Three-dimensional volumetric analysis

Abstract

Described are machine learning platforms for performing three-dimensional volumetric analysis on acoustic neuromas such as vestibular schwannoma. The machine learning platforms contain a machine learning algorithm operably linked to a computer processor. The algorithm is configured to process data originating from medical images from magnetic resonance imaging. Processing the data involves three-dimensional volumetric analysis, utilizing a voxel size of about 0.8x0.8x0.9 mm or smaller, by the algorithm trained using image scans containing raw image scans and validated image scans. The data show that platforms containing a machine learning algorithm described herein not only speed up the segmentation process (less than 4 minutes per scan) but also offer improved depictions and measurements of tumor growth. This can enhance clinical decision-making, provide consistent monitoring of diseased tissue progression / regression (e.g., tumor progression / regression), and empower patients with detailed insights into their condition, thus facilitating better self-advocacy and treatment planning.

Description

[0001]ATTORNEY DOCKET NO. YU 8712 PCT THREE-DIMENSIONAL VOLUMETRIC ANALYSIS CROSS-REFERENCE TO RELATED APPLICATIONS This application claims the benefit of and priority to U.S. Provisional Application No. 63 / 662,928 filed June 21, 2024, which is incorporated herein by reference in its entirety. FIELD OF THE INVENTION This invention is generally related to processing and visualizing medical imaging data, particularly a machine learning algorithm for the segmentation and three-dimensional (3D) volumetric calculations of diseased tissues, so as to shorten the time required to conduct 3D volumetric analysis and improve image processing accuracy, preferably to (i) assess the presence of tumors and / or (ii) monitor the progression / regression of tumors. BACKGROUND OF THE INVENTION Linear and volumetric analysis are the typical methods to measure tumor size. However, limitations of linear and volumetric analysis persist requiring the need to develop accurate methods of measuring their sizes and / or tracking their growth in diseased tissues. For instance, traditional linear and volumetric analysis methods are less effective due to variability in measurement and tumor shape irregularities. Further, current methods of volumetric analysis that employ ellipsoid volume approximation can be inaccurate, particularly with irregularly shaped tumors. Three-dimensional (3D) volumetric analysis is another alternative to assessing tumor volumes. However, 3D volumetric analysis is not currently practical for clinical use as it requires time intensive manual segmentation and requires trained clinicians and engineers to perform. This means that diagnosis relies heavily on the experience and education of these individuals and often can be subjective. These limitations highlight the need for improved systems and / or methods for measuring diseased tissues, and remain an unmet need. Accordingly, it is an object of the invention to provide systems and / or methods for improved measurement and / or diagnosis of diseased tissues. It is also an object of the invention to provide artificial intelligence-led systems and / or methods for improved measurement and / or diagnosis of diseased tissues via 3D volumetric analysis. SUMMARY OF THE INVENTION Described are machine learning platforms for performing three-dimensional volumetric analysis on tumors. In some forms, the tumor is an acoustic neuroma, such as vestibular schwannoma that occurs in individuals with neurofibromatosis type 2 (NF2). 45741567.1 1 ATTORNEY DOCKET NO. YU 8712 PCT In some forms, the machine learning platforms contain a machine learning algorithm operably linked to a computer processor. The algorithm is configured to process data originating from one or more medical images from a medical imaging device, such as an magnetic resonance imaging (MRI) machine, wherein processing the data involves three-dimensional volumetric analysis by the algorithm. In some forms, the machine learning platforms contain a machine learning algorithm. When the algorithm is operably linked to a computer processor, the algorithm is configured to process data originating from one or more medical images from a medical imaging device, such as an MRI machine, wherein processing the data involves three-dimensional volumetric analysis by the algorithm. In some forms, the machine learning platforms contain a machine learning algorithm operably linked to a medical imaging device. The algorithm is configured to process data originating from one or more medical images from MRI, wherein processing the data involves three-dimensional volumetric analysis by the algorithm. In some forms, the machine learning platforms are as described above, except that processing the data further involves segmentation of the one or more medical images from a medical imaging device, such as an MRI machine by the machine learning algorithm. In some forms, the one or more medical images are from a diseased tissue or suspected diseased tissue. In some forms, the machine learning platforms are as described above, except that the machine learning algorithm is trained using image scans selected from raw image scans, validated image scans, or preferably both. Preferably, the image scans are obtained from (i) two or more patients and / or (ii) two or more scanning machines preferably from at least two different vendors of scanning machines. Preferably, at least one of the validated image scans is validated to have a tumor. In preferred forms, the three-dimensional volumetric analysis involves using a voxel size of about 0.8x0.8x0.9 mm or smaller. Also described is a non‐transitory computer-readable medium with computer executable instructions stored thereon executed by a processor to perform a method of processing one or more medical images from a medical imaging device, such as an MRI machine. The method involves: (i) collecting by the medium data originating from one or more medical images from medical imaging device, such as an MRI machine, and (ii) performing three-dimensional volumetric analysis preferably by a machine learning algorithm described herein. Also described is a computer-implemented method of measuring a three-dimensional volume of a tumor, the method involving preferably using a machine learning algorithm described herein 45741567.1 2 ATTORNEY DOCKET NO. YU 8712 PCT The data show that platforms containing a machine learning algorithm described herein not only speed up the segmentation process (less than 4 minutes per scan) but also offer improved depictions and measurements of tumor growth. This can enhance clinical decision- making, provide consistent monitoring of tumor progression / regression, and empower patients with detailed insights into their condition, thus facilitating better self-advocacy and treatment planning. The disclosed methods can be used to detect and characterize tumors, aid the diagnosis of cancer, and inform treatment of subject, and all such methods are provided alone in any combination. BRIEF DESCRIPTION OF THE DRAWINGS Figure 1 is a flowchart outlining the patient recruitment process to collect MRI images for segmentation. Figure 2 shows a pathway describing the process used to manually segment the vestibular schwannomas (VS) tumors from MRI images using the imaging processing software Simpleware ScanIP. Figure 3 shows an example of a tumor visualization using a tool created showing tumor growth. Recent tumor growth is shown by the darker, shaded areas in the model. Figure 4 shows examination of tumors slice by slice in all three planes (coronal, axial, sagittal views) to facilitate correction of any missing or extraneous voxels. Figure 5 is a line graph showing the increase in changes in tumor volumes as a result of increasing voxel sizes. DETAILED DESCRIPTION OF THE INVENTION I. Definitions “Operably linked” refers to the connection of at least two components in a system allowing them to work together via technology including, but not limited to, integrated circuits, electrical cables, ethernet, internet, intranet, Bluetooth, near field communication, WiFi, or a combination thereof. In one exemplary form, “operably linked” refers to a functional interaction between two or more components in a system, allowing them to work together to achieve a shared objective. The term “real-time” refers to data transmission to a user interface of a computer- implemented method, system, tool, or device within 1, 2, 3, 4, 5, 10, 15, 20, or no more than 30 minutes after a platform containing the disclosed machine learning algorithm is used to process data originating from one or more medical images, involving three-dimensional volumetric 45741567.1 3 ATTORNEY DOCKET NO. YU 8712 PCT analysis. The transmitted data can be the concentration of the analyte in a sample based, in part, on processing signals generated by the computer-implemented method, system, tool, or device. II. Machine learning platform for three-dimensional volumetric analysis Described are machine learning platforms for performing three-dimensional volumetric analysis on diseased tissues, particularly diseased tissues that are responsive to diagnostic agents (e.g., contrast agents such as gadolinium-based contrast agents) and / or are amenable to being represented in bulk volumes, e.g., three-dimensional volumes. In some forms, the machine learning platforms contain a machine learning algorithm operably linked to a computer processor. The algorithm is configured to process data originating from one or more medical images, wherein processing the data involves three-dimensional volumetric analysis by the algorithm. In some forms, the machine learning platforms contain a machine learning algorithm. When the algorithm is operably linked to a computer processor, the algorithm is configured to process data originating from one or more medical images, wherein processing the data involves three-dimensional volumetric analysis by the algorithm. In some forms, the machine learning platforms contain a machine learning algorithm operably linked to a medical imaging device. The algorithm is configured to process data originating from one or more medical images, wherein processing the data involves three- dimensional volumetric analysis by the algorithm. In some forms, the machine learning platforms are as described above, except that processing the data further involves segmentation of the one or more medical images by the machine learning algorithm. The term “segmentation” refers to the process of separating data into distinct groups. Typically, data in each group are similar of each other and different from data in other groups. In the context of images, segmentation involves identifying parts of the image and understanding to what object they belong. Segmentation can form the basis for performing object detection and classification. For an image of a biological tissue, for example, segmentation can mean identifying the background, tissue, parts of the tissue, and instruction (where present). Preferably, the one or more medical images are from a subject displaying one or more symptoms of a disease or disorder. In some forms, the one or more medical images are from a diseased tissue or suspected diseased tissue. In some forms, the machine learning platforms are as described above, except that the machine learning algorithm is trained using image scans selected from raw image scans, validated image scans, or preferably both. Optionally, the image scans are obtained from (i) two or more patients and / or (ii) two or more scanning machines preferably from at least two different 45741567.1 4 ATTORNEY DOCKET NO. YU 8712 PCT vendors of scanning machines. Preferably, at least one of the validated image scans is validated to have a diseased tissue, e.g., tumor. In some forms, the machine learning platforms are as described above, except that the three-dimensional volumetric analysis involves using a voxel size of about 0.8x0.8x0.9 mm or smaller. Also described is a non‐transitory computer-readable medium with computer executable instructions stored thereon executed by a processor to perform a method of processing one or more medical images. The method involves: (i) collecting by the medium data originating from one or more medical images, and (ii) performing three-dimensional volumetric analysis preferably by a machine learning algorithm described herein. Also described is a computer-implemented method of measuring a three-dimensional volume of a diseased tissue, the method involving preferably using a machine learning algorithm described herein. The medical images can be from a variety of image scanning methods and / or acquired using machines from one or more (preferably two or more) vendors. In some forms, the medical images are from methods that include, but not limited to, computed tomography (CT) scans, X- ray images, magnetic resonance images (MRI), ultrasound images, positron emission tomography images, magnetic resonance angiograms, and combinations thereof. In some forms, implanting the machine learning algorithm involves one or more convolution layers, activation function layers, pooling layers, or a combination thereof. The machine learning procedures may involve various supervised machine learning techniques, various semi-supervised machine learning techniques, and / or various unsupervised machine learning techniques. For instance, the machine learning procedures may utilize Logistic Regression, Gaussian Naive Bayes, Random Forest, Gradient boosting, Adaptive Boosting, LPBoost, TotalBoost, BrownBoost, MadaBoost, LogitBoost, Extra Trees, Linear Discriminant Analysis, Support Vector Machines, Decision Tree, k-nearest neighbor, alternating decision trees (ADTree), Decision Stumps, functional trees (FT), logistic model trees (LMT), linear classifiers, factor analysis, principal component analysis, neighborhood component analysis, sparse filtering, stochastic neighbor embedding, autoencoders, stacked autoencoders, neural networks, convolutional neural networks, feed forward neural network, Tabular Attention Network, or any other machine learning algorithm or statistical algorithm. Machine learning analyses may be performed using one or more of various programming languages and platforms, such as Python, R, Weka, and / or Matlab, for example. In some forms, the machine learning algorithm is configured to provide output from data processing to an individual in real-time via a user interface preferably a digital screen (e.g., a 45741567.1 5 ATTORNEY DOCKET NO. YU 8712 PCT screen of a computing device such as a screen of a smartphone, a laptop, a desktop computer, a watch, a tablet, etc.), an electro-mechanical acoustic system (e.g., electronic speakers), or a combination thereof. As described herein, the configuration of the machine learning algorithm provides enhancements and / or improvements compared to conventional approaches (e.g., linear analyses approaches) employed to identify defects in tissues from analyzing images. For instance, the machine learning algorithm can perform image registration to overlay 3D diseased models (e.g., preferably all 3D diseased models) onto previous imaging regardless of patient position. In doing so, the machine learning algorithm can identify in multiple dimensions where a diseased tissue is occurring and / or evaluate the impact of the diseased tissue on other anatomical structures. Further, and compared to conventional image processing approaches, the implementation of the imaging processing module shortens the time required to conduct 3D volumetric analysis and also improves image processing accuracy. This disclosed machine learning algorithm-driven volumetric approach emerges as a superior method for assessment of the size of a diseased tissue (e.g., tumor) due to how it rapidly provides clinically accessible three-dimensional information, particularly when compared to linear analysis. While linear measurements focus solely on the largest diameter of a diseased tissue (e.g., a tumor), the disclosed 3D volumetric analysis considers the entire volume, providing a more holistic understanding of the diseased tissue’s size and growth dynamics. The disclosed machine learning algorithm-driven volumetric approach accounts for irregular shapes and orientations of diseased tissues (e.g., tumors) and mitigates the impact of observer variation. By encompassing the entirety of a diseased tissue (e.g., tumor), the disclosed 3D volumetric analysis offers greater accuracy and reliability. Accordingly, the disclosed machine learning algorithm-driven volumetric approach (i) improves diagnoses of diseased tissue and / or help guide treatment and clinical decision making and / or (ii) provides the diagnoses in a shorter period of time. III. Diagnosis The disclosed methods can be used in diagnostic tests to assess diseased tissue in a subject, e.g., to distinguish between normal cells and diseased cells, and disease status. For example, disease status includes, without limitation, the presence or absence of disease (e.g., cancer v. non-cancer), characterization of cells including cancer, the risk of developing disease, the stage of the disease (e.g., non-invasive or early-stage cancer v. invasive or metastatic cancer), the progress of disease (e.g., progress of disease or remission of disease over time) and the effectiveness or response to treatment of disease. Based on this status, further procedures may be 45741567.1 6 ATTORNEY DOCKET NO. YU 8712 PCT indicated, including additional diagnostic tests or therapeutic procedures or regimens. Representative cancers and therapies are discussed in more detail below. The method typically involves, first, measuring the diseased tissue (e.g., tumor(s)) in a subject according using the methods described herein, and, second, comparing the measurement with a diagnostic amount or cut-off control that indicates the presence and / or characterizes the cancer. As is well understood in the art, by adjusting the particular diagnostic cut-off used in an assay, one can increase sensitivity or specificity of the diagnostic assay depending on the preference of the diagnostician. The particular diagnostic cut-off can be determined, for example, by measuring the volume of the diseased tissue (e.g., tumor) in a statistically significant number of samples from subjects with the different cancer statuses and drawing the cut-off to suit the diagnostician's desired levels of specificity and sensitivity. IV. Determining Risk of Developing Disease Methods for determining the risk of developing disease in a subject are also provided. Diseased tissue (e.g., tumor) sizes, volumes, or patterns can be characteristic of various risk states, e.g., high, medium, or low. The risk of developing a disease is determined by measuring the diseased tissue (e.g., tumor) and then either submitting them to a classification algorithm or comparing them with a reference amount and / or pattern and / or location and / or impedance on normal tissue that is associated with the particular risk level. V. Determining Stage of Disease Another form provides methods for determining the stage of disease in a subject. Each stage of the disease can have a characteristic size, volume, and / or pattern of diseased tissue (e.g., tumor(s)). The stage of a disease can be determined by measuring the diseased tissue (e.g., tumor) and then either submitting them to a classification algorithm or comparing them with a reference size, volume, and / or pattern of diseased tissue (e.g., tumor) that is associated with the particular stage. VI. Determining Course (Progression / Remission) of Disease Still another form provides methods for determining the course of disease in a subject. Disease course refers to changes in disease status over time, including disease progression (worsening) and disease regression (improvement). Over time, the amounts or relative amounts (e.g., the pattern) of the diseased tissue (e.g., tumor) changes. This method involves measuring one or more diseased tissue (e.g., tumor)s in a subject at least two different time points, e.g., a first time and a second time, and comparing the change in amounts, if any. The course of disease is determined based on these comparisons. Similarly, this method is useful for determining the response to treatment. If a treatment is effective, then the diseased tissue (e.g., tumor) will 45741567.1 7 ATTORNEY DOCKET NO. YU 8712 PCT stabilize or regress, while if treatment is ineffective, the diseased tissue (e.g., tumor) not regress and will typically grow. VII. Subject Management In certain forms of the method including the detection and / or analysis of one or more diseased tissue (e.g., tumor)s further include managing subject treatment based on the status. Such management includes the actions of the physician or clinician subsequent to determining cancer status. For example, if a physician makes a diagnosis of cancer, then a certain regime of treatment, such as prescription or administration of chemotherapy, radiation, immunotherapy, including, but not limited to administration of the compositions discussed in more detail below, might follow. Alternatively, a diagnosis of non-cancer or benign diseased tissue (e.g., tumor) might be followed with further testing to determine a specific disease that the patient might be suffering from. Also, if the diagnostic test gives an inconclusive result on cancer, further tests may be required. Additional forms relate to the communication of assay results or diagnoses or both to technicians, physicians or patients, for example. In certain forms, computers will be used to communicate assay results or diagnoses or both to interested parties, e.g.: physicians and their patients. In some forms, the assays will be performed or the assay results analyzed in a country or jurisdiction which differs from the country or jurisdiction to which the results or diagnoses are communicated. In a preferred form a diagnosis is communicated to the subject as soon as possible after the diagnosis is obtained. The diagnosis may be communicated to the subject by the subject's treating physician. Alternatively, the diagnosis may be sent to a test subject by email or communicated to the subject by phone. A computer may be used to communicate the diagnosis by email or phone. In certain forms, the message containing results of a diagnostic test may be generated and delivered automatically to the subject using a combination of computer hardware and software which will be familiar to artisans skilled in telecommunications. In certain forms all or some of the method steps, including the assaying of samples, diagnosing of diseases, and communicating of assay results or diagnoses, may be carried out in diverse (e.g., foreign) jurisdictions. VIII. Assessing the Effectiveness of Treatment or Risk for Developing Cancer Methods for determining the course of cancer in a subject are also provided. Disease course refers to changes in disease status over time, including disease progression (worsening) and disease regression (improvement). Over time, the amounts or relative amounts (e.g., the pattern) of the diseased tissue (e.g., tumor(s)) changes. Accordingly, this method involves measuring one or more diseased tissues (e.g., tumors) in a subject at least two different time 45741567.1 8 ATTORNEY DOCKET NO. YU 8712 PCT points, e.g., a first time and a second time, and comparing the change in amounts, if any. The course of disease is determined based on these comparisons. Similarly, this method is useful for determining the response to treatment. If a treatment is effective, then the diseased tissue (e.g., tumor) will stabilize or regress, while if treatment is ineffective, the diseased tissue (e.g., tumor) will typically grow. IX. Methods of Treatment Any of the disclosed methods can be coupled to a method of treating a subject in need thereof. Thus, any of the disclosed methods can further include treating a positive for cancer with a treatment known to be effective and / or preferred for treating subjects with the cancer. In certain forms, the compositions are administered systemically, locally, or regionally. In some forms, the compositions are taken orally, injected, topically applied, or otherwise administered directly into the vasculature or onto vascular tissue at or adjacent to a site of cancerous growth. Typically, local administration causes an increased localized concentration of the compositions, which is greater than that which can be achieved by systemic administration. X. Therapies In some forms, patients are also subject to one or more therapies or procedures for the treatment of the disease or disorder. When two or more therapies or procedures are used, they can be simultaneous or sequential combination therapy. In some forms, the therapy is a conventional treatment for cancer, more preferably a conventional treatment for the particular cancer type, e.g., prostate or breast cancer. For example, in some forms, the additional therapy or procedure is surgery, a radiation therapy, or chemotherapy. In some forms, the conventional cancer therapy is in the form of one or more active agents. Therefore, in some forms, the methods administer compositions in combination with one or more additional active agents. Such active agent can be, for example, chemotherapeutic agents, cytokines, chemokines, radiation therapy, or immunotherapy. The majority of chemotherapeutic drugs can be divided into alkylating agents, antimetabolites, anthracyclines, plant alkaloids, topoisomerase inhibitors, and other antitumor agents. These drugs affect cell division or DNA synthesis and function in some way. Therapeutics include monoclonal antibodies and the tyrosine kinase inhibitors e.g., imatinib mesylate (GLEEVEC® or GLIVEC®), which directly targets a molecular abnormality in certain types of cancer (chronic myelogenous leukemia, gastrointestinal stromal tumors). In some forms, the therapy is a chemotherapeutic agent. Representative chemotherapeutic agents include, but are not limited to, amsacrine, bleomycin, busulfan, camptothecin, capecitabine, carboplatin, carmustine, chlorambucil, cisplatin, cladribine, clofarabine, 45741567.1 9 ATTORNEY DOCKET NO. YU 8712 PCT crisantaspase, cyclophosphamide, cytarabine, dacarbazine, dactinomycin, daunorubicin, docetaxel, doxorubicin, epipodophyllotoxins, epirubicin, etoposide, etoposide phosphate, fludarabine, fluorouracil, gemcitabine, hydroxycarb amide, idarubicin, ifosfamide, innotecan, leucovorin, liposomal doxorubicin, liposomal daunorubici , lomustine, mechlorethamine, melphalan, mercaptopurine, mesna, methotrexate, mitomycin, mitoxantrone, oxaliplatin, paclitaxel, pemetrexed, pentostatin, procarbazine, raltitrexed, satraplatin, streptozocin, teniposide, tegafur-uracil, temozolomide, teniposide, thiotepa, tioguanine, topotecan, treosulfan, vinblastine, vincristine, vindesine, vinorelbine, vorinostat, taxol, trichostatin A and derivatives thereof, trastuzumab (HERCEPTIN®), cetuximab, and rituximab (RITUXAN® or MABTHERA®), bevacizumab (AVASTIN®), and combinations thereof. Representative pro- apoptotic agents include, but are not limited to, fludarabinetaurosporine, cycloheximide, actinomycin D, lactosylceramide, 15d-PGJ(2)5, and combinations thereof. In some forms, the treatment is or includes immunotherapy such as inhibition of checkpoint proteins such as components of the PD-1 / PD-L1 axis or CD28-CTLA-4 axis using one or more immune checkpoint modulators (e.g., PD-1 antagonists, PD-1 ligand antagonists, and CTLA4 antagonists), adoptive T cell therapy, and / or a cancer vaccine. Exemplary immune checkpoint modulators used in immunotherapy include Pembrolizumab (anti-PD1 mAb), Durvalumab (anti-PDL1 mAb), PDR001 (anti-PD1 mAb), Atezolizumab (anti-PDL1 mAb), Nivolumab (anti-PD1 mAb), Tremelimumab (anti-CTLA4 mAb), Avelumab (anti-PDL1 mAb), and RG7876 (CD40 agonist mAb). In some forms, the treatment is or includes adoptive T cell therapy. Methods of adoptive T cell therapy are known in the art and used in clinical practice. Generally adoptive T cell therapy involves the isolation and ex vivo expansion of tumor-specific T cells to achieve greater number of anti-tumor T cells than what could be obtained by vaccination alone. The tumor- specific T cells are then infused into patients with cancer in an attempt to give their immune system the ability to overwhelm remaining tumor via T cells, which can attack and kill the cancer. Several forms of adoptive T cell therapy can be used for cancer treatment including, but not limited to, culturing tumor infiltrating lymphocytes or TIL; isolating and expanding one particular T cell or clone; and using T cells that have been engineered to recognize and attack tumors. In some forms, the T cells are taken directly from the patient's blood. Methods of priming and activating T cells in vitro for adaptive T cell cancer therapy are known in the art. See, for example, Wang, et al, Blood, 109(11):4865-4872 (2007) and Hervas-Stubbs, et al, J. Immunol.,189(7):3299-310 (2012). In some forms, the treatment is or includes a cancer vaccine. Vaccination typically includes administering a subject an antigen (e.g., a cancer antigen) together with an adjuvant to 45741567.1 10 ATTORNEY DOCKET NO. YU 8712 PCT elicit therapeutic T cells in vivo. In some forms, the cancer vaccine is a dendritic cell cancer vaccine in which the antigen is delivered by dendritic cells primed ex vivo to present the cancer antigen. Examples include PROVENGE® (sipuleucel-T), which is a dendritic cell-based vaccine for the treatment of prostate cancer (Ledford, et al., Nature, 519, 17–18 (05 March 2015). Such vaccines and other compositions and methods for immunotherapy are reviewed in Palucka, et al., Nature Reviews Cancer, 12, 265-277 (April 2012). In some forms, the compositions and methods are used prior to or in conjunction with surgical removal of tumors, for example, in preventing primary tumor metastasis. In some forms, the compositions and methods are used to enhance the body’s own anti-tumor immune functions. XI. Subjects The subject of the disclosed methods typically have or are suspected of having diseased tissue In some forms, the diseased tissue is a tumor. The tumor can be cancerous. Cancer is a disease of genetic instability, allowing a cancer cell to acquire the hallmarks proposed by Hanahan and Weinberg, including (i) self-sufficiency in growth signals; (ii) insensitivity to anti-growth signals; (iii) evading apoptosis; (iv) sustained angiogenesis; (v) tissue invasion and metastasis; (vi) limitless replicative potential; (vii) reprogramming of energy metabolism; and (viii) evading immune destruction (Cell.,144:646–674, (2011)). Tumors, which can be treated in accordance with the disclosed methods, are classified according to the embryonic origin of the tissue from which the tumor is derived. Carcinomas are tumors arising from endodermal or ectodermal tissues such as skin or the epithelial lining of internal organs and glands. Sarcomas, which arise less frequently, are derived from mesodermal connective tissues such as bone, fat, and cartilage. The leukemias and lymphomas are malignant tumors of hematopoietic cells of the bone marrow. Leukemias proliferate as single cells, whereas lymphomas tend to grow as tumor masses. Malignant tumors may show up at numerous organs or tissues of the body to establish a cancer. The described compositions and methods are useful for detecting, diagnosing, treating, and / or alleviating benign or malignant tumors in subjects. The disclosed compositions and methods of treatment thereof are generally suited for treatment of carcinomas, sarcomas, lymphomas, etc., provided the tumor or other cancer mass can be detected by the medical imaging device. In some forms, the cancer includes one or more solid tumors. Other types of cancer that can be treated with the provided compositions and methods include, but are not limited to, cancers such as vascular cancer such as multiple myeloma, adenocarcinomas and sarcomas, of bone, bladder, brain, breast, cervical, colorectal, esophageal, 45741567.1 11 ATTORNEY DOCKET NO. YU 8712 PCT kidney, liver, lung, nasopharangeal, pancreatic, prostate, skin, stomach, and uterine. In some forms, the compositions are used to treat multiple cancer types concurrently. The compositions can also be used to treat metastases or tumors at multiple locations. Exemplary tumor cells include, but are not limited to, tumor cells of cancers, including leukemias including, but not limited to, acute leukemia, acute lymphocytic leukemia, acute myelocytic leukemias such as myeloblastic, promyelocytic, myelomonocytic, monocytic, erythroleukemia leukemias and myelodysplastic syndrome, chronic leukemias such as, but not limited to, chronic myelocytic (granulocytic) leukemia, chronic lymphocytic leukemia, hairy cell leukemia; polycythemia vera; lymphomas such as, but not limited to, Hodgkin’s disease, non-Hodgkin’s disease; multiple myelomas such as, but not limited to, smoldering multiple myeloma, nonsecretory myeloma, osteosclerotic myeloma, plasma cell leukemia, solitary plasmacytoma and extramedullary plasmacytoma; Waldenström’s macroglobulinemia; monoclonal gammopathy of undetermined significance; benign monoclonal gammopathy; heavy chain disease; bone and connective tissue sarcomas such as, but not limited to, bone sarcoma, osteosarcoma, chondrosarcoma, Ewing’s sarcoma, malignant giant cell tumor, fibrosarcoma of bone, chordoma, periosteal sarcoma, soft-tissue sarcomas, angiosarcoma (hemangiosarcoma), fibrosarcoma, Kaposi’s sarcoma, leiomyosarcoma, liposarcoma, lymphangiosarcoma, neurilemmoma, rhabdomyosarcoma, synovial sarcoma; brain tumors including, but not limited to, glioma, astrocytoma, brain stem glioma, ependymoma, oligodendroglioma, nonglial tumor, acoustic neurinoma, craniopharyngioma, medulloblastoma, meningioma, pineocytoma, pineoblastoma, primary brain lymphoma; breast cancer including, but not limited to, adenocarcinoma, lobular (small cell) carcinoma, intraductal carcinoma, medullary breast cancer, mucinous breast cancer, tubular breast cancer, papillary breast cancer, Paget’s disease, and inflammatory breast cancer; adrenal cancer, including, but not limited to, pheochromocytom and adrenocortical carcinoma; thyroid cancer such as but not limited to papillary or follicular thyroid cancer, medullary thyroid cancer and anaplastic thyroid cancer; pancreatic cancer, including, but not limited to, insulinoma, gastrinoma, glucagonoma, vipoma, somatostatin-secreting tumor, and carcinoid or islet cell tumor; pituitary cancers including, but not limited to, Cushing’s disease, prolactin-secreting tumor, acromegaly, and diabetes insipius; eye cancers including, but not limited to, ocular melanoma such as iris melanoma, choroidal melanoma, and ciliary body melanoma, and retinoblastoma; vaginal cancers, including, but not limited to, squamous cell carcinoma, adenocarcinoma, and melanoma; vulvar cancer, including, but not limited to, squamous cell carcinoma, melanoma, adenocarcinoma, basal cell carcinoma, sarcoma, and Paget’s disease; cervical cancers including, but not limited to, squamous cell carcinoma, and adenocarcinoma; uterine cancers including, but not limited to, endometrial carcinoma and 45741567.1 12 ATTORNEY DOCKET NO. YU 8712 PCT uterine sarcoma; ovarian cancers including, but not limited to, ovarian epithelial carcinoma, borderline tumor, germ cell tumor, and stromal tumor; esophageal cancers including, but not limited to, squamous cancer, adenocarcinoma, adenoid cyctic carcinoma, mucoepidermoid carcinoma, adenosquamous carcinoma, sarcoma, melanoma, plasmacytoma, verrucous carcinoma, and oat cell (small cell) carcinoma; stomach cancers including, but not limited to, adenocarcinoma, fungating (polypoid), ulcerating, superficial spreading, diffusely spreading, malignant lymphoma, liposarcoma, fibrosarcoma, and carcinosarcoma; colon cancers; rectal cancers; liver cancers including, but not limited to, hepatocellular carcinoma and hepatoblastoma, gallbladder cancers including, but not limited to, adenocarcinoma; cholangiocarcinomas including, but not limited to, papillary, nodular, and diffuse; lung cancers including, but not limited to, non-small cell lung cancer, squamous cell carcinoma (epidermoid carcinoma), adenocarcinoma, large-cell carcinoma and small-cell lung cancer; testicular cancers including, but not limited to, germinal tumor, seminoma, anaplastic, classic (typical), spermatocytic, nonseminoma, embryonal carcinoma, teratoma carcinoma, choriocarcinoma (yolk-sac tumor), prostate cancers including, but not limited to, adenocarcinoma, leiomyosarcoma, and rhabdomyosarcoma; penal cancers; oral cancers including, but not limited to, squamous cell carcinoma; basal cancers; salivary gland cancers including, but not limited to, adenocarcinoma, mucoepidermoid carcinoma, and adenoidcystic carcinoma; pharynx cancers including, but not limited to, squamous cell cancer, and verrucous; skin cancers including, but not limited to, basal cell carcinoma, squamous cell carcinoma and melanoma, superficial spreading melanoma, nodular melanoma, lentigo malignant melanoma, acral lentiginous melanoma; kidney cancers including, but not limited to, renal cell cancer, adenocarcinoma, hypernephroma, fibrosarcoma, transitional cell cancer (renal pelvis and / or uterer); Wilms’ tumor; bladder cancers including, but not limited to, transitional cell carcinoma, squamous cell cancer, adenocarcinoma, and carcinosarcoma. For a review of such disorders, see Fishman et al., 1985, Medicine, 2d Ed., J.B. Lippincott Co., Philadelphia and Murphy et al., 1997, Informed Decisions: The Complete Book of Cancer Diagnosis, Treatment, and Recovery, Viking Penguin, Penguin Books U.S.A., Inc., United States of America). The systems and methods herein described are further illustrated in the following examples, which are provided by way of illustration and are not intended to be limiting. Having described the ground truth data and components of the machine learning platform, it will require no more that routine skill for those of skill in the art to extend to other diseased tissues, particularly diseased tissues that are responsive to diagnostic agents (e.g., contrast agents such as gadolinium-based contrast agents) and / or are amenable to being represented in bulk volumes, e.g., three-dimensional volumes. It will be appreciated that variations in components and 45741567.1 13 ATTORNEY DOCKET NO. YU 8712 PCT alternatives in elements of the components shown will be apparent to those skilled in the art and are within the scope of disclosed forms. Use of the term "about" is intended to describe values either above or below the stated value in a range of approx. + / - 10%. Examples of values within this range are + / - 1%, + / - 2%, + / - 3%, + / - 4%, + / - 5%, + / - 6%, + / - 7%, + / - 8%, + / - 9%, and + / - 10%. Examples Example 1: The Development of an Artificial Intelligence Auto-Segmentation Tool for 3D Volumetric Analysis of Vestibular Schwannomas Neurofibromatosis Type 2 (NF2)-related schwannomatosis (previously known as Neurofibromatosis type II) is a rare, autosomal dominant disorder caused by mutations in NF2 gene on chromosome 22q12. It’s characterized by tumors in the nervous system, such as vestibular schwannomas, spinal meningioma, and peripheral nerve tumors1. Patients with vestibular schwannomas suffer from hearing loss, tinnitus, facial palsy, and a reduced life expectancy2. Radiologic techniques are vital in the diagnosis of NF2 and, regular screening is indicated.1,3Magnetic resonance imaging (MRI) has replaced computed tomography (CT) as the gold-standard imaging for diagnosing and monitoring vestibular schwannomas due to its high sensitivity and specificity.4,5T1-weighted MRI is particularly effective in delineating tumor shape, size, location and reflecting mass effects.6,7Vestibular schwannomas (VS) tumors enhance significantly after intravenous gadolinium contrast.8Use of contrast-enhanced, T1-weighted MRI used with T2- weighted or Fluid-attenuated inversion recovery (FLAIR) sequences can distinguish peritumoral cysts and edema, which enhance heterogenously.9As VS tumors sizes vary, higher resolution MRIs with small voxel sizes are optimal to capture morphologic details. Dombi et. al. found that slice thickness should be less than 1mm.10Linear and volumetric analysis are current methods to measure tumor size. In linear analysis, unidimensional or bidimensional measurements of the largest tumor diameter are assessed on axial or coronal views of MRIs10,12. Numerous factors can decrease sensitivity of linear measurements, including patient orientation, oblique orientations and irregular shape of tumors and high levels of observer variation.11,12Volumetric analysis integrates an additional dimension of measurement. By approximating the tumor as an ellipsoid in every MR slice, volumetric analysis is more sensitive to tumor progression compared to 2D measurements. However, studies exploring volumetric analysis of VS tumors have found that these approximations overestimate volume. Cross- 45741567.1 14 ATTORNEY DOCKET NO. YU 8712 PCT sectional slices of VS tumors deviate from an ellipsoid shape as they develop extra-canalicular components extending into the cerebellopontine angle, adopting the “ice cream cone” shape.13Considering the limitations of linear and volumetric analysis, 3D volumetric analysis has gained recognition as an accurate method of tracking growth, where tumors are segmented on each MRI slice and the area is multiplied by slice thickness. This eliminates error introduced by approximating or assessing the longest diameter. Despite greater sensitivity14, 3D volumetric analysis is not currently practical for clinical use as it requires time intensive manual segmentation.13The goal of the study is to develop an AI-led approach to perform the segmentation and 3D volumetric calculations of these tumors, so as to shorten the time required to conduct 3D volumetric analysis and improve image processing accuracy. Materials and methods Based on initial trials with a dataset of 10 images, we determined that 150 MRI images can be used for a ground truth data set to achieve a desired AI accuracy. This was determined using statistical power analysis based on previous research in DICE score modeling15,16. Patient recruitment: 77 patients were identified through the Yale New Haven Hospital medical database (Figure 1). Of these, 24 patients were eligible for inclusion. The patient records of these patients produced 84 MRIs from which VS tumor models could successfully be made. To obtain a diverse set of data and to increase the numbers of patients, the researchers contacted patients through NF2 specific Non-Governmental Organisation NF2 BioSolutions. An initial recruitment email was sent out to all members on the mailing list (n=1000) with a return rate of 7.2% responding with interest.13 patients were included in the final stage, providing 70 MRIs. Tumor models could be successfully made from 59 MRIs. Scans without contrast, scans with large voxel sizes or scan sequences that did not allow for visualization of the inner auditory canal (less than 150 images) were excluded. From Yale New Haven Hospital and public patient recruitment, 143 MRIs were included in the ground truth dataset. Creation of initial data set: The quality of images was determined by the researchers and was categorized by voxel size. Scans were categorized into high (less than 0.5x0.5x1.0), medium (less than 1.0x1.0x1.0) and low quality (greater than 1.0x1.0x1.0). To create the tumor models an image processing software (Simpleware ScanIP, Synopsys, Mountain View, CA) was used. For the initial implementation, 1103D models were used; 66 high quality, 44 medium quality and 6 low quality scans.3D tumor masks were created for vestibular schwannomas (unilateral or bilateral) as shown in Figure 2. 45741567.1 15 ATTORNEY DOCKET NO. YU 8712 PCT To highlight the vestibular schwannomas, a thresholding algorithm was used on selected slices containing the tumor mass. A ‘split regions’ algorithm was used to isolate the tumors and remove the non-tumorous voxels. Consideration was taken with voxels lining the border of the mask. Missing and surplus voxels were adjusted to include or exclude as needed using the paint function. After the initial mask creation, the tumors were re-examined in all three planes (coronal, axial, sagittal views) to correct for any missing or extra voxels. All the models were reviewed by a neuroradiologist, who made needed adjustments. The ‘volume’ measurement tool was used to calculate the 3D volume (in mm3) of each tumor mask. To visualize the shape, size, and pattern of growth of the masked tumors, a mask was created of the pons at the levels of the tumors. Creation of prototype: The initial data set used by the engineers at Synopsys consisted of 25 high quality MRIs. The ground truth dataset of 143 MRIs was subdivided into train (80%), validation (10%) and test (10%) groups. The helper (DPP V1.0) was trained using proprietary AI- and ML-based algorithms and information. No tumors identified within the ground truth segmentation were missed by the helper. A final testing stage was completed using 30 new segmentations of MRI scans obtained from NF2 Biosolutions. This stage of training was used for validating the segmentations produced by the tool to verify its ability to identify and segment tumors in previously unseen patient data. Following this testing stage, an additional tool was added to the modeler which corrects the orientation of the images when imported into the software to accommodate different imaging protocols. To compare the accuracy of the AI generated 3D models to the radiologist validated manual segmentation models, a DICE score was calculated. The DICE score calculation was calculated using the equation: DICE Coefficient = 2 * the Area of Overlap / by the total number of pixels in both images. Development of visualization tool: To compare the chronological growth of a patient’s tumors, each patient who had multiple scans had their segmented tumor masks imported into a single ScanIP file. An image registration algorithm was used to reformat and align the brain across the DICOM images and the generated 3D tumor models. Accessing the scripting interface in ScanIP, custom code was written to organize bilateral tumors in chronological order and sort them into left and right categories. 45741567.1 16 ATTORNEY DOCKET NO. YU 8712 PCT Plots were generated with a script illustrating the change in size of each tumor over time. Plots indicating percentage change from baseline and volatility were also generated. A color scheme that illustrates chronological tumor growth was developed. The overlapping 3D models of each tumor were displayed in a single color. Models of tumors from earlier scans are displayed in lighter shades whereas models from later scans are displayed in darker shades, Figure 3. Results A mean DICE score of 0.76 (standard deviation 0.21) was achieved in an initial testing of the model. After the final testing stage, the final mean DICE score was 0.88 (range 0.74-0.93, standard deviation 0.04). 45741567.1 17 ATTORNEY DOCKET NO. YU 8712 PCT Table 1. Table showing improvement in DICE scores between the initial tool and the latest version of the AI modeler. . ATTORNEY DOCKET NO. YU 8712 PCT Mean 0.76 0.88 Standard Deviation 021 004 nd the final DPP version of the AI modeler in a set of the same 24 images. For example, significant improvement can be seen in Image 7 where the DICE score improved from 0.14 to 0.84. Our study has demonstrated significant findings of 3D volumetric tumor analysis through AI-driven, automated image processing. We have demonstrated the capability to automate the processing of VS tumors with a credible overall DICE score of 0.88, demonstrating the tool's accuracy and reliability. Investigation and testing revealed the versatility of the AI tool across various MRI T1 sequences, accommodating different voxel sizes without compromising efficacy. We have demonstrated the tool's proficiency in analyzing sequential chronological scans, offering valuable insights into tumor growth progression and treatment response over time. These findings collectively underscore the potential of the AI-driven methodology to revolutionize VS tumor assessment and monitoring. This new AI driven volumetric approach emerges as a superior method for tumor size assessment due to how it rapidly provides clinically accessible three-dimensional information, particularly when compared to linear analysis. Volumetric analysis is superior to linear assessments, but there are limitations to this, including the high-intensive process required to complete volumetric analysis17. While linear measurements focus solely on the largest diameter of the tumor, 3D volumetric analysis considers the entire volume, providing a more holistic understanding of the tumor’s size and growth dynamics18. This approach accounts for irregular shapes and orientations of tumors and mitigates the impact of observer variation. By encompassing the entirety of the tumor, 3D volumetric analysis offers greater accuracy and reliability, important for informing clinical decisions19. The ground truth dataset used in this sample was heterogeneous, in that it included a wide range of patients, MRI scanners (1.5T and 3T scanners), manufacturers and voxel sizes ranging from 0.375x0.375x1.0 to 1.0x1.0x1.0. This heterogenicity in the dataset and the overall strong performance of the AI demonstrated by the high DICE scores allows for the AI driven tool to be applicable to a wide range of clinical settings. While 3D volumetric analysis has been found to yield the most accurate measures of tumor size, manual segmentation of tumors is time intensive and requires trained clinicians and engineers to perform, introducing variability in measurements. Our algorithms use AI algorithms to provide a fully automated tool, without subjectivity, error, and the need for extensive clinical 45741567.1 19 ATTORNEY DOCKET NO. YU 8712 PCT training. This streamlined process provides more accurate measures of tumor size than linear segmentation while taking only a fraction of the time (less than 4 minutes). Additionally, the described AI-segmented models are objective in their creation, reducing subjectivity. The repeatability of the process increases the clinical validity of these measurements. Comparing sequential imaging to determine tumor growth is challenging, given patients’ heads are rarely positioned in the MRI scanner in the exact same position between scans. To reduce the uncertainty introduced by this, the software performs image registration to overlay all the 3D tumor models onto previous imaging regardless of patient position. In doing so, we can identify in multiple dimensions where tumor growth is occurring and evaluate the impact of the tumor on structures such as the pons and cochlear nerve. Clinical trials could also use this tool to determine the effectiveness of their investigational drug or interventional procedures. The 3D morphometric technique would standardize the measurement of VS tumors in clinical trials and treatment. For individuals diagnosed with NF2-SWN, having access to their 3D tumor growth can reinforce an individual’s ability to self-advocate. The 3D modeling tools developed can help patients understand their disease and its progression. This tool has the unique ability to provide patients with the possibility of being actively involved with their treatment plan and understanding their tumors in more detail. There is an unmet need for a reliable and easy to use tool to confidently evaluate tumor growth over time. A standardized process greatly reduces the potential for error in the interpretation, increasing efficiency and accuracy of VS volumetric analysis. This tool will be an aid to experienced clinicians and be beneficial to early-career radiologists to better visualize and assess tumor size, morphology, and growth. As shown above, the AI tool created has not produced a DICE score of 1 indicating that there is still some discrepancy between manual and AI segmentations. This requires the oversight of a radiologist or trained researcher to validate scans to verify segmentations. Validation by a radiologist can be applied in cases when meningiomas are close to the vestibular schwannoma to validate correct identification. Researchers also identified a small number of tumors with a morphology not previously encountered by the AI such as when tumors have undergone debulking procedures. These would also benefit from additional validation. Manual segmentation was done by human operators, which could have introduced errors in their determination of the extent of the tumor. Refining of the AI will continue for improvement in assessing unconventional cases or previously unencountered tumor morphologies. 45741567.1 20 ATTORNEY DOCKET NO. YU 8712 PCT In conclusion, our study has demonstrated an efficient, accurate AI for the 3D volumetric analysis for vestibular schwannomas. The use of this AI will enable faster 3D volumetric analysis compared to manual segmentation. The tool will be a method of assessing tumor growth through volume measurements and allow clinicians to make more informed decisions. One key area of future research will focus on predictive growth of tumors based on the comparison of previous growth rate of analyzed tumors. REFERENCES 1. Asthagiri AR, Parry DM, Butman JA, et al. Neurofibromatosis type 2. Lancet. 2009;373(9679):1974-1986. doi:https: / / doi.org / 10.1016 / s0140-6736(09)60259-2 2. Evans DGR, Birch JM, Ramsden RT. Paediatric presentation of type 2 neurofibromatosis. Archives of disease in childhood.1999;81(6):496-499. doi:https: / / doi.org / 10.1136 / adc.81.6.496 3. Dombi E, Ardern-Holmes SL, Dusica Babovic-Vuksanovic, et al. Recommendations for imaging tumor response in neurofibromatosis clinical trials. Neurology. 2013;81(21_supplement_1). doi:https: / / doi.org / 10.1212 / 01.wnl.0000435744.57038.af 4. Jackler RK, Shapiro MS, Dillon WP, Pitts L, Lanser MJ. Gadolinium-DTPA Enhanced Magnetic Resonance Imaging in Acoustic Neuroma Diagnosis and Management. Otolaryngology–Head and Neck Surgery.1990;102(6):670-677. doi:10.1177 / 019459989010200608 5. Vestibular schwannomas: A Review • APPLIED RADIOLOGY. Appliedradiology.com. Published June 7, 2019. Accessed June 2, 2024. https: / / appliedradiology.com / articles / vestibular- schwannomas-a-review 6. Asthagiri AR, Parry DM, Butman JA, et al. Neurofibromatosis type 2. Lancet. 2009;373(9679):1974-1986. doi:https: / / doi.org / 10.1016 / s0140-6736(09)60259-2 7. Halliday D, Emmanouil B, Pretorius P, et al. Genetic Severity Score predicts clinical phenotype in NF2. Journal of medical genetics.2017;54(10):657-664. doi:https: / / doi.org / 10.1136 / jmedgenet-2017-104519 8. Neurofibromatosis 2. National Organization for Rare Disorders. Published July 31, 2023. Accessed June 2, 2024. https: / / rarediseases.org / rare-diseases / neurofibromatosis-2 / 9. Evans DgR. Neurofibromatosis type 2 (NF2): A clinical and molecular review. Orphanet journal of rare diseases.2009;4(1). doi:https: / / doi.org / 10.1186 / 1750-1172-4-16 10. Harris GJ, Plotkin SR, Maccollin M, et al. Three-dimensional volumetrics for tracking vestibular schwannoma growth in neurofibromatosis type II. Neurosurgery / Neurosurgery online. 2008;62(6):1314-1320. doi:https: / / doi.org / 10.1227 / 01.neu.0000333303.79931.83 45741567.1 21 ATTORNEY DOCKET NO. YU 8712 PCT 11. Morris KA, Parry A, Pretorius PM. Comparing the sensitivity of linear and volumetric MRI measurements to detect changes in the size of vestibular schwannomas in patients with neurofibromatosis type 2 on bevacizumab treatment. The British journal of radiology / British journal of radiology.2016;89(1065):20160110-20160110. doi:https: / / doi.org / 10.1259 / bjr.20160110 12. MacKeith S, Das T, Graves M, et al. A comparison of semi-automated volumetric vs linear measurement of small vestibular schwannomas. European archives of oto-rhino- laryngology / European archives of oto-rhino-laryngology and head & neck.2018;275(4):867- 874. doi:https: / / doi.org / 10.1007 / s00405-018-4865-z 13. Ho HH, Li YH, Lee JC, et al. Vestibular schwannomas: Accuracy of tumor volume estimated by ice cream cone formula using thin-sliced MR images. PloS one. 2018;13(2):e0192411-e0192411. doi:https: / / doi.org / 10.1371 / journal.pone.0192411 14. Kim IK, Starke RM, McRae DA, Nasr NM, Caputy A, Cernica GD, Hong RL, Sherman JH. Cumulative volumetric analysis as a key criterion for the treatment of brain metastases. J Clin Neurosci.2017 May;39:142-146. doi: 10.1016 / j.jocn.2016.12.006. Epub 2017 Jan 11. PMID: 28089195. 15. Eelbode T, Bertels J, Berman M, et al. Optimization for medical image segmentation: theory and practice when evaluating with Dice score or Jaccard Index. IEEE Trans Med Imaging 2020;39:3679–90. 16. Ching-Yung Lin and Shih-Fu Chang "Robust image authentication method surviving JPEG lossy compression", Proc. SPIE 3312, Storage and Retrieval for Image and Video Databases VI, (23 December 1997) 17. Morris KA, Parry A, Pretorius PM. Comparing the sensitivity of linear and volumetric MRI measurements to detect changes in the size of vestibular schwannomas in patients with neurofibromatosis type 2 on bevacizumab treatment. The British Journal of Radiology [Internet]. [cited 2024 Jun 10];89(1065):20160110. 18. Harris GJ, Plotkin SR, MacCollin M, Bhat SP, Urban T, Lev MH, et al. Three- dimensional volumetrics for tracking vestibular schwannoma growth in neurofibromatosis type II. Neurosurgery.2008 Jun 1;62(6):1314–20. 19. Li D, Tsimpas A, Germanwala AV. Analysis of vestibular schwannoma size: A literature review on consistency with measurement techniques. Clinical Neurology and Neurosurgery. 2015 Nov;138:72–7 45741567.1 22 ATTORNEY DOCKET NO. YU 8712 PCT Example 2: Evaluating the Effect of MRI Voxel Size on the Accuracy of 3D Volumetric Analysis Measurements of NF2 Associated Vestibular Schwannomas Neurofibromatosis Type 2 (NF2), a disease caused by a mutation in the NF2 tumor suppressor gene (located on chromosome 22), is an autosomal dominant condition with a poor prognosis. Affected individuals present with meningiomas and vestibular schwannomas (VS), often leading to significant hearing loss and vestibular dysfunction [1,2]. There are limited effective pharmacologic, radiation oncologic and surgical therapies for these tumors. Tumor growth has been traditionally monitored with serial annual surveillance imaging, seeking to identify patterns of tumor growth which threaten injury to adjacent vital structures. Currently, progression of VS is monitored with volumetric analysis allowing volumes to be compared chronologically, as advised by the Response Evaluation in Neurofibromatosis and Schwannomatosis (REiNS) consortium [3]. This process is an estimation employing the volume of an ellipsoid based on the length and width of the tumor at its highest surface area slice(s), as determined by the neuroradiologist. This does not necessarily reflect the actual three- dimensional (3D) shape and size of the tumor. 3D volumetric analysis, which involves creating a 3D model of the tumor based on its boundaries as delineated on the MRI, accounts for changes in tumor growth, including alterations in shape as well as size. This new method has been described in NF2 patients by Evers et al [4]. Such data can be used to extrapolate the potential onset of clinical symptoms due to involvement of adjacent vital structures. The accuracy of any 3D model relies on the image resolution of the source magnetic resonance imaging (MRI) data which is dependent on many factors. The resolution of an MRI is defined by the size of an individual pixel (in two dimensions) or voxel (in 3D). This is related to the field of view, matrix, and slice thickness through the following equation [5]: where the field of is the number of voxels per field of view. An increase in matrix size results in increased resolution as the voxels become smaller with a constant field of view. Other factors to be considered while assessing image quality is the Sound to Noise ratio where a higher ratio would result in more voxels; however, it should be noted that higher signal could also mean that the voxel size would be larger, leading to poor resolution. 45741567.1 23 ATTORNEY DOCKET NO. YU 8712 PCT In this study, we aim to assess a desirable MRI voxel resolution to accurately create 3D models of NF2 patient VS tumors, by using voxel size (in 3 axes) as the variable. The aim is to determine voxel sizes that would lead to 3D model accuracy appropriate for clinical diagnostics. A maximum voxel size has not yet been clearly defined, as there are many different brain MRI protocols with varying voxel sizes, and an unclear consensus regarding which maximum voxel size would lead to measurement inaccuracies. Materials and methods Sampling: Ten patients diagnosed with NF2 were retrospectively recruited at random from our database of NF2 patients who had received annual brain MRIs for vestibular schwannoma (VS) surveillance. Of the 10 patients, the average age was 49.5 years (range 22-65 years). Seven were females and three were males. The NF2 mutation was familial in two cases and mosaic in eight. In total there were 18 tumors examined and included in the study (eight bilateral vestibular schwannomas and two unilateral). Five patients had not previously undergone surgery. The five remaining patients all had one surgery performed; surgeries performed on the other patients included two gamma knife radiosurgeries and three open surgical resections. Imaging: All imaging was conducted at “manuscript excluding author details”. MRI examinations were conducted on a combination of 1.5T Toshiba (1 scan) and 3T Siemens or GE Systems (9 scans) systems. All MRI scans used for volumetric analysis were high resolution thin slice, axial longitudinal relaxation time (T1) postcontrast scans, as seen in Table 2. Table 2. Voxel Sizes of the Original Scans Used 3D Model Creation: 3D tumor masks were created for the vestibular schwannomas (either unilateral or bilateral) of the 10 patients using the image processing software (Simpleware ScanIP, Synopsis, Mountain View, CA). In order to highlight the vestibular schwannomas, a thresholding 45741567.1 24 ATTORNEY DOCKET NO. YU 8712 PCT algorithm tool was used. Isolation of the tumors themselves, as well as the exclusion of non- tumorous soft tissue, was performed using the ‘split regions’ algorithm. In the cases where bilateral VS tumors were present, this tool also allowed these tumors to be separated as independent left and right masks. To assist with mask creation, tumor masks were outlined with a border while the mask was viewed as transparent. Special consideration was taken with the voxels lining the border of the mask and adjustments were made to include missing voxels and exclude extraneous ones. After the initial mask creation, the tumors were re-examined slice by slice in all three planes (coronal, axial, sagittal views) to facilitate correction of any missing or extraneous voxels, Figure 4. All scans were reviewed by three researchers to agree on the voxels to be included as part of the tumor. The original scans for all 18 models were reviewed by a neuroradiologist in order to validate the work of the researchers, and to make adjustments to the 3D masks by examining the voxel selection. The ‘volume’ measurement tool was used to calculate the 3D volume of each tumor mask. This is given in mm3.To evaluate the accuracy of ellipsoid volumes versus 3D volumetric analysis, the researchers calculated the ellipsoid volume of each tumor and compared this to the above 3D volumetric calculation. Resampling: To compare the volumes calculated with the original smallest voxel size to larger voxel sizes, the resampling function was utilized within the image processing software to change the voxel size of the background image using linear interpolations between neighboring voxels. The researchers recreated the masks de novo using the method described above to reduce bias. The desired larger voxel sizes were 0.5x0.5x0.8, 0.8x0.8x0.9, 0.8x0.8x1.6, 1.2x0.9x4.0. These voxel sizes were chosen because they are the most commonly used voxel sizes for 3T internal auditory canal MRIs [6]. Percentage change was calculated between the volume of the tumor mask in the new larger voxel size and the original high-resolution volume. To compare the 3D volumetric calculations with the ellipsoid method, the ellipsoid volume of each tumor was calculated using the equation: volume = 4 / 3 * π * A * B * C, where: A, B, and C are the lengths of all three semi-axes of the ellipsoid. The axes were measured on the slice of the MRI where the tumor appeared the largest. Diagonal lines for the ellipsoid calculation were carried out by the same researcher to minimize intra-individual variability. The percentage change between the ellipsoid and original 3D volumetric was then calculated. 45741567.1 25 ATTORNEY DOCKET NO. YU 8712 PCT Results Table 3. Resampling Analysis, Corresponding Ellipsoid Volume and Percent Change M N ii l V l i R l % % % % % % % % % % % % % % % % % % % % % % % % % 45741567.1 26 ATTORNEY DOCKET NO. YU 8712 PCT 8 0.8 x 0.8 x 0.9 1,778.455 -1.57% 0.4688 x 0.4688 x 1.2 0.8 x 0.8 x 1.6 1,714.946 -5.09% % % % % % % % % % % % % % % % % % % % % % % % % % % % 45741567.1 27 ATTORNEY DOCKET NO. YU 8712 PCT 16 0.5 x 0.5 x 0.8 2,227.791 0.58% 0.4688 x 0.4688 x 0.9 0.8 x 0.8 x 0.9 2,186.641 -1.27% % % % % % % % % % % 45741567.1 28 ATTORNEY DOCKET NO. YU 8712 PCT Table 4. Voxel Volume Change Original Voxel Size (mm) Voxel Size- Resampled (mm) % Voxel Vol. Δ % % % % % % %%% % % % % % % % % % % % % % % % % % % % % % %%% Tumor Volume Percentage Change Volumes of the 3D masks were calculated with each scan’s smallest voxel size and then compared to the 3D masks created from resampled larger voxel sizes. For a voxel size of 0.5x0.5x0.8 millimeter (mm), the researchers found an average percentage change of 1.68% with 45741567.1 29 ATTORNEY DOCKET NO. YU 8712 PCT a range of –0.64% to +13.78%. With a voxel size of 0.8x0.8x.0.9mm, the researchers found an average percentage change of 1.75% with a range of –6.19% to +2.69%. With a voxel size of 0.8x0.8x1.6mm the researchers found an average percentage change of 5.61% with a range of - 42.79% to + 2.95%. With a voxel size 1.2x0.9x4.0mm, the researchers found an average percentage change of 26.87% with a range of –43.37% to -6.43%. This study aims to evaluate the optimum voxel size required for accurate 3D volumetric analysis of VS tumors in NF2 patients. Based on the results of the resampling analysis, the use of voxel sizes up to 0.8x0.8x0.9mm for 3D volumetrics is contemplated. The average change of 1.75% achieved with this voxel size would enable accurate 3D modelling. However, the two larger sizes assessed, 0.8x0.8x1.6mm and 1.2x0.9x4.0mm produced exceptionally large differences in tumor volume and should not be used for accurate 3D volumetric analysis, Figure 5. Due to human error and subjectivity, there is some inherent error in the volume calculations. In addition, Simpleware ScanIP uses partial voxels and smoothing to calculate final volumes; as voxel size increases, the software’s estimations to create smooth models introduces an error into the volume calculations. There is also some heterogeneity in the acquisition of scans due to the incorporation of magnets of different companies and different magnetic fields, however the scans were chosen in order to represent a variety of different voxel sizes to include very fine and very thick voxel thicknesses. The results also demonstrate the inaccuracies of volumetric analysis. As current practice is to use an ellipsoid approximation to calculate tumor volume, the researchers calculated the ellipsoid volume of each tumor and compared these calculations to the 3D volumetric calculation. The results demonstrate that even 3D volumetric calculations that were based on a very large voxel size (1.2x0.9x4.0) were more accurate than the ellipsoid volume produced by the linear calculation. The voxel sizes we evaluated in our resampling analysis represent the most commonly used MRI voxel sizes for brain scans. Based on our results, we recommend using only thin slice scans (0.8x0.8x0.9) in 3D volumetric analysis. In addition, we do not believe that common larger voxel sizes (0.8x0.8x1.6 and 1.2x0.9x4.0) are appropriate for 3D volumetric analysis due to the increased inaccuracy introduced. As the voxel size increases, the tumor volume decreases due to fewer complete voxels being included in the 3D model. Further, due to large discrepancies observed in tumor volumes because of varying voxel sizes, we do not recommend comparing a patient’s consecutive scans that do not have the same voxel size. 45741567.1 30 ATTORNEY DOCKET NO. YU 8712 PCT Although we have only completed mapping and analysis for vestibular schwannomas, we believe this 3D volumetric analysis may be utilized for any solid based tumor model, and we are in the process of systematically evaluating tumors to see its applicability. In conclusion, the current method of linear measurements using ellipsoid volume calculations for measurement of vestibular schwannomas can be inaccurate, particularly with irregularly shaped tumors, since it does not consider the variability in tumor shape in consecutive slices. We believe we have demonstrated that quantitative 3D volumetric analysis on 3D modeled tumors yields a more precise measurement. For this to be carried, the researchers advocate for the use of a voxel size of 0.8x0.8x0.9 or smaller when conducting 3D volumetric analysis for clinical decision making. REFERENCES 1. Asthagiri AR, Parry D, Butman J, et al (2009) Neurofibromatosis type 2. Lancet (London, England), 373(9679), 1974–1986. https: / / doi.org / 10.1016 / S0140-6736(09)60259-2 2. Tamura R. (2021) Current Understanding of Neurofibromatosis Type 1, 2, and Schwannomatosis. International journal of molecular sciences, 22(11), 5850. https: / / doi.org / 10.3390 / ijms22115850 3. Dombi E, Ardern-Holmes SL, Babovic-Vuksanovic D, et al. (2013) Recommendations for imaging tumor response in neurofibromatosis clinical trials. Neurology, 81, 33-40. doi:10.1212 / 01.wnl.0000435744.57038.af 4. Evers S, Verbaan D, Sanchez E, Peerdeman, S. (2015) 3D Volumetric Measurement of Neurofibromatosis Type 2-Associated Meningiomas: Association between Tumor Location and Growth Rate. World Neurosurgery, 84 (4), 1062–1069. https: / / doi.org / 10.1016 / j.wneu.2015.05.068. 5. Bodurka J., Ye N, Petridou K, et al. (2007) Mapping the MRI Voxel Volume in Which Thermal Noise Matches Physiological Noise—Implications for FMRI. NeuroImage, 34 (2), 542– 549.10.1016 / j.neuroimage.2006.09.039. 6. Benson, JC. Carlson ML, Lane JI. (2020) MRI of the Internal Auditory Canal, Labyrinth, and Middle Ear: How We Do It. Radiology, 297(2), 252–265.10.1148 / radiol.2020201767. Those skilled in the art will recognize, or be able to ascertain using no more than routine experimentation, many equivalents to the specific forms of the invention described herein. Such equivalents are intended to be encompassed by the following claims. 45741567.1 31

Claims

ATTORNEY DOCKET NO. YU 8712 PCT We claim:

1. A machine learning algorithm operably linked to a computer processor, wherein the algorithm is configured to process data originating from one or more medical images, wherein processing the data comprises three-dimensional volumetric analysis by the algorithm.

2. A machine learning algorithm, wherein when operably linked to a computer processor, the algorithm is configured to process data originating from one or more medical images, wherein processing the data comprises three-dimensional volumetric analysis by the algorithm.

3. A machine learning algorithm operably linked to a medical imaging device, wherein the algorithm is configured to process data originating from one or more medical images, wherein processing the data comprises three-dimensional volumetric analysis by the algorithm.

4. The machine learning algorithm of any one of claims 1 to 3, wherein processing the data further comprises segmentation of the one or more medical images by the machine learning algorithm.

5. The machine learning algorithm of any one of claims 1 to 4, wherein the one or more medical images are from a subject displaying one or more symptoms of a disease or disorder, such as a disease or disorder associated with acoustic neuroma, e.g., vestibula schwannoma.

6. The machine learning algorithm of any one of claims 1 to 5, wherein the one or more medical images are from a diseased tissue or suspected diseased tissue.

7. The machine learning algorithm of any one of claims 1 to 6, trained using image scans selected from raw image scans, validated image scans, or preferably both, optionally wherein the image scans are obtained from (i) two or more patients and / or (ii) two or more scanning machines preferably from at least two different vendors of scanning machines.

8. The machine learning algorithm of claim 7, wherein at least one of the validated image scans is validated to have a diseased tissue, e.g., tumor.

9. The machine learning algorithm of any one of claims 1 to 8, wherein the three- dimensional volumetric analysis comprises using a voxel size of about 0.8x0.8x0.9 mm or smaller.

10. The machine learning algorithm of any one of claims 1 to 9, configured to perform and / or is capable of performing image registration to overlay generated 3D diseased models onto previous imaging regardless of position of subject when an image used to generate the 3D diseased models were generated. 45741567.132ATTORNEY DOCKET NO. YU 8712 PCT 11. The machine learning algorithm of any one of claims 1 to 10, configured to provide output from data processing to an individual in real-time via a user interface preferably a digital screen (e.g., a screen of a computing device such as a screen of a smartphone, a laptop, a desktop computer, a watch, a tablet, etc.), an electro-mechanical acoustic system (e.g., electronic speakers), or a combination thereof.

12. The machine learning algorithm of any one of claims 1 to 11: (a) configured to process data originating from one or more medical images, wherein processing the data comprises 3D volumetric analysis by the algorithm; (b) configured to perform and / or is capable of performing image registration to overlay generated 3D diseased models onto previous imaging regardless of position of subject when an image used to generate the 3D diseased models were generated, and (c) configured to provide output of its data processing to an individual in real-time.

13. A non‐transitory computer-readable medium with computer executable instructions stored thereon executed by a processor to perform a method of processing one or more medical images, the method comprising: (i) collecting by the medium data originating from one or more medical images, and (ii) performing three-dimensional volumetric analysis preferably by the machine learning algorithm of any one of claims 1 to 12.

14. A computer-implemented method of measuring a three-dimensional volume of a diseased tissue, the method comprising using the machine learning algorithm of any one of claims 1 to 13.

15. A method of diagnosing a subject with a disease comprising measuring a three- dimensional volume of a diseased tissue according to the method of claim 14.

16. The method of claim 15, wherein the diseased tissue is a tumor.

17. The method of claim 16, wherein the tumor is cancerous.

18. The method of any one of claims 15 to 17, further comprising devising a treatment regimen for the subject.

19. The method of any one of claims 14 to 18, wherein output from data processing is provided to an individual in real-time via a user interface preferably a digital screen (e.g., a screen of a computing device such as a screen of a smartphone, a laptop, a desktop computer, a watch, a tablet, etc.), an electro-mechanical acoustic system (e.g., electronic speakers), or a combination thereof.45741567.133

Images