System and methods for cancer prediction and detection

Abstract

Systems and methods for detecting cancer in a subject comprising receiving an image of a first region of tissue from a subject, calculating a roughness exponent for the first region of tissue, comparing the roughness exponent of the first region of tissue to a reference value of 0.5, wherein a difference of less than 0.2 between the roughness exponent and the reference value indicates the first region of tissue is active tissue, determining a range of multiscale anisotropic factors of the active tissue, wherein a subset of the multiscale anisotropic factors greater than one standard deviation above a benchmark average indicates that the active tissue is cancerous or pre-cancerous, and if the active tissue is determined to be cancerous, applying anticancer therapy.

Description

TITLE SYSTEM AND METHODS FOR CANCER PREDICTION AND DETECTION RELATED APPLICATIONS

[0001] This application claims priority to United States Provisional Application No.63 / 693,054 filed under 35 U.S.C. § 111(b) on September 10, 2024. The entire disclosure is expressly incorporated herein by reference for all purposes. STATEMENT REGARDING FEDERALLY SPONSORED RESEARCH

[0002] This invention was made without government support. The government has no rights in this invention. BACKGROUND

[0003] Cancer, a group of diseases characterized by the uncontrolled growth and spread of abnormal cells, remains one of the leading causes of death worldwide. The growth and behavior of cancer cells differs from normal cells. Despite advancements in medical research and treatment, the complex nature of cancer poses significant challenges in diagnosis, management, and treatment. Early detection may be crucial for correctly addressing the issue, which may improve patient outcomes. This benefit has led to the development of various screening methods and diagnostic tools for early detection.

[0004] Known methods of detecting cancer are disclosed in US Patent Nos.10,769,790 and 10,467,755, and in European Patent No. EP 2,988,659, the disclosures of which are incorporated herein by reference in their entirety.

[0005] 1 in 8 women are diagnosed with invasive breast cancer during their lifetime. When breast cancer is caught in early stages treatment is over 99% successful and earlier detection methods are currently a major area of focus to take advantage of our collective advancements in treatment strategies. Risk models and genetic testing can help identify high-risk individuals for triaging to secondary imaging and shorter- screening intervals for earlier detection, but 50% of women who develop invasive breast cancer have no risk-factors from our current traditional approaches. The traditional risk models include the Tyrer-Cuzick, Gail, and Breast Cancer Surveillance Consortium (BCSC) models which incorporate factors such as age, age at first live birth, age at menopause, personal history of ovarian cancer, family history of breast and ovarian cancer, number of biopsies, and other basic information from personal and family history.

[0006] As identified by The U.S. Preventive Services Task Force (USPSTF) mammography is the only imaging modality shown to reduce breast-cancer mortality, yet it is less effective for the ~40% of screened women with dense tissue. Up to 50% of cancers are missed in this group, and mortality risk isnearly doubled. Furthermore, the Cancer Moonshot Program has identified improved methods for risk stratification as an opportunity to reduce breast cancer incidence rates. Current mammography screening guidelines have improved detection of pre-invasive DCIS, but invasive breast cancer detection has remained stable. It is currently unknown how to predict when DCIS will progress into invasive cancer causing some patients to begin treatment who may never progress to invasive cancer and other patients who are unknowingly at high-risk for invasive breast cancer progression not receiving enough treatment.

[0007] Therefore, there is a pressing need to find novel technologies to improve cancer prediction and detection. SUMMARY

[0008] Provided herein is a method for detecting cancer in a subject comprising receiving an image of a first region of tissue from a subject, calculating a roughness exponent for the first region of tissue, comparing the roughness exponent of the first region of tissue to a reference value of 0.5, wherein a difference of less than 0.2 between the roughness exponent and the reference value indicates the first region of tissue is active tissue, determining a range of multiscale anisotropic factors of the active tissue, wherein a subset of the multiscale anisotropic factors greater than one standard deviation above a benchmark average indicates that the active tissue is cancerous or pre-cancerous, and if the active tissue is determined to be cancerous, applying anticancer therapy.

[0009] In certain examples, the roughness exponent is calculated using one or more multiscale analytical methods selected from a wavelet-transform modulus maxima, a wavelet leader, detrended fluctuation, and Fourier analysis.

[0010] In certain examples, the steps of receiving, calculating, comparing, and determining are each performed a plurality of time, each on a different region of tissue from the subject.

[0011] In certain examples, the first region of tissue is or comprises tissue selected from a breast region, a brain region, a colon region, a dermal region, an esophagus region, a kidney region, a liver region, a lung region, an ovary region, a pancreatic region, a prostate region, a stomach region, and a uterine region.

[0012] In certain examples, the difference between the roughness exponent and the reference value is less than or equal to 0.15.

[0013] In certain examples, the difference between the roughness exponent and the reference value is less than or equal to 0.1.

[0014] In certain examples, the difference between the roughness exponent and the reference value is less than or equal to 0.05.

[0015] Provided herein also is a system for detecting cancer in a subject comprising a memory medium having processor-executable instructions stored thereon, wherein the memory is tangible, non-transitory, and computer readable, a processor configured to execute processor-executable instructions, wherein the processor-executable instructions, when executed, are configured to receive an image of a first region of tissue of a subject, calculate a roughness exponent for the first region of tissue, compare the roughness exponent of the first region of tissue to a reference value of 0.5, wherein a difference of less than 0.2 between the roughness exponent and the reference value indicates the first region of tissue is active tissue, determine a range of multiscale anisotropic factors of the active tissue, generate a detection output based on the multiscale anisotropic factors of the active tissue, wherein a subset of the multiscale anisotropic factors greater than one standard deviation above a benchmark average indicates that the active tissue is cancerous or pre-cancerous, and a display configured to render the detection output thereon.

[0016] In certain examples, the roughness exponent is calculated using one or more multiscale analytical methods selected from a wavelet-transform modulus maxima, a wavelet leader, detrended fluctuation, and Fourier analysis.

[0017] In certain examples, the steps of receiving, calculating, comparing, and determining are each performed a plurality of time, each on a different region of tissue from the subject.

[0018] In certain examples, the first region of tissue is or comprises tissue selected from a breast region, a brain region, a colon region, a dermal region, an esophagus region, a kidney region, a liver region, a lung region, an ovary region, a pancreatic region, a prostate region, a stomach region, and a uterine region.

[0019] In certain examples, the difference between the roughness exponent and the reference value is less than or equal to 0.15.

[0020] In certain examples, the difference between the roughness exponent and the reference value is less than or equal to 0.1.

[0021] In certain examples, the difference between the roughness exponent and the reference value is less than or equal to 0.05.

[0022] Also provided herein is A method for predicting cancer in a subject, comprising: receiving a mammographic view, wherein the mammographic view is a left mediolateral oblique mammogram view, a left craniocaudal mammogram view, a right mediolateral oblique mammogram view, or a right craniocaudal mammogram view, wherein the mammographic view includes a plurality of subregions; calculating a roughness exponent for each of the subregions; comparing the roughness exponent of each of the subregions to a reference value of 0.5, wherein a difference of less than 0.2 between the roughness exponent of one of the subregions and the reference value indicates the one of the subregions is active tissue; calculating an anisotropy factor for each of the subregions; calculating an anisotropic mass for the mammographic view; normalizing the anisotropic mass for the mammographic view by a mammographic area; determining a mean of the anisotropic mass of the mammogram view; and comparing the active anisotropic mass with a threshold active anisotropy mass value derived from a library of mammograms todetermine a likelihood of cancer, wherein the library of mammograms includes mammograms from patients who developed cancer.

[0023] In certain examples, the roughness exponent is calculated using one or more multiscale analytical methods selected from a wavelet-transform modulus maxima, a wavelet leader, detrended fluctuation, and Fourier analysis.

[0024] In certain examples, the difference between the roughness exponent and the reference value is less than or equal to 0.15.

[0025] In certain examples, the difference between the roughness exponent and the reference value is less than or equal to 0.1.

[0026] In certain examples, the difference between the roughness exponent and the reference value is less than or equal to 0.05.

[0027] In certain examples, the library of mammograms includes mammograms from patients who subsequently developed cancer about three years after the mammograms were obtained.

[0028] In certain examples, the anisotropy factors and the anisotropy mass can be calculated at multiple size cales.

[0029] Also provided herein is a method for predicting cancer in a subject comprising receiving at least four mammographic views, including a left mediolateral oblique mammogram view, a left craniocaudal mammogram view, a right mediolateral oblique mammogram view, and a right craniocaudal mammogram view, wherein each of the four mammographic views includes a plurality of subregions, calculating a roughness exponent for each of the subregions, comparing the roughness exponent of each of the subregions to a reference value of 0.5, wherein a difference of less than 0.2 between the roughness exponent of one of the subregions and the reference value indicates the one of the subregions is active tissue, calculating an anisotropy factor for each of the subregions, calculating an anisotropic mass for each of the four mammographic views, normalizing the anisotropic mass for each of the four mammographic views by a mammographic area, determining a mean of the anisotropic mass of the left mediolateral oblique mammogram view and the left craniocaudal mammogram view, and a mean of the anisotropic mass of the right mediolateral oblique mammogram view and the right craniocaudal mammogram view, summing the mean of the anisotropic mass of the left mediolateral oblique mammogram view and the left craniocaudal mammogram view with the mean of the anisotropic mass of the right mediolateral oblique mammogram view and the right craniocaudal mammogram view to determine the active anisotropic mass for both breasts of the subject, and comparing the active anisotropic mass for both breasts of the subject with a threshold active anisotropy mass value derived from a library of mammograms to determine a likelihood of cancer, wherein the library of mammograms includes mammograms from patients who developed cancer.BRIEF DESCRIPTION OF THE DRAWINGS

[0030] The patent or application file contains at least one drawing executed in color. Copies of this patent or patent application publication with color drawing(s) will be provided by the Office upon request and payment of the necessary fee.

[0031] FIG.1: American College of Radiology (ACR) Breast Imaging and Reporting Data System (BI-RADS) density scores, where radiologists assign a density score to each mammogram exam to quantify the amount of dense breast tissue.

[0032] FIG.2: Overview of the 2D wavelet-transform modulus maxima (WTMM) multifractal sliding window approach.

[0033] FIG.3: illustrates the discriminatory power of different dense tissue types.

[0034] FIG.4: illustrates a sample mammogram subregion undergoing anisotropic analysis.

[0035] FIG.5: illustrates a system for detecting cancer in a subject.

[0036] FIG.6: Boxplot (top) and probability density function (bottom) of active anisotropic mass from patients’ mammograms 3 years prior to their cancer diagnosis compared to a control group at 0.5 mm scale.

[0037] FIG.7: Boxplot (top) and probability density function (bottom) of active anisotropic mass from patients’ mammograms 3 years prior to their cancer diagnosis compared to a control group at 1.4 mm scale.

[0038] FIG.8: Boxplot (top) and probability density function (bottom) of active anisotropic mass from patients’ mammograms 3 years prior to their cancer diagnosis compared to a control group at 3.9 mm scale.

[0039] FIG.9: Graph of cancer frequency and log2(Wavelet Scale in Millimeters).

[0040] FIG.10: Graph of anisotropy mass threshold for high vs low risk and log2(Wavelet Scale in Millimeters).

[0041] FIG.11A: Boxplot (top) and probability density function (bottom) of active anisotropic mass from patients’ mammograms 3-years prior to their cancer diagnosis compared to a control group at 0.5 mm scale for the RMLO mammographic view.

[0042] FIG.11B: Boxplot (top) and probability density function (bottom) of active anisotropic mass from patients’ mammograms 3-years prior to their cancer diagnosis compared to a control group at 0.5 mm scale for the LMLO mammographic view.

[0043] FIG.11C: Boxplot (top) and probability density function (bottom) of active anisotropic mass from patients’ mammograms 3-years prior to their cancer diagnosis compared to a control group at 0.5 mm scale for the RCC mammographic view.

[0044] FIG.11D: Boxplot (top) and probability density function (bottom) of active anisotropic massfrom patients’ mammograms 3-years prior to their cancer diagnosis compared to a control group at 0.5 mm scale for the LCC mammographic view.

[0045] FIG.12A: Boxplot (top) and probability density function (bottom) of active anisotropic mass from patients’ mammograms 3-years prior to their cancer diagnosis compared to a control group at 1.4 mm scale for the RMLO mammographic view.

[0046] FIG.12B: Boxplot (top) and probability density function (bottom) of active anisotropic mass from patients’ mammograms 3-years prior to their cancer diagnosis compared to a control group at 1.4 mm scale for the LMLO mammographic view.

[0047] FIG.12C: Boxplot (top) and probability density function (bottom) of active anisotropic mass from patients’ mammograms 3-years prior to their cancer diagnosis compared to a control group at 1.4 mm scale for the RCC mammographic view.

[0048] FIG.12D: Boxplot (top) and probability density function (bottom) of active anisotropic mass from patients’ mammograms 3-years prior to their cancer diagnosis compared to a control group at 1.4 mm scale for the LCC mammographic view.

[0049] FIG.13A: Boxplot (top) and probability density function (bottom) of active anisotropic mass from patients’ mammograms 3-years prior to their cancer diagnosis compared to a control group at 3.9 mm scale for the RMLO mammographic view.

[0050] FIG.13B: Boxplot (top) and probability density function (bottom) of active anisotropic mass from patients’ mammograms 3-years prior to their cancer diagnosis compared to a control group at 3.9 mm scale for the LMLO mammographic view.

[0051] FIG.13C: Boxplot (top) and probability density function (bottom) of active anisotropic mass from patients’ mammograms 3-years prior to their cancer diagnosis compared to a control group at 3.9 mm scale for the RCC mammographic view.

[0052] FIG.13D: Boxplot (top) and probability density function (bottom) of active anisotropic mass from patients’ mammograms 3-years prior to their cancer diagnosis compared to a control group at 3.9 mm scale for the LCC mammographic view.

[0053] FIG.14A: Top: Graph of cancer frequency and log2(Wavelet Scale in Millimeters). Bottom: Graph of anisotropy mass threshold for high vs low risk and log2(Wavelet Scale in Millimeters) for the RMLO mammographic view.

[0054] FIG.14B: Top: Graph of cancer frequency and log2(Wavelet Scale in Millimeters). Bottom: Graph of anisotropy mass threshold for high vs low risk and log2(Wavelet Scale in Millimeters) for the LMLO mammographic view.

[0055] FIG.14C: Top: Graph of cancer frequency and log2(Wavelet Scale in Millimeters). Bottom: Graph of anisotropy mass threshold for high vs low risk and log2(Wavelet Scale in Millimeters) for the RCCmammographic view.

[0056] FIG.14D: Top: Graph of cancer frequency and log2(Wavelet Scale in Millimeters). Bottom: Graph of anisotropy mass threshold for high vs low risk and log2(Wavelet Scale in Millimeters) for the LCC mammographic view. DETAILED DESCRIPTION

[0057] Throughout this disclosure, various publications, patents, and published patent specifications are referenced by an identifying citation. The disclosures of these publications, patents, and published patent specifications are hereby incorporated by reference into the present disclosure in their entirety to more fully describe the state of the art to which this invention pertains.

[0058] As used herein, term “anisotropy” refers to the property of being directionally dependent, characterized by differences in a material’s structural properties when measured along different directions. In general, those skilled in the art will appreciate that the term “anisotropy,” as understood from context, refers to variations in pixel intensity fluctuations observed in an image when measured in different directions.

[0059] As used herein, the term “anisotropic” refers to the property or characteristic of an image where pixel intensity fluctuations vary when measured along different directions.

[0060] As used herein, the term “cancer” refers to a group of diseases characterized by unregulated cell growth. Examples of cancers include, but are not limited to: Acute lymphoblastic leukemia, Acute myeloid leukemia, Adrenocortical carcinoma, AIDS-related cancers, AIDS-related lymphoma, Anal cancer, Appendix cancer, Astrocytoma (childhood cerebellar or cerebral), Basal cell carcinoma, Extrahepatic bile duct cancer, Bladder cancer, Bone cancer (Osteosarcoma / Malignant fibrous histiocytoma), Brainstem glioma, Brain tumors (cerebellar astrocytoma, cerebral astrocytoma / malignant glioma, ependymoma, medulloblastoma, supratentorial primitive neuroectodermal tumors, visual pathway and hypothalamic glioma), Breast cancer, Bronchial adenomas / carcinoids, Burkitt lymphoma, Childhood carcinoid tumor (gastrointestinal), Carcinoma of unknown primary, Primary central nervous system lymphoma, Childhood cerebellar astrocytoma, Childhood cerebral astrocytoma / malignant glioma, Cervical cancer, Childhood cancers, Chronic lymphocytic leukemia, Chronic myelogenous leukemia, Chronic myeloproliferative disorders, Colon cancer, Cutaneous T-cell lymphoma, Desmoplastic small round cell tumor, Endometrial cancer, Ependymoma, Esophageal cancer, Ewing's sarcoma, Childhood extracranial germ cell tumor, Extragonadal germ cell tumor, Extrahepatic bile duct cancer, Intraocular melanoma, Retinoblastoma, Gallbladder cancer, Gastric (stomach) cancer, Gastrointestinal carcinoid tumor, Gastrointestinal stromal tumor (GIST), Germ cell tumors (extracranial, extragonadal, or ovarian), Gestational trophoblastic tumor, Brainstem glioma, Childhood cerebral astrocytoma, Childhood visual pathway and hypothalamic glioma,Gastric carcinoid, Hairy cell leukemia, Head and neck cancer, Heart cancer, Hepatocellular (liver) cancer, Hodgkin lymphoma, Hypopharyngeal cancer, Childhood hypothalamic and visual pathway glioma, Intraocular melanoma, Islet cell carcinoma (endocrine pancreas), Kaposi sarcoma, Kidney cancer (renal cell carcinoma), Laryngeal cancer, Leukemias (acute lymphoblastic / acute lymphocytic, acute myeloid / acute myelogenous, chronic lymphocytic, chronic myelogenous, hairy cell), Lip and oral cavity cancer, Liposarcoma, Primary liver cancer, Non-small cell lung cancer, Small cell lung cancer, Lymphomas (Burkitt, cutaneous T-cell, Hodgkin, primary central nervous system), Waldenström macroglobulinemia, Malignant fibrous histiocytoma of bone / osteosarcoma, Childhood medulloblastoma, Melanoma, Intraocular melanoma (eye), Merkel cell carcinoma, Adult malignant mesothelioma, Childhood mesothelioma, Metastatic squamous neck cancer with occult primary, Mouth cancer, Childhood multiple endocrine neoplasia syndrome, Multiple myeloma / plasma cell neoplasm, Mycosis fungoides, Myelodysplastic syndromes, Myelodysplastic / myeloproliferative diseases, Chronic myelogenous leukemia, Adult acute myeloid leukemia, Childhood acute myeloid leukemia, Multiple myeloma (cancer of the bone marrow), Chronic myeloproliferative disorders, Nasal cavity and paranasal sinus cancer, Nasopharyngeal carcinoma, Neuroblastoma, Non-Hodgkin lymphoma, Non-small cell lung cancer, Oral cancer, Oropharyngeal cancer, Osteosarcoma / malignant fibrous histiocytoma of bone, Ovarian cancer, Ovarian epithelial cancer (surface epithelial-stromal tumor), Ovarian germ cell tumor, Ovarian low malignant potential tumor, Pancreatic cancer, Pancreatic islet cell cancer, Paranasal sinus and nasal cavity cancer, Parathyroid cancer, Penile cancer, Pharyngeal cancer, Pheochromocytoma, Pineal astrocytoma, Pineal germinoma, Childhood pineoblastoma and supratentorial primitive neuroectodermal tumors, Pituitary adenoma, Plasma cell neoplasia / multiple myeloma, Pleuropulmonary blastoma, Primary central nervous system lymphoma, Prostate cancer, Rectal cancer, Renal cell carcinoma (kidney cancer), Transitional cell cancer of the renal pelvis and ureter, Retinoblastoma, Childhood rhabdomyosarcoma, Salivary gland cancer, Ewing family of tumors (sarcoma), Kaposi sarcoma, Soft tissue sarcoma, Uterine sarcoma, Sézary syndrome, Nonmelanoma skin cancer, Merkel cell skin carcinoma, Small intestine cancer, Soft tissue sarcoma, Squamous cell carcinoma, Metastatic squamous neck cancer with occult primary, Stomach cancer, Childhood supratentorial primitive neuroectodermal tumor, Cutaneous T-cell lymphoma, Testicular cancer, Throat cancer, Childhood thymoma, Thymoma and thymic carcinoma, Thyroid cancer, Childhood thyroid cancer, Transitional cell cancer of the renal pelvis and ureter, Gestational trophoblastic tumor, Adult carcinoma of unknown primary site, Childhood cancer of unknown primary site, Transitional cell cancer of the ureter and renal pelvis, Urethral cancer, Endometrial uterine cancer, Uterine sarcoma, Vaginal cancer, Childhood visual pathway and hypothalamic glioma, Vulvar cancer, and Childhood Wilms tumor (kidney cancer).

[0061] As used herein, the term “pre-cancer” or “pre-cancerous” refers to tissue in or from a subject that is not yet cancerous but has a higher likelihood of becoming cancerous compared to normal tissue.

[0062] As will be understood from context, “risk” of a disease, disorder, and / or condition comprises a likelihood that a particular individual will develop a disease, disorder, and / or condition (e.g., a cancer). In some embodiments, risk is expressed as a percentage. In some embodiments, risk is from 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 20, 30, 40, 50, 60, 70, 80, 90, up to 100%. In some embodiments risk is expressed as a risk relative to a risk associated with a reference sample or group of reference samples. In some embodiments, a reference sample or group of reference samples have a known risk of a disease, disorder, condition, and / or event (e.g., a muscular dystrophy). In some embodiments a reference sample or group of reference samples are from individuals comparable to a particular individual. In some embodiments, relative risk is 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, or more.

[0063] As used herein, the term “subject” refers to a human or any non-human animal (e.g., mouse, rat, rabbit, dog, cat, cattle, swine, sheep, horse, or primate). A human includes pre-natal and post-natal forms. In many embodiments, a subject is a human being. A subject can be a patient, which refers to a human presenting to a medical provider for diagnosis or treatment of a disease. The term “subject” is used herein interchangeably with “individual” or “patient.” A subject can be afflicted with or is susceptible to a disease or disorder but may or may not display symptoms of the disease or disorder.

[0064] Although the detailed description provides an in-depth exploration of the disclosure’s application in detecting breast cancer, it will be understood that the methods described herein are applicable to a wide variety of cancers. These methods use advanced imaging techniques, computational algorithms, and multifractal analysis to identify cancerous tissues in various organs. Therefore, although breast cancer serves as the primary illustrative example, the disclosure's utility is not confined to it. As such, the disclosure may be adapted for use in detecting other types of cancer, thereby broadening the scope and potential impact of this approach in the field of oncology.

[0065] FIG.1 illustrates the ACR BI-RADS density scores, where radiologists assign a density score to each mammogram exam to quantify the amount of dense breast tissue. High mammographic breast density is a recognized independent risk factor for breast cancer. Women with dense breasts face an elevated risk of developing breast cancer compared to those with non-dense breasts. Radiologists typically quantify breast density using the Breast Imaging Reporting and Data System (BI-RADS) (FIG.1). Alteration of the stromal architecture and composition of the extracellular matrix is a well-recognized component of breast pathologies. Signs of aberrant stroma and epithelia may exist long before there is overt carcinoma and in fact, alterations might not always be reactive, but might sometimes play an initial role in breast carcinogenesis. Histologically normal breast tissues adjacent to breast tumors frequently exhibit methylation changes in multiple genes. Highly methylated tumor suppressor genes have been found in peritumoral breast tissue cells as far as 4 cm away from tumors. Therefore, the properties that distinguish between healthy and tumorous tissues may not be limited to the tumor characteristics but rather extend tosurrounding non-tumorous tissues. Characterizing the microenvironment surrounding the tumor will support the early detection of breast cancer and the development of novel therapies. This issue was addressed by expanding whole-breast tissue analysis via mammography, notably via the analysis of anisotropic active dense mammographic tissue.

[0066] Significantly, algorithms may be developed to provide longitudinal analysis on several years of screening mammography from a large dataset of patients in four cohorts, including: three groups of patients who eventually obtained a pathology-proven diagnostic of 1) a cancerous lesion during a screening visit, 2) those whose cancer was discovered outside of a screening visit (interval cancer), and 3) those with a benign lesion, respectively, and 4) a group of control (healthy) patients. This analysis will preferably include a rigorous statistical study of the variability of healthy mammographic tissue.

[0067] Previous studies have shown that the amounts and temporal change of mammographic dense tissue and dense tissue subtypes could provide insights into cancer risk. Such previous studies have been used to computationally quantify mammographic breast tissue composition and identify changes over time that indicate the potential for malignancy using routine screening mammograms. Mammogram exams were analyzed from patients with pathology-proven cancer within one year of their last screening and from patients with no history of breast disease. This reference suggests that the extensive restructuring associated with early tumor onset is detectable via computational mammography prior to radiological diagnosis of breast cancer.

[0068] Although the embodiments presented herein describe the use of mammographic images, other radiological and non-radiological images may be used. A radiological image may refer to an image obtained with the use of ionizing radiation typically associated with radiographic techniques like X-rays, CT scans, or mammography. A non-radiological image may refer to an image obtained without the use of ionizing radiation, which may be acquired using various methods, including but not limited to ultrasound, magnetic resonance imaging (MRI), optical imaging, and thermal imaging. For example, ultrasound uses high-frequency sound waves to create images of internal body structures, while MRI uses magnets and radio waves to generate images of organs and tissues. Optical imaging techniques, such as optical coherence tomography (OCT), use light to capture images of tissues. Non-radiological images can be obtained through at least these methods to analyze tissue characteristics without exposing the patient to radiation.

[0069] FIG.2 illustrates an overview of the 2D WTMM multifractal sliding window approach, showing the segmentation process of breast tissue and the analysis of mammographic sub-images for different tissue types through wavelet transforms and WTMM construction.

[0070] Although wavelet transforms are discussed, other transforms and / or image processing techniques may be used in the segmentation process. In addition to the wavelet transform, other types oftransforms and / or image processing techniques may also be used to enhance the accuracy of cancer detection methods, including, for example, the wavelet leader analysis, which involves generating a roughness exponent using various multi-scale analytical methods. In some embodiments, the roughness exponent is calculated using one or more multi-scale methods selected from at least one of a wavelet- transform modulus maxima, a wavelet leader, detrended fluctuation, and Fourier analysis. Another transform is a detrended fluctuation analysis (DFA), in which a roughness exponent may be calculated using any mathematically related and / or similar multi-scale density fluctuation assessment method yielding an exponent or spectrum of exponents that is similar and / or can complement the use of the Hurst exponent. The Fourier analysis is an additional approach to generating a roughness exponent through multi-scale analytical methods.

[0071] Performing a transform on a radiological or a non-radiological image involves converting the image data using mathematical techniques to highlight specific features and patterns within the image. For example, a wavelet transform decomposes the image into components at multiple scales, allowing for the detailed examination of texture and structure. Performance of a transform on a radiological or non- radiological image may allow identification of active dense tissue within the transformed radiological or non-radiological image.

[0072] A 360x360 pixel sliding window was positioned at the top left of the segmented breast tissue. The sliding window shifted from left to right and top to bottom with a step size of 32 pixels between subregions. If the subregion was entirely contained inside the mask, the subregion was accepted for further analysis (FIG.2D-H). Each sub-image was wavelet transformed across 50 different size scales. Maxima chains, which form naturally by taking the maximal values of the wavelet-transform modulus, i.e., the WTMM, and represent locations where the image gradient is locally maximal. The WTMM maxima, or WTMMM form maxima lines through scales, from which partition functions can be calculated and a Hurst exponent (H) can be obtained following the methods described in (FIG.2I-K). This allows us to determine mammographic tissue subtypes: fatty tissue (H<0.45), passive dense tissue (H>0.55), or active dense tissue (0.45≤H≤0.55) (FIG.2L, M). The amount of passive and active dense tissue was predicted by creating linear mixed-effects (LME) models. The fixed effects for this model included time (to diagnosis for cancer cases and to the last visit for controls), cancer status, and breast status. An interaction term between time and breast status was fitted to test the hypothesis of changes occurring in dense breast tissue in tumorous breasts vs non-tumorous breasts. Random effects included breast (left or right) nested within participant nested within case control strata obtained from age-matching. The MMC mammogram data and accompanying pathology reports were collected from 50 patients (27 controls and 23 malignant cases), with mammograms obtained using an Hologic® mammography system.

[0073] Critically important observations can be missed when using a computational image analysistool that only considers a single scale (or a single frequency, if using a Fourier-based method). It would be impossible to observe the drastically different behaviors at different size scales using a single-scale tool. For example, if a single-scale analysis had been performed on the melanoma data at scale of ~15µm (see FIG.7 from "Multiscale Anisotropy Analysis of Second-Harmonic Generation Collagen Imaging of Mouse Skin" by Karissa Tilbury, Xiang Hua Han, Peter C. Brooks, and Andre Khalil, the disclosure of which is incorporated herein by reference in its entirety), then it would have been impossible to discriminate between the two subtypes of diseased skin collagen fibers (Tilbury et al.2021). This would have forced the analyst to report inconclusive results. But perhaps even more importantly, suppose two different analysts were studying these data at two different individual scales, say analyst 1 at one given scale less than 10 microns, and analyst 2 at one given scale greater than 20 microns. Then the two analysts would have reached contradicting conclusions.

[0074] FIG.3 illustrates the discriminatory power of different dense tissue types, focusing on overall dense tissue, active dense tissue, and anisotropic active dense tissue. To explore the potential applicability of the multiscale anisotropy on non-radiological or radiological images, such as mammograms, a previously studied dataset was revisited. The dataset consists of mammograms from patients with either a malignancy (n=72) or a benign lesion (n=18) at their last screening mammogram prior to pathology-confirmed diagnosis. The overall amount of (active and passive) dense tissue (FIG.3A, purple) was a significant marker to discriminate between cancer vs. benign patients (p=0.04, Wilcoxon Ranksum Test), and that restricting the analysis to only the active dense tissue (FIG.3B, red) increases the significance level (p=0.01). Given the presence of active dense tissue in either benign or non-tumorous breasts, a complementary textural analysis was sought to restrict the areas of active dense tissue. To do so, an anisotropic analysis was performed on the benign dataset and empirically determined benchmark average and standard deviation of the anisotropy factors at each wavelet scale. The subset of active dense tissue subregions that had anisotropy factors was kept greater than one standard deviation above the benchmark average. Significantly, when performing the same statistical analysis on the amount of active tissue that are anisotropic (FIG.3C, orange), the level of significance vastly improves (p=0.00001). Note that the analysis shown in FIG.3C was done using a single wavelet scale (~15 mm), but the level of significance was observed on a wide range of scales.

[0075] In one embodiment, a user may use a database of longitudinal 2D mammograms from 199 cancer patients and 3717 control patients from a dataset, to verify that multiscale anisotropy augments the cancer predictability of active mammographic density.

[0076] This embodiment may include calculating the multiscale anisotropic signatures of mammographic subregions of active dense breast tissue no more than one year before diagnosis, as well as on subsets of patients with past longitudinal mammograms at 3 and 6 years prior to diagnosis. Cancer,benign, and interval cancer patients are defined as women having received a pathology-proven diagnosis, while control patients have all negative mammograms.

[0077] Patients may be age-matched based on their age at their final mammography visit. The four standard bilateral mammographic views (left MLO, right MLO, left CC, and right CC) may be analyzed in this study. While all 9,300 patients have at least one set of four bilateral views from a screening visit, most have data available for prior visits. For example, approximately 90% of the controls have at least one prior screening visit. For the cancer, benign, and interval cancer patients, the percentages that have at least one prior screening visit are 46%, 77%, and 81%, respectively. The final size of the subset for the past longitudinal study at 3- and 6-year priors is determined after the age-matching analysis is performed. Note that the prior periods of 3 and 6 years may be chosen because of data availability. In the UK, mammography screening is recommended every 3 years, unlike in the US, where it is typically either every year or every other year.

[0078] FIG.4 outlines a 2D WTMM anisotropy calculation strategy. The 2D WTMM calculations for both the multifractal and anisotropy methods may be performed using a software suite that uses a combination of script-wrapped C routines with bash secure shell transfers. More particularly, FIG.4 illustrates a detailed analysis of a mammogram subregion through the application of wavelet transforms. In part (A), a sample mammogram subregion is shown. Part (B) presents the convolution of the Gaussian wavelet with this mammogram subregion at three representative wavelet size scales. In part (C), maxima chains are identified, representing the local maximal gradients, known as the wavelet-transform modulus maxima (WTMM). These gradients show variations in image density. Part (D) presents zoomed-in views of areas outlined in black from part (C), focusing on the WTMM vectors, which indicate both the magnitude and direction of local gradients corresponding to maximum image density fluctuations. Part (E) presents probability density functions (PDF) for the three representative wavelet scales, generated from the angles of the local gradient vectors with a theoretical isotropic 1 / 2π line shown in black. The anisotropy factor is obtained by integrating the difference between each PDF and this theoretical line.

[0079] Three statistical models may be created to predict disease status using 1) amounts of overall dense tissue, 2) amounts of active dense tissue, and 3) amounts of anisotropic active dense tissue. Area under the receiver operating characteristic curve (AUC) and delta AUC are used to evaluate each model’s performance. The amounts of 1) overall dense tissue (passive + active dense), 2) active dense tissue, and 3) anisotropic active dense tissue may be calculated for each patient category (control, cancer, benign, and interval cancer) at all time points available. Validation through a held-out dataset is used by splitting the dataset into training and testing sets. The training set includes 2 / 3 of the age-matched subclasses (randomly chosen), and the testing set consists of the remaining 1 / 3 age-matched subclasses. Conditional logistic regression models with strata of age-matched cancer, benign, interval cancer, and control patients may bebuilt using the training set. The explanatory variable for Model 1 is the amount of overall dense tissue, for Model 2 it is the amount of active dense tissue, and for Model 3 it is the amount of anisotropic active dense tissue. The response variable is the patient outcome of cancer or interval cancer compared with no cancer or benign, for all three models. Models 1, 2, and 3 are used to predict the observation in the testing set. An ROC curve and the area under the curve (AUC) are created and calculated for each model. Delta AUC is used to compare the discriminatory performance of the models. A quantitative measure of success is the demonstration that anisotropic active dense tissue (model 3) can predict breast cancer with a 10% higher sensitivity than either of the other two models, without increasing the false-positive rate.

[0080] A method for detecting cancer in a subject can comprise receiving an image of a first region of tissue from a subject; calculating a roughness exponent for the first region of tissue; comparing the roughness exponent of the first region of tissue to a reference value of 0.5, wherein a difference of less than 0.2 between the roughness exponent and the reference value indicates the first region of tissue is active tissue; and determining a range of multiscale anisotropic factors of the active tissue, wherein a subset of the multiscale anisotropic factors greater than one standard deviation above a benchmark average indicates that the active tissue is cancerous or pre-cancerous.

[0081] If the active tissue is determined to be cancerous, the method can further include applying anticancer therapy to the subject. Non-limiting examples of anticancer therapy can include surgery, chemotherapy, radiation therapy, hormone therapy, targeted therapy, immunotherapy, stem cell or bone marrow transplant, and experimental therapies like gene therapy, oncolytic virus therapy, and nanoparticle- based therapies.

[0082] In other examples, if the active tissue is determined to be cancerous, the method can further include performing additional imaging and triaging of the subject.

[0083] The image can include mammographic images, other radiological, and non-radiological images of the subject. The images can be sent to a computer, for example, a mobile device, a tablet, etc. In certain examples, the first region of tissue is or comprises tissue selected from a breast region, a brain region, a colon region, a dermal region, an esophagus region, a kidney region, a liver region, a lung region, an ovary region, a pancreatic region, a prostate region, a stomach region, a uterine region, or combinations thereof.

[0084] The first region is tissue can be considered active tissue if there is a difference of less than 0.2 between the roughness exponent and the reference value. In certain examples, the difference can be less than or equal to 0.15, 0.1, or 0.05.

[0085] The roughness exponent can be calculated using one or more multiscale analytical methods selected from a wavelet-transform modulus maxima, a wavelet leader, detrended fluctuation, and Fourier analysis.

[0086] In certain examples, the steps of receiving, calculating, comparing, and determining are each performed a plurality of time, each on a different region of tissue from the subject.

[0087] With reference to FIG.5, a system 100 for detecting cancer in a subject can comprise a processor 102; a memory medium 104; processer-executable instructions 106; and a display 108. In certain examples, the processor 102, the memory medium 104, and the display 108 are contained in one device, e.g., a computer. However, it should be appreciated that the processor 102, the memory medium 104, and the display 108 can be split across multiple devices and / or servers.

[0088] The processer 102 can be in communication with the memory medium 104. In certain examples, the processer 102 includes a plurality of processers. The memory medium 104 is tangible, non- transitory, and computer readable. The processor-executable instructions 106 are stored on the medium 104. When the processor-executable instructions 106, when executed, are configured to receive an image of a first region of tissue from a subject; calculate a roughness exponent for the first region of tissue; compare the roughness exponent of the first region of tissue to a reference value of 0.5, wherein a difference of less than 0.2 between the roughness exponent and the reference value indicates the first region of tissue is active tissue; and determine a range of multiscale anisotropic factors of the active tissue; and generate a detection output based on the range of multiscale anisotropic factors of the active tissue, wherein a subset of the multiscale anisotropic factors greater than one standard deviation above a benchmark average indicates that the active tissue is cancerous or pre-cancerous.

[0089] The detection output can be rendered on the display 108. The output can include a report which reflects whether the active tissue is cancerous or pre-cancerous.

[0090] The 2D Wavelet-Transform Modulus Maxima (WTMM) anisotropy method can be performed by considering vectors that are on so-called maxima chains, which are composed of positions where the modulus of the wavelet transform is locally maximal. The process of quantifying the departure from uniform directional gradients is by considering the distribution of angles, stored as a histogram with angle bins on the x-axis and counts on the y-axis. This distribution is then normalized to a probability density function (PDF). This can facilitate comparing images regardless of the number of local maxima. The area between the image PDF and a line at 1 / 2pi is then calculated. A line at 1 / 2pi is also a PDF with an area of one, but each angle bin from -180 degrees (-1 / 2pi) to 180 degrees (+1 / 2pi) has the same probability. This means every angle is equiprobable, purely isotropic. Deviations from this line occur due to angular information present and therefore the area calculated between the image PDF and the line is a quantitative representation of deviation from isotropy, i.e., the anisotropy factor. Its minimum value is 0, no deviation from equal angle probability, to 2, the maximum possible area between the two PDFs. The larger the anisotropy factor the more angular preference an image has at that wavelet size scale.

[0091] In certain embodiments, a method for predicting cancer in a subject can comprise receiving atleast one mammographic view. The mammographic view can include a left mediolateral oblique mammogram view, a left craniocaudal mammogram view, a right mediolateral oblique mammogram view, or a right craniocaudal mammogram view. The mammographic view includes a plurality of subregions. It should be appreciated that other medical imaging may also be employed for the method, especially for different types of cancer, e.g., ultrasound scans, computed tomography (CT) scans, etc.

[0092] The method further includes calculating a roughness exponent for each of the subregions. The roughness exponent can be calculated using one or more multiscale analytical methods selected from a wavelet-transform modulus maxima, a wavelet leader, detrended fluctuation, and Fourier analysis.

[0093] The method further includes comparing the roughness exponent of each of the subregions to a reference value of 0.5. A difference of less than 0.2 between the roughness exponent of one of the subregions and the reference value can indicate the one of the subregions is active tissue. The roughness exponent can be calculated using one or more multiscale analytical methods selected from a wavelet- transform modulus maxima, a wavelet leader, detrended fluctuation, and Fourier analysis. In certain examples, the difference between the roughness exponent and the reference value is less than or equal to 0.15, 0.1, or 0.05. In certain examples, the library of mammograms includes mammograms from patients who subsequently developed cancer about three years after the mammograms were obtained. Other methods for determining if tissue is active or passive may also be employed with the scope of this disclosure.

[0094] The method further includes a step of calculating an anisotropy factor for each of the subregions. The anisotropy factor measures directional preference by quantifying departure from uniform directional gradients. It can be calculated from the distribution of directional vector angles obtained from local maxima after a wavelet transform, typically with the first derivative of the Gaussian smoothing function. The Gaussian wavelet can be used at different smoothing size scales which leads this approach to being multiscale. Each vector angle corresponds to the direction of highest intensity gradient, which is represented by the modulus (amplitude) of the wavelet transform.

[0095] The method further includes a step of calculating an anisotropic mass for the mammographic view. The anisotropic mass can be a sum of all anisotropy factors for subregions subtyped as active, passive, fatty, or combinations thereof using their Hurst exponent. Therefore, the mammographic view can include an active dense tissue anisotropic mass, a passive dense tissue anisotropic mass, a fatty tissue anisotropic mass, or combinations thereof.

[0096] The method can further include a step of normalizing the anisotropic mass for the mammographic view by a mammographic area. The anisotropic mass is dependent on the overall number of subregions, so the size of the breast should be considered. In certain examples, the anisotropy factors and the anisotropy mass can be calculated at multiple size scales.

[0097] The method can further include determining a mean of the anisotropic mass of the mammogram view.

[0098] The method can further include comparing the active anisotropic mass for a mammographic view with a threshold active anisotropy mass value derived from a library of mammograms to determine a likelihood of cancer. The library of mammograms can include mammograms from patients who developed cancer. A threshold can be calibrated by calculating the number of cancer patients over all patients (cancer frequency) above and below different threshold values checking 1 percentile increments from the entire patient population. This can be performed at each wavelet scale as the WTMM anisotropy method is a multiscale approach. Patients above the threshold can be referred to as high-risk and patients below the threshold can be referred to as low-risk.

[0099] A method for predicting cancer in a subject can comprise receiving at least four mammographic views. The at least four mammographic views can include a left mediolateral oblique mammogram view, a left craniocaudal mammogram view, a right mediolateral oblique mammogram view, and a right craniocaudal mammogram view. Each of the four mammographic views includes a plurality of subregions. It should be appreciated that other medical imaging may also be employed for the method, especially for different types of cancer, e.g., ultrasound scans, computed tomography (CT) scans, etc.

[0100] The method further includes comparing the roughness exponent of each of the subregions to a reference value of 0.5. A difference of less than 0.2 between the roughness exponent of one of the subregions and the reference value can indicate the one of the subregions is active tissue. The roughness exponent can be calculated using one or more multiscale analytical methods selected from a wavelet- transform modulus maxima, a wavelet leader, detrended fluctuation, and Fourier analysis. In certain examples, the difference between the roughness exponent and the reference value is less than or equal to 0.15, 0.1, or 0.05. In certain examples, the library of mammograms includes mammograms from patients who subsequently developed cancer about three years after the mammograms were obtained. Other methods for determining if tissue is active or passive may also be employed with the scope of this disclosure.

[0101] The method further includes a step of calculating an anisotropy factor for each of the subregions. The anisotropy factor measures directional preference by quantifying departure from uniform directional gradients. It can be calculated from the distribution of directional vector angles obtained from local maxima after a wavelet transform, typically with the first derivative of the Gaussian smoothing function. The Gaussian wavelet can be used at different smoothing size scales which leads this approach to being multiscale. Each vector angle corresponds to the direction of highest intensity gradient, which is represented by the modulus (amplitude) of the wavelet transform.

[0102] The method further includes a step of calculating an anisotropic mass for each of the fourmammographic views. The anisotropic mass can be a sum of all anisotropy factors for subregions subtyped as active, passive, fatty, or combinations thereof using their Hurst exponent. Therefore, each mammographic view can include an active dense tissue anisotropic mass, a passive dense tissue anisotropic mass, a fatty tissue anisotropic mass, or combinations thereof.

[0103] The method can further include a step of normalizing the anisotropic mass for each of the four mammographic views by a mammographic area. The anisotropic mass is dependent on the overall number of subregions, so the size of the breast should be considered. In certain examples, the anisotropy factors and the anisotropy mass can be calculated at multiple size cales.

[0104] The method can further include determining a mean of the anisotropic mass of the left mediolateral oblique mammogram view and the left craniocaudal mammogram view, and a mean of the anisotropic mass of the right mediolateral oblique mammogram view and the right craniocaudal mammogram view; summing the mean of the anisotropic mass of the left mediolateral oblique mammogram view and the left craniocaudal mammogram view with the mean of the anisotropic mass of the right mediolateral oblique mammogram view and the right craniocaudal mammogram view to determine the active anisotropic mass for both breasts of the subject. This can facilitate an active anisotropic mass per a patient from both breasts, normalized by size, without double counting information from CC and MLO views.

[0105] The method can further include comparing the active anisotropic mass for both breasts of the subject with a threshold active anisotropy mass value derived from a library of mammograms to determine a likelihood of cancer. The library of mammograms can include mammograms from patients who developed cancer. A threshold can be calibrated by calculating the number of cancer patients over all patients (cancer frequency) above and below different threshold values checking 1 percentile increments from the entire patient population. This can be performed at each wavelet scale as the WTMM anisotropy method is a multiscale approach. Patients above the threshold can be referred to as high-risk and patients below the threshold can be referred to as low-risk.

[0106] EXAMPLE

[0107] Retrospective mammograms from 199 patients who developed cancer ~3 years later (tumor cases) vs.3717 controls were analyzed. Typically, the number of active dense tissue subregions in a mammogram is used as a predictor for cancer in our general logistic models. Other predictors used are the number of passive dense regions, fatty regions, mammographic percent density, and patient age.

[0108] In this novel approach, instead of counting each active dense tissue subregion, the anisotropy factor for that subregion is used, the overall sum of which is referred to as the anisotropic mass. If an active dense tissue subregion had a value near 0 for its anisotropy factor (i.e., purely isotropic) it would add a value near 0 to the total active dense anisotropic mass and if the value was near 2 (the maximum theoreticalvalue for the anisotropy factor), it would add 2. Therefore, active anisotropic mass can be a measure of the overall departure from isotropy for all the active dense tissue subregions in a respective mammogram. This can also be done for passive dense tissue or fatty tissue as well.

[0109] Individual mammogram analysis

[0110] The mammographic view active anisotropic mass for this dataset was calculated as follows. First, for each patient, either the left mediolateral oblique (LMLO), left craniocaudal (LCC), right mediolateral oblique (RMLO), or right craniocaudal (RCC), active tissue subregions were determined by calculating the roughness exponent and comparing the roughness exponent to a reference value. Then, the anisotropy factor for each subregion was determined. Next, for either of the four mammographic views, the anisotropy factors of all active subregions were summed to calculate the anisotropic mass for either of the four mammographic views. Then, the anisotropic mass for either of the four mammographic views was normalized by the mammographic area, i.e., the size of the breast, as the anisotropic mass is also dependent on the overall number of subregions.

[0111] The active anisotropic mass from the cancer and control patients’ individual views 3-years prior to their cancer diagnosis shows that cancer patients tend to have a higher active anisotropic mass across all wavelet scales, including the three scales shown in FIGS.11A-11D (0.5 mm), FIGS.12A-12D (1.4 mm), and FIGS.13A-13D (3.9 mm).

[0112] The predictive capability of anisotropic mass was determined by calibrating a threshold by calculating the number of cancer patients over all patients (cancer frequency) above and below different threshold values checking 1 percentile increments from the entire patient population from their individual mammographic views. This was performed at each wavelet scale as the WTMM anisotropy method is a multiscale approach. Patients’ individual mammographic views above the threshold were used to refer patients as high-risk and patients’ individual mammographic views below the threshold were referred as low-risk. For this particular dataset, the overall population cancer frequency is 5.08% (= 199 / (199+3717)). The best wavelet scale for thresholding was 0.5 mm for LCC, for LMLO, and for RCC, and it was 1.3 mm for RMLO, with threshold active anisotropy mass values of 0.103, 0.129, 0.098, and 0.199, for the LCC, LMLO, RCC, and RMLO views, respectively. This found 314, 275, 510, and 275 high-risk patients for the LCC, LMLO, RCC, and RMLO views, respectively, of which 9.5% (LCC), 9.4% (LMLO), 7.3% (RCC), and 9.8% (RMLO) went on to develop cancer in 3 years. The remaining 3604 (LCC), 3642 (LMLO), 3408 (RCC), and 3642 (RMLO) low-risk patients had a cancer frequency of 4.7% (LCC), 4.8% (LMLO), 4.8% (RCC), and 4.7% (RMLO). In other words, while ~1 in 20 women in the original dataset developed cancer in 3 years, this approach can identify a subset where 1 in 10 women develop cancer in 3-years with a reduced cancer frequency in the larger lower risk group. This is a new risk factor for breast cancer prediction and could be combined with other risk factors to create more powerful cancer risk predictionmodels. In addition, there is also a direct linear relationship between the log2(Wavelet Scale in Millimeters) and the threshold for this dataset showing stability of the method across all scales (FIGS.14A- 14D).

[0113] Patient level analysis

[0114] The patient level active anisotropic mass for this dataset was calculated as follows. First, for each patient, on all four mammographic views, e.g., left mediolateral oblique (LMLO), left craniocaudal (LCC), right mediolateral oblique (RMLO), and right craniocaudal (RCC), active tissue subregions were determined by calculating the roughness exponent and comparing the roughness exponent to a reference value. Then, the anisotropy factor for each subregion was determined. Next, for each of the four mammographic views, the anisotropy factors of all active subregions were summed to calculate the anisotropic mass for each of the four mammographic views. Then, the anisotropic mass for each of the four mammographic views was normalized by the mammographic area, i.e., the size of the breast, as the anisotropic mass is also dependent on the overall number of subregions.

[0115] The mean of the LMLO and LCC was calculated and summed with the mean of the RMLO and RCC view. This gives an active anisotropic mass per patient from both breasts, normalized by size, without double counting information from CC and MLO views.

[0116] As shown in FIGS.6-8, the active anisotropic mass from the cancer and control patients’ mammograms 3-years prior to their cancer diagnosis shows that cancer patients tend to have a higher active anisotropic mass across all wavelet scales, including the three scales shown in FIG.6 (0.5 mm), FIG.7 (1.4 mm), and FIG.8 (3.9 mm).

[0117] The predictive capability of anisotropic mass was determined by calibrating a threshold by calculating the number of cancer patients over all patients (cancer frequency) above and below different threshold values checking 1 percentile increments from the entire patient population. This was performed at each wavelet scale as the WTMM anisotropy method is a multiscale approach. Patients’ combined four mammographic views above the threshold were referred as high-risk and patients below the threshold were referred as low-risk. For this particular dataset, the overall population cancer frequency is 5.08% (= 199 / (199+3717)). The best wavelet scale for thresholding was 0.5 mm with a threshold active anisotropy mass value of 0.226. This found 235 high-risk patients of which 10.9 % went on to develop cancer in 3 years. The remaining 3681 low-risk patients had a cancer frequency of 4.75%. In other words, while ~1 in 20 women in the original dataset developed cancer in 3 years, this approach can identify a subset where 1 in 10 women develop cancer in 3-years with a reduced cancer frequency in the larger lower risk group. This is a new risk factor for breast cancer prediction and could be combined with other risk factors to create more powerful cancer risk prediction models. In addition, there is also a direct linear relationship between the log2(Wavelet Scale in Millimeters) and the threshold for this dataset showing stability of the method acrossall scales (FIGS.9-10).

[0118] To take advantage of the multi-scale power of this technique, every unique patient across all 38 wavelet scales who was placed into the high-risk category was determined. This was 372 unique high- risk patients with a cancer frequency of 8.89%, which is lower than a majority of individual scales, but captured more cancer cases than any individual scale. The low-risk group using all scales had the lowest cancer prevalence of any individual scale of 4.67%.

[0119] Certain embodiments of the devices and methods disclosed herein are defined in the above examples. It should be understood that these examples, while indicating particular embodiments of the invention, are given by way of illustration only. From the above discussion and these examples, one skilled in the art can ascertain the essential characteristics of this disclosure, and without departing from the spirit and scope thereof, can make various changes and modifications to adapt the devices and methods described herein to various usages and conditions. Various changes may be made and equivalents may be substituted for elements thereof without departing from the essential scope of the disclosure. In addition, many modifications may be made to adapt a particular situation or material to the teachings of the disclosure without departing from the essential scope thereof.

Claims

CLAIMS What is claimed is:

1. A method for detecting cancer in a subject, comprising: receiving an image of a first region of tissue from a subject; calculating a roughness exponent for the first region of tissue; comparing the roughness exponent of the first region of tissue to a reference value of 0.5, wherein a difference of less than 0.2 between the roughness exponent and the reference value indicates the first region of tissue is active tissue; determining a range of multiscale anisotropic factors of the active tissue, wherein a subset of the multiscale anisotropic factors greater than one standard deviation above a benchmark average indicates that the active tissue is cancerous or pre-cancerous; and applying anticancer therapy if the active tissue is determined to be cancerous.

2. The method of claim 1, wherein the roughness exponent is calculated using one or more multiscale analytical methods selected from a wavelet-transform modulus maxima, a wavelet leader, detrended fluctuation, and Fourier analysis.

3. The method of claim 1, wherein the steps of receiving, calculating, comparing, and determining are each performed a plurality of time, each on a different region of tissue from the subject.

4. The method of claim 1, wherein the first region of tissue is or comprises tissue selected from a breast region, a brain region, a colon region, a dermal region, an esophagus region, a kidney region, a liver region, a lung region, an ovary region, a pancreatic region, a prostate region, a stomach region, and a uterine region.

5. The method of claim 1, wherein the difference between the roughness exponent and the reference value is less than or equal to 0.

15.

6. The method of claim 1, wherein the difference between the roughness exponent and the reference value is less than or equal to 0.1.

7. The method of claim 1, wherein the difference between the roughness exponent and the reference value is less than or equal to 0.

05.

8. A system for detecting cancer in a subject, comprising: a memory medium having processor-executable instructions stored thereon, wherein the memory is tangible, non-transitory, and computer readable; a processor configured to execute processor-executable instructions, wherein the processor-executable instructions, when executed, are configured to: receive an image of a first region of tissue of a subject; calculate a roughness exponent for the first region of tissue; compare the roughness exponent of the first region of tissue to a reference value of 0.5, wherein a difference of less than 0.2 between the roughness exponent and the reference value indicates the first region of tissue is active tissue; determine a range of multiscale anisotropic factors of the active tissue; generate a detection output based on the multiscale anisotropic factors of the active tissue, wherein a subset of the multiscale anisotropic factors greater than one standard deviation above a benchmark average indicates that the active tissue is cancerous or pre-cancerous; and a display configured to render the detection output thereon.

9. The system of claim 8, wherein the roughness exponent is calculated using one or more multiscale analytical methods selected from a wavelet-transform modulus maxima, a wavelet leader, detrended fluctuation, and Fourier analysis.

10. The system of claim 8, wherein the steps of receiving, calculating, comparing, and determining are each performed a plurality of time, each on a different region of tissue from the subject.

11. The system of claim 8, wherein the first region of tissue is or comprises tissue selected from a breast region, a brain region, a colon region, a dermal region, an esophagus region, a kidney region, a liver region, a lung region, an ovary region, a pancreatic region, a prostate region, a stomach region, and a uterine region.

12. The system of claim 8, wherein the difference between the roughness exponent and the reference value is less than or equal to 0.

15.

13. The system of claim 8, wherein the difference between the roughness exponent and the reference value is less than or equal to 0.

1.

14. The system of claim 8, wherein the difference between the roughness exponent and the reference value is less than or equal to 0.

05.

15. A method for predicting cancer in a subject, comprising: receiving a mammographic view, wherein the mammographic view is a left mediolateral oblique mammogram view, a left craniocaudal mammogram view, a right mediolateral oblique mammogram view, or a right craniocaudal mammogram view, wherein the mammographic view includes a plurality of subregions; calculating a roughness exponent for each of the subregions; comparing the roughness exponent of each of the subregions to a reference value of 0.5, wherein a difference of less than 0.2 between the roughness exponent of one of the subregions and the reference value indicates the one of the subregions is active tissue; calculating an anisotropy factor for each of the subregions; calculating an anisotropic mass for the mammographic view; normalizing the anisotropic mass for the mammographic view by a mammographic area; determining a mean of the anisotropic mass of the mammogram view; and comparing the active anisotropic mass with a threshold active anisotropy mass value derived from a library of mammograms to determine a likelihood of cancer, wherein the library of mammograms includes mammograms from patients who developed cancer.

16. The method of claim 15, wherein the roughness exponent is calculated using one or more multiscale analytical methods selected from a wavelet-transform modulus maxima, a wavelet leader, detrended fluctuation, and Fourier analysis.

17. The method of claim 15, wherein the difference between the roughness exponent and the reference value is less than or equal to 0.15.

18. The method of claim 15, wherein the difference between the roughness exponent and the reference value is less than or equal to 0.

1.

19. The method of claim 15, wherein the difference between the roughness exponent and the reference value is less than or equal to 0.05.

Images