancer patient 202 and stored in the memory 101. In some embodiments, the memory 101 may comprise electronic memory (e.g., solid state memory, SRAM (static random-access memory), DRAM (dynamic random-access memory), and / or the like).

[0027] In some embodiments, the digitized pathology imaging data 102 may be generated by an image generation stage 203 that is configured to digitize a stained slide generated from a tissue sample taken from the cancer patient 202. The tissue sample may be from a colon, an oral cavity (e.g., from a soft palate, tonsil, tongue, etc.), breast tissue, cervical tissue, and / or the like. In some embodiments, the image generation stage 203 may comprise a tissue resection tool 204 (e.g., a scalpel, a needle, scissors, and / or the like) that is used to surgically excise tissue from the cancer patient 202. The tissue may be embedded in paraffin and then provided to a tissue sectioning and staining tool 206. The tissue sectioning and staining tool 206 is configured to slice the tissue into thin slices (e.g., using a microtome), which are placed on one or more transparent slides (e.g., one or more glass slides). The tissue on the one or more transparent slides is then stained to generate one or more tissue slides. The one or more tissue slides are subsequently converted to one or more digitized pathology images by a slide digitization tool 208 (e.g., comprising a CMOS image sensor, a CCD camera, and / or the like). In some embodiments, the digitized pathology imaging data 102 may comprise a whole slide image (WSI) of a H&E (Hematoxylin and Eosin) stained slide, a Masson’s trichome stained slide, or the like. In other embodiments, the digitized pathology imaging data 102 may comprise patches of a WSI, segmented images, and / or the like.

[0028] In some embodiments, a patch generator 210 is configured to access the digitized pathology imaging data 102. The patch generator 210 comprises an additional machine learning model (e.g., deep learning model) that is configured to identify tumor regions within the digitized pathology imaging data 102 (e.g., WSI) and to separate the tumor regions into a plurality of non-overlapping patches 212 (e.g.,tiles). The plurality of non-overlapping patches 212 may subsequently stored within the digitized pathology imaging data 102 and / or be provided to one or more downstream machine learning models to mitigate computation intensity. In some embodiments, the plurality of non-overlapping patches 212 may have a size of approximately 50 pixels x 50 pixels, 64 pixels x 64 pixels, or other similar values. In some embodiments, non-overlapping patches 212 that are extracted from the digitized pathology imaging data 102 and that contain less than approximately 50% of viable tissue may be discarded.

[0029] In some embodiments, a segmentation tool 106 is configured to access the digitized pathology imaging data 102 (e.g., the plurality of non-overlapping patches 212). The segmentation tool 106 is further configured to identify one or more regions of interest (ROI) 104. The one or more ROI 104 may include tumorstroma regions 108 and / or collagen fibers 1 10 within a tumor microenvironment in the digitized pathology imaging data 102. In some embodiments, the segmentation tool 106 may comprise a first segmentation stage and a second segmentation stage. The first segmentation stage is configured to perform a first segmentation process that segments the tumor-stroma regions 108 and epithelial regions within the tumor microenvironment (e.g., using morphological operations such as image dilation, erosion, and hole-filling). The second segmentation stage is configured to perform a second segmentation process that identifies the collagen fibers 1 10 within the tumorstroma regions 108 (e.g., using a Derivative of Gaussian model). In some embodiments, after segmentation is complete the segmentation tool 106 may be configured to store the segmented digitized pathology images in the memory 101 as the digitized pathology imaging data 102.

[0030] In some embodiments, the segmentation tool 106 is configured to generate one or more binary masks that comprise the tumor-stroma regions 108. In some such embodiments, the one or more binary masks comprise images having a value of “1 ” in image units (e.g., pixels, voxels, etc.) identified as being within the tumor-stroma regions 108 and having a value of “0” in image units outside of the tumor-stroma regions 108. In some embodiments, the segmentation tool 106 may comprise a machine learning model. In some embodiments, the machine learning model may comprise or be a HoVer-Net model. In some embodiments, the machine learning model may be implemented as computer code run on one or moreprocessors (e.g., a central processing unit including one or more transistor devices configured to operate computer code to achieve a result, a microcontroller, or the like).

[0031] A feature extraction tool 1 12 is configured to extract a plurality of collagen fiber features 114 from the one or more ROI 104. In some embodiments, the plurality of collagen fiber features 114 may comprise one or more of a collagen fiber fragmentation measure 214, a collagen fiber bundling percentage 216, a collagen fiber rigidity 218, a collagen fiber anisotropy index 220, and a collagen fiber density index 222. In some additional embodiments, the plurality of collagen fiber features 114 may comprise all of the collagen fiber fragmentation measure 214, the collagen fiber bundling percentage 216, the collagen fiber rigidity 218, the collagen fiber anisotropy index 220, and the collagen fiber density index 222. In some embodiments, the plurality of collagen fiber features 114 may have values that are average values derived from a plurality of patches. In some embodiments, the feature extraction tool 112 may be implemented as computer code run on one or more processors (e.g., a central processing unit including one or more transistor devices configured to operate computer code to achieve a result, a microcontroller, or the like).

[0032] The collagen fiber fragmentation measure 214 describes a breakdown or degradation of collagen fibers within the tumor-stroma regions 108. In some embodiments, the collagen fiber fragmentation measure 214 for one patch may be calculated by dividing a measure of a collagen fiber fragment length by an overall CF f collagen fiber lengths (e.g., Collagen fiber fragmentation measure = where CFf is the collagen fiber fragment length and CFI is the collagen fiber length). In some embodiments, the collagen fiber fragment length may be calculated as being equal to a sum of lengths of multiple different branches of a collagen fiber (e.g., for a collagen fiber with three branches the collagen fiber fragment length would be equal to a sum of the length of each of the three branches). In such embodiments, the collagen fiber fragmentation measure 214 may be determined by measuring lengths of different branches of a collagen fiber, adding the lengths of the different branches of the collagen fiber together to determine a collagen fiber fragment length, and then dividing the collagen fiber fragment length by a length of the collagen fiber (e.g., measured between opposing ends of the collagen fiber). It has been appreciatedthat as cancer progresses, the number and / or length of branches within a collagen fiber will increase, so that the collagen fiber fragmentation measure 214 is indicative of cancer progression.

[0033] The collagen fiber bundling percentage 216 describes a process of collagen fibers aligning and organizing themselves into tight bundles within the tumor-stroma regions 108. In some embodiments, the collagen fiber bundling percentage 216 for one patch may be calculated by dividing a collagen fiber bundle area by a number of collagen fiber bundles and multiplying the result by 100 (e.g., Collagen fiber bundling percentage = x 100, where CFba is the collagen fiber bundle area and CFb is the collagen fiber bundles).

[0034] The collagen fiber rigidity 218 describes a straightness (e.g., “noncurliness”) of collagen fibers within the tumor-stroma regions 108. In some embodiments, the collagen fiber rigidity 218 for one patch may be calculated by multiplying a thickness of the collagen fibers by an orientation of the collagen fibers (e.g., Collagen fiber rigidty measure = CFt x CFo, where CFt is the thickness of the collagen fibers and CFo is the orientation of the collagen fibers). In some embodiments, the collagen fiber rigidity 218 may be determined by multiplying a thickness of a collagen fiber by a trigonometric function (e.g., a sin, cosine, efc.) of an orientation of the collagen fiber. In some embodiments, a Hough transform may be used in determining an average orientation and / or thickness of collagen fibers. It has been appreciated that as cancer progresses, collagen fibers will become straighter (e.g., less curly), so that the collagen fiber rigidity 218 is indicative of cancer progression.

[0035] The collagen fiber anisotropy index 220 describes a directional preference or arrangement of collagen fibers along a certain direction within the tumor-stroma regions 108. In some embodiments, the collagen fiber anisotropy index 220 may be as Collagen fiber anisotropy index (CFdx, CFdyf where CFdx is a collagen fiber change in direction in the x-direction and Cfdy is a collagen fiber change in direction in the y-direction.

[0036] The collagen fiber density index 222 describes an abundance of collagen fibers within a specific volume or area of the tumor-stroma regions 108. In some embodiments, the collagen fiber density index 222 for one patch may be calculatedby multiplying a dividing an area occupied by collagen fibers in a tumor-stroma region by a total area of the tumor-stroma region (e.g.,Collagen fiber density index =wherein CFa is an area occupied by collagen fibers within a tumor-stroma region and TSa is a total area of the tumor-stroma region).

[0037] The plurality of collagen fiber features 114 are provided to a machine learning model 116 that has been trained to generate a medical prognosis 118. In various embodiments, the medical prognosis 118 may relate to overall survival of the cancer patient, a treatment response of the cancer patient, and / or the like. In some embodiments, the machine learning model 116 may comprise one or more of a Cox proportional hazards model, a random forest classifier, a support vector machine, a Naive Bayes classifier, and / or the like. The machine learning model 116 may be implemented as computer code run on one or more processors (e.g., a central processing unit including one or more transistor devices configured to operate computer code to achieve a result, a microcontroller, a graphics processing unit (GPU), and / or the like).

[0038] By generating the medical prognosis 118 using the plurality of collagen fiber features 114, the disclosed cancer assessment system 200 is able to achieve a highly accurate medical prognosis 118. For example, it has been appreciated that collagen fibers are amongst the amplest components of extracellular matrix within the tumor-stroma regions. Because dysregulation of an extracellular matrix is often present in oral cancers (e.g., in HPV related cancers), the plurality of collagen fiber features may be used by the machine learning model 116 to make an accurate prognosis relating to the survival of oral cancer patients. Furthermore, using the plurality of collagen fiber features 114 to generate the medical prognosis 118 gives the medical prognosis 118 a higher level of explainability than black box models (e.g., deep learning models), thereby potentially improving a confidence of health care professionals relying upon the medical prognosis 118.

[0039] In some embodiments, the machine learning model 116 may be configured to perform survival analysis to generate a risk score 224 corresponding to a risk of death for the cancer patient 202. The machine learning model 116 is further configured to compare the risk score 224 to a threshold 226 to generate the medical prognosis 118 to identify the cancer patient 202 as a low risk 120 (e.g., a patienthaving a low risk of death) or a high risk 122 (e.g., a patient having a high risk of death). In some embodiments, the machine learning model 116 may be configured use the risk score 224 to stratify risk within individual cancer stages (e.g., AJCC (American Joint Committee on Cancer) cancer stages II, III, and IV). In other embodiments, the machine learning model 1 16 may be configured to use the risk score 224 to stratify risk within all cancer stages combined.

[0040] Fig. 3 illustrates a table 300 showing exemplary multivariable survival analysis of the disclosed cancer assessment system applied to cancer patients having oral cancer.

[0041] Table 300 illustrates a hazard ratio with a 95% confidence interval and a p- value for different variables e.g., age, sex, smoking status, T-stage, N-stage, collagen fiber features (high risk vs. low risk)) for overall survival and for disease free survival. As can be seen in table 300, the disclosed cancer assessment system provides for a larger hazard ratio than other variables in determining overall survival and a comparable hazard ratio as other variables in disease free survival. For example, the disclosed cancer assessment system is able to achieve a higher hazard ratio for overall survival than age, sex, smoking status, and cancer staging.

[0042] Furthermore, the disclosed cancel assessment system is able to be highly prognostic of overall survival and disease free survival independent of T / N stages, age, sex, and smoking status with HR=1.99 (95% Cl=1 .46-2.72, p=0.00001 ) for overall survival and HR=1.73 (95% Cl=1 .21 -247, p=0.002) for disease free survival. Therefore, the disclosed cancer assessment system improves a computer’s ability to analyze medical images in a manner that accurately identify cancer patients that are likely to experience overall survival and / or disease free survival over other variables that have been used in the past. The improved ability to accurately identify cancer patients that are likely to experience overall survival and / or disease free survival can improve treatment of the cancer patients.

[0043] Fig. 4 illustrates some additional embodiments of a block diagram of a cancer assessment system 400 configured to utilize collagen fiber features extracted from a digitized pathology image to make a medical prognosis.

[0044] The cancer assessment system 400 comprises a memory 101 configured to store digitized pathology imaging data 102 from a cancer patient 202. The digitized pathology imaging data 102 may be segmented to identify one or more ROI104. In some embodiments, the memory 101 may be further configured to store additional patient information 402. The additional patient information 402 may include demographic variables 404 (e.g., age, race, gender, etc.) and / or clinical variables 406 (e.g., clinical stages).

[0045] A feature extraction tool 1 12 is configured to operate upon the one or more ROI 104 within the digitized pathology imaging data 102 to extract a plurality of collagen fiber features 114, which describe an architecture of collagen fibers within a tumor microenvironment. The plurality of collagen fiber features 114 are provided to a machine learning model 116 that has been trained to generate a medical prognosis 118. In some embodiments, the machine learning model 116 may be trained to provide different weightings to one or more of the plurality of collagen fiber features 114. In some embodiments, the machine learning model 116 may be further trained to utilize the additional patient information 402 to determine different weightings that are applied to one or more of the plurality of collagen fiber features 114. For example, the machine learning model 116 may be trained to utilize a first set of weightings for a first cancer patient that is less than 65 years old and a second set of weightings for a second cancer patient that is older than 65 years old. By utilizing the additional patient information 402 to provide the different weightings to one or more of the plurality of collagen fiber features 114, the machine learning model 116 can improve a prognostic ability of the medical prognosis 118.

[0046] In some embodiments, the machine learning model 1 16 may be able to use one or more of the plurality of collagen fiber features 114 to generate a biological variable estimate 408. In various embodiments, the biological variable estimate 408 may comprise an estimate of a genomic mutation status (e.g., a Kirsten rat sarcoma virus (KRAS) genetic mutation, a BRAF genetic mutation, a NRAS genetic mutation, etc.), a molecular subtype classification (e.g., CMS1-CMS4), and / or clinical status (e.g., overall stage, TNM stage, etc.) of a cancer patient.

[0047] It has been appreciated that it is possible to generate the biological variable estimate 408 using the plurality of collagen fiber features 114 since different collagen fiber features have a strong correlation (e.g., indicated by mean area under curve (AUC) values of greater than or equal to 0.7) with demographic variables, clinical variables, molecular subtypes and genetic variables. For example, Fig. 5 illustrates a heat map 500 showing an association between collagen fiber featuresand demographic variables, clinical variables, molecular subtypes and genetic variables.

[0048] The heat map 500 illustrates clinical variables on the x-axis and collagen fiber features on the y-axis. As can be seen in the heat map, there are a number of collagen fiber features that have a high hazard ratio. For example, CF rigidity has a high HR for genetic mutation variables, thereby showing that CF rigidity is particularly relevant for identifying genetic mutation variables. The high hazard ratio shows that collagen fiber features may be able to distinguish patient subgroups based on demographic, clinical, molecular, and / or genetic characteristics.

[0049] Referring again to Fig. 4, in some embodiments the disclosed machine learning model 116 may be configured to generate the medical prognosis 1 18 that is specific to one or more biological variables. For example, it has been appreciated that combining MEK and SRC inhibitor therapy (e.g., which combines drugs targeting both the MEK (Mitogen-activated protein kinase) and SRC (Src tyrosine kinase) pathways) may be an effective strategy for targeting RAS-mutant colon cancer, especially in specific CMS subtypes. Typically, RAS mutations are identified using genomic testing. However, due to the strong correlation between different collagen fiber features and genetic variables, the biological variable estimate 408 may be formed to comprise a RAS mutation status. The generation of a biological variable estimate 408 that identifies a RAS genetic mutation can allow for the disclosed machine learning model 116 to generate a medical prognosis 118 that includes a treatment response identifying a cancer patient as being likely to benefit from combined MEK and SRC inhibitor therapy or not likely to benefit from combined MEK and SRC inhibitor therapy.

[0050] Therefore, the disclosed cancer assessment system 400 can determine the biological variable estimate 408 directly from a cancer patient’s digitized pathology images. While genomic and / or molecular testing can be expensive and may not be widely accessible in some geographic areas (e.g. middle to low income countries), the disclosed use of collagen fiber features 114 to generate the biological variable estimate 408 provides for a highly accessible and cost-effective alternative to genomic testing. The more accessible and cost-effective alternative to genomic testing may be significant in low-to middle income countries, where negativerepercussions of cancer may be more pronounced (e.g., where a majority of colon cancer deaths occur).

[0051] Fig. 6 illustrates a flow diagram showing some embodiments of a method 600 of generating a medical prognosis using collagen fiber features extracted from a digitized pathology image.

[0052] While the disclosed method 600 is illustrated and described herein as a series of acts or events, it will be appreciated that the illustrated ordering of such acts or events are not to be interpreted in a limiting sense. For example, some acts may occur in different orders and / or concurrently with other acts or events apart from those illustrated and / or described herein. In addition, not all illustrated acts may be required to implement one or more aspects or embodiments of the description herein. Further, one or more of the acts depicted herein may be carried out in one or more separate acts and / or phases.

[0053] At act 602, digitized pathology imaging data from a cancer patient is generated. In some embodiments, the digitized pathology imaging data may be generated according to acts 604-612.

[0054] At act 604, a digitized pathology image is obtained from a pathology sample taken from a cancer patient. In some embodiments, the cancer patient may have or be suspected to have colon cancer, breast cancer, oral cancer (e.g., oral cavity squamous cell carcinoma, human papilloma virus (HPV) induced oropharyngeal squamous cell carcinoma, etc.), and / or the like. In some embodiments, the digitized imaging data may be formed

[0055] At act 606, the digitized pathology image is segmented to identify a tumor stroma region.

[0056] At act 608, the tumor stroma region is separated into a plurality of patches.

[0057] At act 610, the tumor stroma region is segmented to generate segmented digitized imaging data that identifies one or more regions of interest (ROI) including collagen fibers within the tumor stroma region.

[0058] At act 612, the segmented digitized imaging data is saved within electronic memory.

[0059] At act 614, the digitized pathology imaging data from the cancer patient, which has been segmented to identify ROI including tumor-stroma regions and / or collagen fibers within tumor microenvironment, is accessed.

[0060] At act 616, a plurality of collagen fiber features are extracted from the one or more ROI (e.g., tumor-stroma regions and / or collagen fibers). In some embodiments, the plurality of collagen fiber features may comprise one or more of a collagen fiber (CF) fragmentation measure, a CF bundling percentage, a CF rigidity measure, a CF anisotropy index, and a CF density index.

[0061] At act 618, a machine learning model is operated onto the plurality of collagen fiber features to generate a medical prognosis. In some embodiments, the machine learning model may generate the medical prognosis according to acts 620- 622.

[0062] At act 620, the machine learning model is operated upon the plurality of collagen fiber features to generate a biological variable estimate.

[0063] At act 622, the medical prognosis is generated based upon the biological variable estimate.

[0064] At act 624, a treatment may be provided to the cancer patient based upon the medical prognosis, in some embodiments. In various embodiments, the treatment may comprise chemotherapy, immunotherapy, adjuvant chemotherapy, MEK and SRC inhibitor therapy, and / or the like.

[0065] Therefore, the disclosed method 600 utilizes collagen fiber features extracted from tumor-stroma regions and / or collagen fibers of a cancer patient to make a medical prognosis.

[0066] It will be appreciated that the disclosed methods and / or block diagrams may be implemented as computer-executable instructions, in some embodiments. Thus, in one example, a computer-readable storage device (e.g., a non-transitory computer-readable medium) may store computer executable instructions that if executed by a machine (e.g., computer, processor) cause the machine to perform the disclosed methods and / or block diagrams. While executable instructions associated with the disclosed methods and / or block diagrams are described as being stored on a computer-readable storage device, it is to be appreciated that executable instructions associated with other example disclosed methods and / or block diagramsdescribed or claimed herein may also be stored on a computer-readable storage device.

[0067] Fig. 7 illustrates a workflow diagram 700 corresponding to an exemplary method of calculating collagen fiber features within a digitized pathology image.

[0068] As shown in workflow diagram 700, a whole slide image 702 is obtained from a cancer patient (e.g., a colon cancer patient). In various embodiments, the whole slide image 702 may be stained with H&E (Hematoxylin and Eosin), Masson’s trichome staining, or the like. As shown in image 704, the whole slide image 702 is overlaid with a mask identifying a tumor region 706 and a non-tumor region 708. The tumor region 710 is separated into a plurality of tiles 712 (e.g., patches) comprising tumor stroma, which are shown in image 714. Image 716 shows that collagen fibers within a stroma region of a tile are detected within the respective tiles. In some embodiments, the collagen fibers may be detected using a derivative-of Gaussian (DtG) based model. In some embodiments, the DtG model may classify each pixel of the tiles into one of seven image structures: flat, slope, blob (dark or light), line (dark or light), and saddle. Pixels identified as a dark line structure may subsequently be isolated to detect collagen fibers within the tiles.

[0069] Images 718-726 illustrate different collagen fiber features. Image 718 shows the identification of collagen fiber fragments. Image 720 shows the formation of collagen fiber bundles. Image 722 shows the identification of collagen fiber rigidity. Image 724 shows the identification of collagen fiber anisotropy. As the anisotropy values increase, the colormap transitions to a different color at the highest end of the scale (corresponding to a well aligned or anisotropic orientation). Image 726 shows the identification of a collagen fiber density. Different density values are represented as different colors.

[0070] Fig. 8 illustrates a block diagram of some additional embodiments of a cancer assessment apparatus 800 configured to make a medical prognosis using collagen fiber features extracted from a digitized pathology image.

[0071] The cancer assessment apparatus 800 comprises a memory 101 configured to store digitized pathology imaging data 102 including a plurality of digitized pathology images from cancer patients (e.g., oral cancer patients, colorectal cancer patients, breast cancer patients, cervical cancer patients, and / or the like). In various embodiments, the plurality of digitized pathology images may be obtained byan image generation stage 203 and / or from an on-line database 802 and / or archive containing digitized pathology images from patients generated at different sites (e.g., different hospitals, research laboratories, and / or the like). Prior to including digitized pathology images within the digitized pathology imaging data 102, the digitized pathology images may be subjected to a pre-processing stage 804. The preprocessing stage 804 may be configured to discard images that have defects, normalize image characteristics (e.g., color, brightness, contrast, etc.) so as to mitigate batch effects (e.g., differences between images obtained from different sites), and / or the like.

[0072] In some embodiments, the digitized pathology imaging data 102 may include a training set 102t and a validation set 102v. The training set 102t comprises digitized pathology images from a first plurality of patients. The validation set 102v comprises digitized pathology images from a second plurality of patients. In some embodiments, the memory 101 may also be configured to store ground truth segmentation data (e.g., segmentation results provided by an expert human pathologist).

[0073] In some embodiments, the training set 102t may be used to train a downstream segmentation tool 106 to perform segmentations that identify tumorstroma regions 108 and / or collagen fibers 110 within a tumor microenvironment. The training set 102t may also be used to train a downstream machine learning model 116 to generate a medical prognosis 118 relating to a cancer patient 202.

[0074] The validation set 102v may be used to validate the results of the segmentation tool 106 to perform segmentations that identify tumor-stroma regions 108 and / or collagen fibers 110. The validation set 102v may also be used to validate the results of the machine learning model 116 and the medical prognosis 1 18.

[0075] In some embodiments, machine learning model 116 may include a feature selection element 806 configured to select a set of most prognostic collagen fiber features to generate the medical prognosis 118. For example, the features extraction tool 1 12 may extract a first number of collagen fiber features and then the feature selection element 806 may select a smaller second number of the collagen fiber features (from the first number of collagen fiber features) that are most prognostic (e.g., that have a most significant impact in determining survival). In some embodiments, the second number of collagen fiber features may be used totrain and validate the machine learning model 116. In some embodiments, the machine learning model 116 may comprise a Cox regression model (e.g., a Cox proportional hazards model). In some embodiments, the Cox regression model may comprise a LASSO (least absolute shrinkage and selection operator) algorithm (e.g., a LASSO Cox regression model) that is configured to operate as the feature selection element 806.

[0076] In some embodiments, the machine learning model 116 may be configured to generate a risk score 224 related to survival of the cancer patient 202. In some embodiments, a median risk score (e.g., a median of a plurality of risk scores) obtained by the machine learning model 116 for the training set 102t may be used for risk stratification in the validation set 102v. For example, in some embodiments, the machine learning model 1 16 may operate upon a first plurality of collagen fiber features extracted from digitized pathology images within the training set 102t to determine a plurality of risk scores and to subsequently determine a median risk score. The median risk score may be subsequently set as a threshold 226 that is configured to distinguish between low risk and high risk patients. The machine learning model 116 is then operated upon a second plurality of collagen fiber features extracted from one or more digitized pathology images within the validation set 102v to determine risk scores 224 associated with one or more patients. The risk scores 224 are compared to the threshold 226 (e.g., the median risk score) to classify the one or more patients as low risk or high risk.

[0077] In some embodiments, the machine learning model 116 may be configured to determine correlations between certain features of the collagen fiber features 114 and one or more biological and / or clinical variables 808. In various embodiments, the one or more biological and / or clinical variables 808 may include one or more of a demographic variable 810 (e.g., age, race, sex, etc.), a clinical variable 812 (e.g., overall stage, TNM stage, etc.), a molecular subtype classification 814 (e.g., CMS1-CMS4), and / or a genomic mutation status 816 (e.g., a Kirsten rat sarcoma virus (KRAS) genetic mutation, a BRAF genetic mutation, a NRAS genetic mutation, etc.) of a cancer patient. In some embodiments, the one or more biological and / or clinical variables 808 may have binary values (e.g., an age may have a value of ‘0’ if below 65 and a value of ‘1’ if above 65). It has been appreciated that certain features of the collagen fiber features 1 14 have a high correlation with the one ormore biological and / or clinical variables 808. Therefore, by identifying correlations between certain features of the collagen fiber features 114 and one or more biological and / or clinical variables 808, the machine learning model 116 may be able to subsequently identify the biological and / or clinical variables 808 within a cancer patient based upon the certain features of the collagen fiber features 114. This can enable the disclosed cancer assessment apparatus 800 to offer a more accessible and cost-effective alternative to genomic, molecular, and / or clinical testing.

[0078] Fig. 9 illustrates some embodiments of a block diagram of an apparatus 900 configured to make a medical prognosis relating to a cancer patient using collagen fiber features extracted from a digitized pathology image.

[0079] The apparatus 900 comprises a cancer assessment apparatus 902. The cancer assessment apparatus 902 is coupled to an image generation stage 203, which is configured to generate a digitized pathology imaging data of tissue samples collected from a cancer patient 202.

[0080] The cancer assessment apparatus 902 comprises a processor 906 and a memory 904. The processor 906 can, in various embodiments, comprise circuitry such as, but not limited to, one or more single-core or multi-core processors. The processor 906 can include any combination of general-purpose processors and dedicated processors (e.g., graphics processors, application processors, etc.). The processor(s) 906 can be coupled with and / or can comprise memory (e.g., memory 904) or storage and can be configured to execute instructions stored in the memory 904 or storage to enable various apparatus, applications, or operating systems to perform operations and / or methods discussed herein.

[0081] The memory 904 can be configured to store digitized pathology imaging data 102. The digitized pathology imaging data 102 may comprise digitized biopsy images having a plurality of pixels, each pixel having an associated intensity. In some additional embodiments, the digitized pathology imaging data 102 may be stored in the memory 904 as one or more training sets of digitized images for training a classifier and / or one or more test sets (e.g., validation sets) of digitized images.

[0082] The cancer assessment apparatus 902 also comprises an input / output (I / O) interface 908 (e.g., associated with one or more I / O devices), a display 910, and an interface 912 that connects the processor 906, the memory 904, and the I / O interface 908. The I / O interface 912 can be configured to transfer data between thememory 904, the processor 906, and external devices, for example, the image generation stage 203.

[0083] In some embodiments, the cancer assessment apparatus 902 may further comprise one or more circuits 914 that include one or more of a segmentation circuit 916, a collagen fiber (CF) feature extraction circuit 918, and a machine learning circuit 920. In some embodiments, the one or more circuits 914 may operate according to machine learning algorithms stored in the memory 904.

[0084] In some embodiments, the segmentation circuit 916 is configured to segment the plurality of digitized pathology images to determine one or more regions of interest (ROI) 104 that identify tumor-stroma regions and / or collagen fibers within a tumor microenvironment. The CF feature extraction circuit 918 is configured to extract a plurality of collagen fiber features 114 from the one or more ROI 104. The machine learning circuit 920 is configured to utilize the plurality of collagen fiber features 114 to generate a medical prognosis 118 for the cancer patient 202. In some embodiments, the display 910 is configured to output or display the medical prognosis 118 generated by the cancer assessment apparatus 902.1stExample use case:

[0085] Background: Human papilloma virus (HPV) induced oropharyngeal squamous cell carcinoma (OPSCC) is a common subtype of head and neck carcinoma. Dysregulation of the extracellular matrix (ECM), an important part of the tumor microenvironment (TME), is shown in HPV related cancers. We studied collagen fibers (CF), the amplest component of the ECM, and evaluated whether their characteristics within the TME can offer a prognostic value in OPSCC.

[0086] Methods: Whole slide images (WSIs) of OPSCC patients were obtained from Washington University (WU, n=107), Houston Veterans Administration Hospital (HV, n=94), Cleveland Clinic (CC, n=336), John Hopkins University (JH, n=121 ) and Vanderbilt Medical Centre (VM, n=158). For the identified CF in tumorstroma regions of WSIs, the extracted CF features included: CF fragmentation measure, CF bundling percentage, CF rigidity measure, CF anisotropy index, and CF density index. For survival analysis using overall survival (OS) and Disease-Free Survival (DFS) as endpoints, the median risk score in the training set (WU) was applied for risk stratification in the validation sets (HV, CC, JH, VM) by means of aLeast Absolute Shrinkage and Selection Operator-Cox regression model. Multivariable Cox Proportional Hazards Regression was done for both OS and DFS.

[0087] Results: The patients defined as “high risk” had worse OS than those identified as “low risk” (HV(p=0.03), CC(p=0.03), JH(p=0.02), VM(p=0.04)). Multivariable analysis showed that CF features were prognostic independent of T / N stages, age, sex, and smoking status with HR=1 .99 (95% Cl= 1 .46-2.72, p=0.00001 )for OS and HR=1 .73 (95% Cl= 1 .21 -247, p=0.002) for DFS.

[0088] Conclusion: Artificial intelligence derived features of CF architecture can help identify high vs low risk patients in HPV related OPSCC. These findings could impact future treatment paradigms.2ndExample use case:

[0089] Background: Oral cavity squamous cell carcinoma (OCSCC), the most common head and neck cancer, exhibit low survival rates despite aggressive treatment, with patients in the advanced stage of the disease, as classified by the American Joint Committee on Cancer (AJCC), having significantly worse prognosis than early-stage patients. Collagen fibers (CF) in the tumor microenvironment (TME) are fundamental in tumor development and progression and extensive research has highlighted CF features as effective prognostic indicators in various cancers. Evaluating the arrangement and microscopic characteristics of CF within the TME can provide predictive and prognostic value in OCSCC patients.

[0090] Methods: Whole slide images (WSIs) obtained from surgically resected OCSCC patients from Vanderbilt University (VU, n=173) and The Cancer Genome Atlas (TCGA, n=275) were used for this study. CF features like CF fragmentation measure, CF bundling percentage, CF anisotropy index, CF density index and CF rigidity measure were measured from the identified CF in the tumorstroma regions of the WSIs. Using overall survival (OS) as the endpoint, a Least Absolute Shrinkage and Selection Operator-Cox regression model selected relevant features to generate risk scores for each patient. The median risk score in the training set (VU) was applied for risk stratification in the validation dataset (TCGA), for the individual AJCC (American Joint Committee on Cancer) stages (i.e., stages II, III and IV), as well as for all stages combined. Multivariable analysis was done using Cox Proportional Hazards Regression.

[0091] Results: In the TCGA cohort, the patients defined as “high risk” had significantly worse OS than those identified as “low risk” (p=0.01 ). The model was also prognostic of OS in Stages II (p=0.04), I ll(p=0.04) and IV(p=0.03). Multivariable survival analysis showed that our model was prognostic independent of T / N stages, age, sex, and race with HR=1.85 (95% 01=1.14-3.01 , p=0.01 ).

[0092] Conclusion: Al extracted features of CF architecture are associated with prognosis of OCSCC patients. This finding could be helpful in identifying patients, especially those in the early stages of the disease, who would need more aggressive postoperative therapy. Additional independent validation of these findings is warranted in multi-site and in stage I patients.

[0093] Therefore, the present disclosure relates to a method and apparatus configured to generate a medical prognosis using collagen fiber features extracted from a digitized pathology image of a cancer patient.

[0094] In some embodiments, the present disclosure relates to method that includes accessing digitized pathology imaging data from a cancer patient, the digitized pathology imaging data including segmented imaging data that identifies one or more regions of interest (ROI) including tumor-stroma regions and / or collagen fibers; extracting a plurality of collagen fiber features from the one or more ROI; and providing the plurality of collagen fiber features to a machine learning model that is trained to generate a medical prognosis relating to the cancer patient. In some embodiments, the method further includes accessing one or more digitized pathology images from the cancer patient; and performing a segmentation process on the one or more digitized pathology images to identify the one or more ROI. In some embodiments, the plurality of collagen fiber features include one or more of a collagen fiber fragmentation measure, a collagen fiber bundling percentage, a collagen fiber rigidity, a collagen fiber anisotropy index, and a collagen fiber density index. In some embodiments, the plurality of collagen fiber features include a collagen fiber fragmentation measure, a collagen fiber bundling percentage, a collagen fiber rigidity, a collagen fiber anisotropy index, and a collagen fiber density index. In some embodiments, the plurality of collagen fiber features consist of a collagen fiber fragmentation measure, a collagen fiber bundling percentage, acollagen fiber rigidity, a collagen fiber anisotropy index, and a collagen fiber density index. In some embodiments, the machine learning model is configured to generate a risk score using the plurality of collagen fiber features and to compare the risk score to a threshold to determine the medical prognosis. In some embodiments, the method further includes operating the machine learning model on the plurality of collagen fiber features to generate a biological variable estimate; and generating the medical prognosis using the biological variable estimate. In some embodiments, the biological variable estimate is a determination if the cancer patient has a Kirsten rat sarcoma virus (KRAS) genetic mutation.

[0095] In other embodiments, the present disclosure relates to a non-transitory computer-readable medium storing computer-executable instructions that, when executed, cause a processor to perform operations, including accessing digitized pathology imaging data from a cancer patient, the digitized pathology imaging data including one or more digitized pathology images that have been segmented to identify one or more regions of interest (ROI) including one or more tumor-stroma regions and collagen fibers; extracting a plurality of collagen fiber features from the one or more ROI, wherein the plurality of collagen fiber features comprise one or more of a collagen fiber fragmentation measure, a collagen fiber bundling percentage, a collagen fiber rigidity, a collagen fiber anisotropy index, and a collagen fiber density index; and operating a machine learning model to use the plurality of collagen fiber features to generate a medical prognosis relating to the cancer patient. In some embodiments, the medical prognosis relates to an overall survival of the cancer patient or a treatment response. In some embodiments, the operations further include accessing the one or more digitized pathology images from the cancer patient; performing a first segmentation process on the one or more digitized pathology images to identify the one or more tumor-stroma regions; and performing a second segmentation process on the one or more digitized pathology images to identify the collagen fibers within the one or more tumor-stroma regions. In some embodiments, the tumor-stroma regions are identified using morphological operations and the collagen fibers are identified using a Derivative of a Gaussian model. In some embodiments, the cancer patient has colon cancer. In some embodiments, the cancer patient has human papilloma virus (HPV) induced oropharyngeal squamous cell carcinoma. In some embodiments, the plurality ofcollagen fiber features comprise the collagen fiber fragmentation measure, the collagen fiber bundling percentage, the collagen fiber rigidity, the collagen fiber anisotropy index, and the collagen fiber density index.

[0096] In yet other embodiments, the present disclosure relates to an apparatus, including a memory configured to store digitized pathology imaging data from a cancer patient, the digitized pathology imaging data including one or more digitized pathology images that have been segmented to identify one or more tumor-stroma regions; a feature extraction tool configured to access the one or more digitized pathology images and to extract a plurality of collagen fiber features from the one or more tumor-stroma regions, the plurality of collagen fiber features including one or more of a collagen fiber fragmentation measure, a collagen fiber bundling percentage, a collagen fiber rigidity, a collagen fiber anisotropy index, and a collagen fiber density index; and a machine learning model configured to use the plurality of collagen fiber features to generate a medical prognosis relating to the cancer patient. In some embodiments, the machine learning model is configured to generate a risk score using the plurality of collagen fiber features and to compare the risk score to a threshold to determine the medical prognosis. In some embodiments, the cancer patient has colorectal cancer. In some embodiments, the cancer patient has human papilloma virus (HPV) induced oropharyngeal squamous cell carcinoma. In some embodiments, the machine learning model comprises a Cox regression model.

[0097] It will be appreciated that the disclosed methods and / or block diagrams may be implemented as computer-executable instructions, in some embodiments. Thus, in one example, a computer-readable storage device (e.g.. a non-transitory computer-readable medium) may store computer executable instructions that if executed by a machine (e.g., computer, processor) cause the machine to perform the disclosed methods and / or block diagrams. While executable instructions associated with the disclosed methods and / or block diagrams are described as being stored on a computer-readable storage device, it is to be appreciated that executable instructions associated with other example disclosed methods and / or block diagrams described or claimed herein may also be stored on a computer-readable storage device.

[0098] Examples herein can include subject matter such as an apparatus, including a digital whole slide scanner, a CT system, an MRI system, a personalizedmedicine system, a CADx system, a processor, a system, circuitry, a method, means for performing acts, steps, or blocks of the method, at least one machine-readable medium including executable instructions that, when performed by a machine (e.g., a processor with memory, an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), or the like) cause the machine to perform acts of the method or of an apparatus or system, according to embodiments and examples described.

[0099] References to “one embodiment”, “an embodiment”, “one example”, and “an example” indicate that the embodiment(s) or example(s) so described may include a particular feature, structure, characteristic, property, element, or limitation, but that not every embodiment or example necessarily includes that particular feature, structure, characteristic, property, element or limitation. Furthermore, repeated use of the phrase “in one embodiment” does not necessarily refer to the same embodiment, though it may.[000100] “Computer-readable storage device”, as used herein, refers to a device that stores instructions or data. “Computer-readable storage device” does not refer to propagated signals. A computer-readable storage device may take forms, including, but not limited to, non-volatile media, and volatile media. Non-volatile media may include, for example, optical disks, magnetic disks, tapes, and other media. Volatile media may include, for example, semiconductor memories, dynamic memory, and other media. Common forms of a computer-readable storage device may include, but are not limited to, a floppy disk, a flexible disk, a hard disk, a magnetic tape, other magnetic medium, an application specific integrated circuit (ASIC), a compact disk (CD), other optical medium, a random access memory (RAM), a read only memory (ROM), a memory chip or card, a memory stick, and other media from which a computer, a processor or other electronic device can read. [000101] “Circuit”, as used herein, includes but is not limited to hardware, firmware, software in execution on a machine, or combinations of each to perform a function(s) or an action(s), or to cause a function or action from another logic, method, or system. A circuit may include a software controlled microprocessor, a discrete logic (e.g., ASIC), an analog circuit, a digital circuit, a programmed logic device, a memory device containing instructions, and other physical devices. A circuit may include one or more gates, combinations of gates, or other circuit components. Where multiplelogical circuits are described, it may be possible to incorporate the multiple logical circuits into one physical circuit. Similarly, where a single logical circuit is described, it may be possible to distribute that single logical circuit between multiple physical circuits.[000102] To the extent that the term “includes” or “including” is employed in the detailed description or the claims, it is intended to be inclusive in a manner similar to the term “comprising” as that term is interpreted when employed as a transitional word in a claim.[000103] Throughout this specification and the claims that follow, unless the context requires otherwise, the words 'comprise' and 'include' and variations such as 'comprising' and 'including' will be understood to be terms of inclusion and not exclusion. For example, when such terms are used to refer to a stated integer or group of integers, such terms do not imply the exclusion of any other integer or group of integers.[000104] To the extent that the term “or” is employed in the detailed description or claims e.g., A or B) it is intended to mean “A or B or both”. When the applicants intend to indicate “only A or B but not both” then the term “only A or B but not both” will be employed. Thus, use of the term “or” herein is the inclusive, and not the exclusive use. See, Bryan A. Garner, A Dictionary of Modern Legal Usage 624 (2d. Ed. 1995).[000105] While example systems, methods, and other embodiments have been illustrated by describing examples, and while the examples have been described in considerable detail, it is not the intention of the applicants to restrict or in any way limit the scope of the appended claims to such detail. It is, of course, not possible to describe every conceivable combination of components or methodologies for purposes of describing the systems, methods, and other embodiments described herein. Therefore, the invention is not limited to the specific details, the representative apparatus, and illustrative examples shown and described. Thus, this application is intended to embrace alterations, modifications, and variations that fall within the scope of the appended claims.

Claims

What is claimed is:1 . A method, comprising: accessing digitized pathology imaging data from a cancer patient, wherein the digitized pathology imaging data includes segmented imaging data that identifies one or more regions of interest (ROI) including tumor-stroma regions and / or collagen fibers; extracting a plurality of collagen fiber features from the one or more ROI; and providing the plurality of collagen fiber features to a machine learning model that is trained to generate a medical prognosis relating to the cancer patient.

2. The method of claim 1 , further comprising: accessing one or more digitized pathology images from the cancer patient; and performing a segmentation process on the one or more digitized pathology images to identify the one or more ROI.

3. The method of claim 1 , wherein the plurality of collagen fiber features comprise one or more of a collagen fiber fragmentation measure, a collagen fiber bundling percentage, a collagen fiber rigidity, a collagen fiber anisotropy index, and a collagen fiber density index.

4. The method of claim 1 , wherein the plurality of collagen fiber features comprise a collagen fiber fragmentation measure, a collagen fiber bundling percentage, a collagen fiber rigidity, a collagen fiber anisotropy index, and a collagen fiber density index.

5. The method of claim 1 , wherein the plurality of collagen fiber features consist of a collagen fiber fragmentation measure, a collagen fiber bundling percentage, a collagen fiber rigidity, a collagen fiber anisotropy index, and a collagen fiber density index.

6. The method of claim 1 , wherein the machine learning model is configured to generate a risk score using the plurality of collagen fiber features and to compare the risk score to a threshold to determine the medical prognosis.

7. The method of claim 1 , further comprising: operating the machine learning model on the plurality of collagen fiber features to generate a biological variable estimate; and generating the medical prognosis using the biological variable estimate.

8. The method of claim 7, wherein the biological variable estimate is a determination if the cancer patient has a Kirsten rat sarcoma virus (KRAS) genetic mutation.

9. A non-transitory computer-readable medium storing computerexecutable instructions that, when executed, cause a processor to perform operations, comprising: accessing digitized pathology imaging data from a cancer patient, wherein the digitized pathology imaging data includes one or more digitized pathology images that have been segmented to identify one or more regions of interest (ROI) including one or more tumor-stroma regions and collagen fibers; extracting a plurality of collagen fiber features from the one or more ROI, wherein the plurality of collagen fiber features comprise one or more of a collagen fiber fragmentation measure, a collagen fiber bundling percentage, a collagen fiber rigidity, a collagen fiber anisotropy index, and a collagen fiber density index; and operating a machine learning model to use the plurality of collagen fiber features to generate a medical prognosis relating to the cancer patient.

10. The non-transitory computer-readable medium of claim 9, wherein the medical prognosis relates to an overall survival of the cancer patient or a treatment response.11 . The non-transitory computer-readable medium of claim 10, wherein the operations further comprise: accessing the one or more digitized pathology images from the cancer patient; performing a first segmentation process on the one or more digitized pathology images to identify the one or more tumor-stroma regions; and performing a second segmentation process on the one or more digitized pathology images to identify the collagen fibers within the one or more tumor-stroma regions.

12. The non-transitory computer-readable medium of claim 9, wherein the tumor-stroma regions are identified using morphological operations and the collagen fibers are identified using a Derivative of a Gaussian model.

13. The non-transitory computer-readable medium of claim 9, wherein the cancer patient has colon cancer.

14. The non-transitory computer-readable medium of claim 9, wherein the cancer patient has human papilloma virus (HPV) induced oropharyngeal squamous cell carcinoma.

15. The non-transitory computer-readable medium of claim 9, wherein the plurality of collagen fiber features comprise the collagen fiber fragmentation measure, the collagen fiber bundling percentage, the collagen fiber rigidity, the collagen fiber anisotropy index, and the collagen fiber density index.

16. An apparatus, comprising: a memory configured to store digitized pathology imaging data from a cancer patient, wherein the digitized pathology imaging data includes one or more digitized pathology images that have been segmented to identify one or more tumor-stroma regions; a feature extraction tool configured to access the one or more digitized pathology images and to extract a plurality of collagen fiber features from the one ormore tumor-stroma regions, wherein the plurality of collagen fiber features comprise one or more of a collagen fiber fragmentation measure, a collagen fiber bundling percentage, a collagen fiber rigidity, a collagen fiber anisotropy index, and a collagen fiber density index; and a machine learning model configured to use the plurality of collagen fiber features to generate a medical prognosis relating to the cancer patient.

17. The apparatus of claim 16, wherein the machine learning model is configured to generate a risk score using the plurality of collagen fiber features and to compare the risk score to a threshold to determine the medical prognosis.

18. The apparatus of claim 16, wherein the cancer patient has colorectal cancer.

19. The apparatus of claim 16, wherein the cancer patient has human papilloma virus (HPV) induced oropharyngeal squamous cell carcinoma.

20. The apparatus of claim 16, wherein the machine learning model comprises a Cox regression model.

Images