Systems and methods for automated generation of reports on disease progression

Abstract

A system and method for generating a report of disease progression are provided. The system includes a data collection module configured to collect magnetic resonance imaging (MRI) data and clinical data from a patient. The system further includes a multi-stage machine learning model that includes a first stage model, a second stage model, and a third stage model. The first stage model is configured to predict regional-level changes in muscle composition based on the MRI data. The second stage model is configured to predict muscle-level changes in muscle composition based on outputs from the first stage model and the clinical data from the patient. The third stage model is configured to predict functional changes based on outputs from the second stage model. The system further includes an output module configured to generate a report, including a personalized disease progression prediction for the patient based on the multi-stage machine learning model.

Description

Quarles Ref. 175340.00034SYSTEMS AND METHODS FOR AUTOMATED GENERATION OF REPORTS ON DISEASE PROGRESSIONCROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application is based on, claims priority to, and incorporates herein by reference for all purposes, U.S. Provisional Patent Application No. 63 / 756,484 filed on February, 10, 2025.BACKGROUND

[0002] Modern medicine is confronted with a variety of diseases, conditions, and disorders of great complexity, such that it can be difficult for a clinical to make a diagnosis, let alone to track and make care decisions based on a myriad of potential progression paths. In particular, many musculoskeletal and neuromuscular diseases, such as genetic muscle diseases, osteoarthritis, inflammatory conditions, and age- or injury-related muscle decline, follow highly heterogeneous and complex patterns of progression, making it challenging to standardize disease monitoring and treatment assessment. As but one non-limiting example, facioscapulohumeral muscular dystrophy [FSHD] is a hereditary neuromuscular disorder characterized by progressive muscle weakness and wasting. The condition affects approximately 1 in 7,500 individuals and typically manifests with facial weakness, followed by weakness in the shoulders and upper arms. As the disease progresses, it can impact muscles throughout the body, leading to significant functional impairments and reduced quality of life for affected individuals.

[0003] The genetic basis of FSHD involves mutations in the DUX4 gene, resulting in the aberrant expression of the DUX4 protein, which is toxic to muscle cells. While the underlying genetic mechanism is well understood, the clinical presentation and progression of FSHD can be highly variable among patients. This heterogeneity poses challenges for both diagnosis and treatment development.-1- QB\175340.00034\100682361.5Quarles Ref. 175340.00034

[0004] Currently, there are no approved treatments specifically targeting the underlying cause of FSHD. Management primarily focuses on supportive care and symptom management. The development of effective therapies has been hindered by several factors, including the complex nature of the disease progression and the lack of sensitive biomarkers to track disease activity and treatment response.

[0005] Traditional approaches to monitoring FSHD progression often rely on functional assessments, such as manual muscle testing and timed movement tests. While these methods provide valuable information, they may lack sensitivity to detect subtle changes over short periods, particularly in the context of clinical trials evaluating potential therapies.

[0006] The heterogeneous nature of FSHD progression complicates the design and interpretation of clinical trials. Different muscles may be affected at varying rates and to different degrees among patients, making it challenging to select appropriate outcome measures that are relevant across a diverse patient population.

[0007] As research in FSHD continues to advance, there is a growing need for more sophisticated methods to analyze and interpret the complex data generated from imaging studies and clinical assessments. Improved approaches for predicting disease progression and treatment response could enhance the efficiency of clinical trials and ultimately lead to more personalized management strategies for individuals with FSHD and other diseases.

[0008] Furthermore, FSHD is just one of the many complex and progressive diseases presenting daily in clinical medicine. Thus, there is a substantial need to assist clinicians that cannot be experts in all protocols for diagnosing or tracking complex diseases or conditions. Thus, there is an ongoing need for systems and methods to assist or guide clinicians with disease or condition diagnosis and tracking, particularly for -2- QB\175340.00034\100682361.5Quarles Ref. 175340.00034progressive diseases or conditions.SUMMARY OF THE DISCLOSURE

[0009] The present disclosure addresses the aforementioned drawbacks by providing a system and method for characterizing disease progression that is personalized to each patient. The described systems and methods can advantageously be used to predict regional-level changes and muscle-level changes in muscle composition throughout the course of disease progression. Moreover, the systems and methods can be used to predict changes in functional outcomes.

[0010] In some aspects, a system for generating a report of disease progression is provided. The system includes a data collection module that is configured to collect magnetic resonance imaging (MRI) data and clinical data from a patient. The system further includes a multi-stage machine learning model that includes a first stage model, a second stage model, and a third stage model. The first stage model is configured to predict regional-level changes in muscle composition based on at least the MRI data. The second stage model is configured to predict muscle-level changes in muscle composition based on outputs from the first stage model and the clinical data from the patient. The third stage model is configured to predict changes in functional outcomes based on outputs from the second stage model. The system further includes an output module that is configured to generate a report that includes a personalized disease progression prediction for the patient based on the multi-stage machine learning model.

[0011] In other aspects, a computer implemented method is provided for generating a report of disease progression for a patient. The method includes controlling an MRI system to acquire MRI data and providing the MRI data into a first stage machine learning model to predict regional-level changes in muscle composition. The method-3- QB\175340.00034\100682361.5Quarles Ref. 175340.00034further includes providing outputs from the first stage model into a second stage machine learning model to predict muscle-level changes in muscle composition. The method further includes providing outputs from the second stage model into a third stage machine learning model to predict changes in functional outcomes. The method further includes generating a report on personalized disease progression prediction for the patient based on outputs from the third stage model.

[0012] In still other aspects, the present disclosure provides a non-transitory computer-readable medium storing instructions. When executed by a processor, the instructions cause the processor to perform operations for generating reports predicting disease progression. The operations include receiving MRI data and clinical data from a patient and applying a multi-stage machine learning model to the received data. The multi-stage model includes a first stage for predicting regional-level changes in muscle composition, a second stage for predicting muscle-level changes in muscle composition, and a third stage for predicting changes in functional outcomes. The operations further include outputting a personalized disease progression prediction for the patient based on results from the multi-stage machine learning model.

[0013] In other aspects, the present disclosure provides a system for generating a report of disease progression. The system includes a data collection module that is configured to collect MRI data and clinical data from a patient. The system further includes a staged machine learning model that includes one or more model stages. The model stages include at least one of a first stage model, a second stage model, or a third stage model. The first stage model is configured to predict regional-level changes in muscle composition based on at least the MRI data. The second stage model is configured to predict muscle-level changes in muscle composition based on at least the MRI data. The third stage model is configured to predict changes in functional outcomes based on .4.QB\175340.00034\100682361.5Quarles Ref. 175340.00034at least one of the MRI data, outputs from the first stage model, or outputs from the second stage model. The system further includes an output module that is configured to generate a report including a personalized disease progression prediction for the patient based on the staged machine learning model.

[0014] These are but a few, non-limiting examples of aspects of the present disclosures. Other features, aspects and implementation details will be described hereinafter.BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Various objects, features, and advantages of the disclosed subject matter can be more fully appreciated with reference to the following detailed description of the disclosed subject matter when considered in connection with the following drawings, in which like reference numerals identify like elements.

[0016] FIG. 1 shows a flowchart setting forth steps of an example process for training a machine learning algorithm to predict disease progression.

[0017] FIG. 2 shows a flowchart setting forth steps of an example process for applying a machine learning algorithm to predict disease progression.

[0018] FIG. 3A schematically illustrates an example multi-stage machine learning algorithm that can be used to predict disease progression.

[0019] FIG. 3B schematically illustrates an example system that can be used to predict disease progression.

[0020] FIG. 4 illustrates an example multi-scale machine learning model that can be used to predict disease progression according to the present disclosure.

[0021] FIG. 5 shows an example of 3D segmentations of muscles in patients with FSHD of varying disease severity.-5- QB\175340.00034\100682361.5Quarles Ref. 175340.00034

[0022] FIG. 6 provides example experimental data that characterizes muscle involvement at baseline in patients with FSHD.

[0023] FIG.7A shows experimental results from an example implementation of the first stage model used to predict regional-level fat fraction changes according to the present disclosure.

[0024] FIG. 7B shows additional experimental results from an example implementation of the first stage model used to predict regional-level fat fraction changes according to the present disclosure.

[0025] FIG. 7C shows additional experimental results from an example implementation of the first stage model used to predict regional-level fat fraction changes according to the present disclosure.

[0026] FIG. 7D shows additional experimental results from an example implementation of the first stage model used to predict regional-level fat fraction changes according to the present disclosure.

[0027] FIG. 7E shows additional experimental results from an example implementation of the first stage model used to predict regional-level fat fraction changes according to the present disclosure.

[0028] FIG. 7F shows additional experimental results from an example implementation of the first stage model used to predict regional-level fat fraction changes according to the present disclosure.

[0029] FIG. 7G shows additional experimental results from an example implementation of the first stage model used to predict regional-level fat fraction changes according to the present disclosure.

[0030] FIG.8A shows experimental results from an example implementation of the second stage model used to predict muscle-level fat fraction changes according to the -6- QB\175340.00034\100682361.5Quarles Ref. 175340.00034present disclosure.

[0031] FIG. 8B shows additional experimental results from an example implementation of the second stage model used to predict muscle-level fat fraction changes according to the present disclosure.

[0032] FIG. 8C shows additional experimental results from an example implementation of the second stage model used to predict muscle-level fat fraction changes according to the present disclosure.

[0033] FIG. 8D shows additional experimental results from an example implementation of the second stage model used to predict muscle-level fat fraction changes according to the present disclosure.

[0034] FIG. 8E shows additional experimental results from an example implementation of the second stage model used to predict muscle-level fat fraction changes according to the present disclosure.

[0035] FIG. 8F shows additional experimental results from an example implementation of the second stage model used to predict muscle-level fat fraction changes according to the present disclosure.

[0036] FIG. 8G shows additional experimental results from an example implementation of the second stage model used to predict muscle-level fat fraction changes according to the present disclosure.

[0037] FIG.9A shows experimental results from an example implementation of the second stage model used to predict lean muscle volume changes according to the present disclosure.

[0038] FIG. 9B shows additional experimental results from an example implementation of the second stage model used to predict lean muscle volume changes according to the present disclosure.-7- QB\175340.00034\100682361.5Quarles Ref. 175340.00034

[0039] FIG. 9C shows additional experimental results from an example implementation of the second stage model used to predict lean muscle volume changes according to the present disclosure.

[0040] FIG. 9D shows additional experimental results from an example implementation of the second stage model used to predict lean muscle volume changes according to the present disclosure.

[0041] FIG. 9E shows additional experimental results from an example implementation of the second stage model used to predict lean muscle volume changes according to the present disclosure.

[0042] FIG. 9F shows additional experimental results from an example implementation of the second stage model used to predict lean muscle volume changes according to the present disclosure.

[0043] FIG. 9G shows additional experimental results from an example implementation of the second stage model used to predict lean muscle volume changes according to the present disclosure.

[0044] FIG. 10A shows experimental results evaluating the performance of the second stage model used to predict fact fraction changes.

[0045] FIG. 10B shows experimental results evaluating the performance of the second stage model used to predict lean muscle volume changes.

[0046] FIG. E8A shows experimental results from an example implementation of the third stage model used to predict functional changes.

[0047] FIG. E8B shows additional experimental results from an example implementation of the third stage model used to predict functional changes.

[0048] FIG. 11 is a block diagram of an example magnetic resonance imaging ("MRI”) system that can implement the methods described in the present disclosure.-8- QB\175340.00034\100682361.5Quarles Ref. 175340.00034

[0049] FIG. 12 is a block diagram of an example disease progression prediction system that can implement the methods of the present disclosure.

[0050] FIG. 13 is a block diagram of example components that can implement the system of FIG. 12.DETAILED DESCRIPTION

[0051] Before any aspects of the present disclosure are explained in detail, it is to be understood that the invention is not limited in its application to the details of construction and the arrangement of components set forth in the following description or illustrated in the following drawings. The invention is capable of other embodiments and of being practiced or of being carried out in various ways. Also, it is to be understood that the phraseology and terminology used herein is for the purpose of description and should not be regarded as limiting. The use of "including,” "comprising,” or "having” and variations thereof herein is meant to encompass the items listed thereafter and equivalents thereof as well as additional items. Unless specified or limited otherwise, the terms "mounted,” "connected,” "supported,” and "coupled” and variations thereof are used broadly and encompass both direct and indirect mountings, connections, supports, and couplings. Further, "connected” and "coupled” are not restricted to physical or mechanical connections or couplings.

[0052] The following discussion is presented to enable a person skilled in the art to make and use embodiments of the invention. Various modifications to the illustrated embodiments will be readily apparent to those skilled in the art, and the generic principles herein can be applied to other embodiments and applications without departing from embodiments of the invention. Thus, embodiments of the invention are not intended to be limited to embodiments shown but are to be accorded the widest scope-9- QB\175340.00034\100682361.5Quarles Ref. 175340.00034consistent with the principles and features disclosed herein. The following detailed description is to be read with reference to the figures, in which like elements in different figures have like reference numerals. The figures, which are not necessarily to scale, depict selected embodiments and are not intended to limit the scope of embodiments of the invention. Skilled artisans will recognize the examples provided herein have many useful alternatives and fall within the scope of embodiments of the invention.

[0053] The clinical management and treatment development of many musculoskeletal and neuromuscular diseases and conditions face challenges due to the heterogeneity of the disease progression, limited biomarkers and standardized characterization methods, and difficulty tracking progression. For example, genetic muscle diseases (e.g., Limb-girdle muscular dystrophy, Duchenne muscular dystrophy, congenital myopathies, facioscapulohumeral muscular dystrophy), osteoarthritis, rheumatoid arthritis and other inflammatory conditions, and even age- or injury-related muscle decline (e.g., sarcopenia, osteoporosis, spinal deformity, frailty, chronic low back pain, and so forth) often present with a high degree of heterogeneity across patients or over the course of progression for a given patient. Even more, these conditions are often complex, as onset and progression is influenced by genetics, lifestyle, aging, diet, inflammation, and environmental factors. This variability makes it difficult to establish standard biomarkers for diagnosis, tracking disease progression, and assessing treatment efficacy.

[0054] As one non-limiting example, facioscapulohumeral muscular dystrophy (FSHD) is a genetic neuromuscular disorder characterized by progressive muscle degeneration, affecting approximately 1 in 7,500 individuals. The genetic basis for FSHD — mutations to the DUX4 gene leading to aberrant expression of the protein DUX4, which is toxic to muscles — is well established. Disease progression typically occurs over -10- QB\175340.00034\100682361.5Quarles Ref. 175340.00034years in adult-onset FSHD and can affect skeletal muscles across the entire body. However, there are no approved treatments for FSHD, and no effective approaches exist to mitigate the progressive weakness and associated loss of key functions such as ambulation, reaching, and swallowing. Moreover, FSHD is characterized by progressive muscle degeneration with substantial variability in severity and progression patterns. Current biomarkers for tracking disease progression lack sensitivity for personalized assessment, posing challenges in therapeutic development.

[0055] Current clinical metrics for tracking FSHD progression primarily rely on functional assessments, including manual muscle testing, timed movement tests (e.g., timed up-and-go (TUG)), computer- vision-analyzed movements (e.g., reachable workspace), and patient-reported outcomes. Among these, reachable workspace (RWS) has been used most recently in FSHD clinical trials. However, recent top-line results from the Fulcrum Therapeutics Phase HI trial (NCT05397470) revealed that both placebo and treatment groups demonstrated improvements in RWS performance; as a result, the measure failed to detect meaningful impacts of treatment. The causes for placebo improvement remain unclear but could include sensitivity to motor learning effects or placebo responses rather than true therapeutic efficacy, especially during short time frames. Due to the heterogeneous progression of FSHD, the functional task that would be the most reflective of progression likely differs across patients. For example, only a subset of participants in a clinical trial would exhibit measurable decline in shoulder girdle muscles involved in the reachable workspace task over a short trial duration, particularly in the absence of treatment. Other functional metrics, like TUG or 6-minute-walktest, also have a similar limitation.

[0056] In order to address the limitations of global clinical metrics, magnetic resonance imaging (MRI) has emerged as the gold standard for quantifying muscle-level -11- QB\175340.00034\100682361.5Quarles Ref. 175340.00034involvement in FSHD. MRI enables a detailed assessment of muscle composition, particularly the amount of fat infiltration, which serves as a reliable metric of disease expression. Despite the relative ease of acquiring whole-body imaging data, current analytic methods for segmenting and quantifying features at the individual muscle level remain limited. Consequently, much of the existing literature in adults and children has relied on qualitative ratings or quantification confined to single slices or central muscle sub-regions. Recent advances have introduced semi-automated and fully automated methods to quantify select individual muscles or muscle groups, leading to updated conclusions about muscle involvement and fat progression patterns. Key revisions include: (1) while FSHD was historically thought to primarily affect specific muscles (e.g., scapular fixation muscles, hamstrings, tibialis anterior, and gastrocnemius), it is now evident that all muscles can be affected across the lifespan; (2) the pattern of muscle involvement and progression is highly variable across patients and muscle groups; and (3) within individual muscles, disease progression is heterogeneous, with varying degrees of fat infiltration and diverse patterns of structural degradation.

[0057] As trials have progressed, such as the recently terminated trial (NCT05397470) and ongoing studies (NCT05747924 and NCT05548556), the common approach has been to monitor a subset of individual muscles or muscle groups. This involves grouping muscles with a prespecified range of fat infiltration, deemed "at risk," into a single category and using this aggregated number as a treatment response metric. While this streamlined strategy aligns with the single-biomarker approach used in DMD, it has several notable limitations: (1) it omits muscles above and below the specified threshold that may also be changing; (2) it disregards baseline muscle status and identity, which are known to influence change rates; (3) it overlooks fat distribution patterns, with confluence being a characteristic feature of the disease; and (4) it regresses data toward -12- QB\175340.00034\100682361.5Quarles Ref. 175340.00034the mean. These cumulative limitations reduce the sensitivity of MRI to detect subtle changes and impair its ability to accurately map muscle features onto task performance.

[0058] While a whole-body MRI analysis provides a more in-depth, personalized description of FSHD disease state, the vast amount of data generated also introduce significant analytical challenges. Furthermore, muscle MRI measurements do not provide the full picture of an individual’s disease state: other data types, such as clinical assessments and disease descriptors (both descriptive modulators like D4Z4 length, and functional task performance), can also be incorporated to provide an integrated patient assessment tool. These challenges point to the need for updated methods to interpret and analyze the data effectively. Machine learning techniques, particularly ensemble methods, have shown considerable promise in addressing the complexities of heterogeneous disease in order to predict patient-specific disease progression.

[0059] To address the aforementioned challenges, the present disclosure provides systems and methods for a machine learning model that incorporates detailed MRI measures along with clinical data to predict progression of muscular diseases and conditions, such as FSHD. Such systems and methods can advantageously be used as a diagnostic tools, for monitoring disease progression, and to advance disease understanding in research for a wide range of complex diseases and conditions (e.g., musculoskeletal conditions, neuromuscular diseases, genetic muscle diseases, osteoarthritis, rheumatoid arthritis and other inflammatory conditions, age- and injury-related muscle decline, and so forth).

[0060] For example, the described systems and methods support predictions of outcome and disease progression both clinically (e.g., for an individual patient) and in research (for a cohort in clinical trials). Even more, the systems and methods can facilitate a so-called ‘patient digital twin,’ which acts as a model that simulates the natural course -13- QB\175340.00034\100682361.5Quarles Ref. 175340.00034of disease progression in untreated patients. By integrating these methodologies, the described model enhances trial efficiency and improves the evaluation of treatment efficacy in FSHD and other clinical research. This digital twin can serve as a benchmark in clinical trials, enabling a comparison of the effects of novel therapies against a personalized model-predicted baseline of untreated progression. The described model has the potential to enhance the efficiency of FSHD and other clinical trials by enabling more informed power analyses and eliminating the need for randomly assigned placebo arms.

[0061] Referring now to FIG. 1, a flowchart is illustrated as setting forth the steps of an example method for predicting disease progression using suitably trained machine learning algorithm. The described method enables the prediction of which muscles, and to what extent, will undergo changes over time during the natural course of a disease (e.g., FSHD), offering a more precise and individualized approach to monitoring disease progression (e.g., in clinical trials).

[0062] The method can be implemented using a computer or suitable computer system to generate a report of disease progression for a patient, which may include predictions of regional-level and muscle-level changes in muscle composition and functional outcome changes. As a non-limiting example, the patient may be experiencing or being assessed for a disease that affects muscle composition, which may be referred to as a muscular, musculoskeletal, or neuromuscular disease or condition. For example, the patient may be a patient diagnosed with FSHD.

[0063] As will be described further with respect to FIG. 3, the trained machine learning algorithm may include a staged machine learning algorithm that employs one or more trained machine learning algorithms used in series or stages. In some implementations, the input for each algorithm can include or be based on output disease -14- QB\175340.00034\100682361.5Quarles Ref. 175340.00034progression data from earlier-stage algorithms. In this way, the trained machine learning algorithm provides a pipeline for predicting disease progression at various levels of analysis (e.g., regional-level, muscle-level, functional level). As non-limiting examples, the machine learning algorithm may include a series of ensemble learning algorithms (e.g., random forest-based models, gradient boosting models), neural network-based algorithms (e.g., convolutional neural networks, recurrent neural networks, and so forth), support vector machine models, other machine learning algorithms, or a combination thereof.

[0064] Referring again to FIG. 1, the machine learning algorithm takes MRI-based data as input data and generates disease progression data as output data at one or more analysis levels (e.g., regional-level, muscle-level, functional level). As an example, the disease progression data can be indicative of a prediction of a regional-level change in muscle composition, a prediction of a muscle-level change in muscle composition, a prediction in a functional outcome associated with muscle composition, or a combination thereof.

[0065] The method includes accessing MRI and clinical data with a computer system, as indicated at step 102. In some implementations, step 102 can be performed by a data collection module that collects the MRI and clinical data from a patient or subject. Accessing the MRI and clinical data may include retrieving such data from a memory or other suitable data storage device, such as an electronic health records system. Additionally or alternatively, accessing the MRI and clinical data may include acquiring such data with an MRI system and transferring or otherwise communicating the data to the computer system, which may be a part of the MRI system.

[0066] In some implementations, the MRI data include raw MRI data that can be processed to generate MRI images with a desired contrast (e.g., water / fat separation). In -15- QB\175340.00034\100682361.5Quarles Ref. 175340.00034other implementations, the MRI data include MRI images with a desired contrast (e.g., water / fat separation, water only images, fat only images, in-phase images, out-of-phase images, Dixon images, modified Dixon images, short tau inversion recovery (STIR) images, T2-weighted images, T2 maps, water T2 maps, and so forth). In some implementations, the MRI images include whole-body images or a series of images that can be stitched together to form whole-body images. In this way, the described methods can be used to fully characterize or predict disease progression across a patient’s full musculoskeletal system. In other implementations, the MRI images may depict muscle groups or body regions of interest (e.g., upper body for a patient experiencing degeneration in their arms).

[0067] The MRI images can be processed to generate MRI-based parameters that characterize musculoskeletal distribution metrics based on the MRI images. For example, the images can be segmented (e.g., automatically, manually, or automatically with manual correction or confirmation) to delineate bones and muscles of interest. In some implementations, the images are segmented using an artificial intelligence (Al) -based algorithm.

[0068] The segmented images can be analyzed to provide MRI-based musculoskeletal metrics or parameters that characterize aspects (e.g., muscle and fat distribution) of the patient’s muscles or bones based on the MRI data. Generating the musculoskeletal metrics may include identifying muscle and fat content within the muscles or bones.

[0069] The musculoskeletal metrics can be quantified for each muscle or for each cross-sectional slice throughout the length of each muscle (e.g., along the superiorinferior length of the muscle). Because fat and fat changes are often regionally localized, slice-level metrics may advantageously provide greater sensitivity than muscle-wide -16- QB\175340.00034\100682361.5Quarles Ref. 175340.00034metrics. As a non-limiting example, the MRI-based musculoskeletal metrics may include the muscle boundary cross-sectional area (CSA), lean muscle CSA, area fat fraction (as a percentage), fat variation, fat fraction kurtosis, fat fraction skewness, and fat fraction Moran’s index. In some implementations, such metrics can be generated based on Dixonbased MRI images that characterize and separate water and fat signals in the tissue. As another non-limiting example, the MRI-based musculoskeletal metrics may include STIR-based metrics (e.g., measure of inflammation or edema) or T2 metrics (e.g., water T2 values). The musculoskeletal metrics can also include bone metrics, such as bone (e.g., femur) volume, cross-sectional area, fat fraction, or a combination thereof.

[0070] In some implementations, process block 102 also includes accessing clinical or patient data. Such data can be retrieved from a health records system or input by a user via a user interface. The clinical data may include patient demographic data, such as age, sex, weight, height, and so forth. The clinical data may include biological data, such as biopsy results, measurement of a blood biomarker, or a characterization of a genetic factors. For example, the clinical data may include D4Z4 repeat length. The clinical data may also include relevant data that characterizes a patient’s disease or condition, such as presence or severity of disease, age of disease onset, family history, and so forth. The clinical data may also include baseline functional performance data that characterize the patient’s functional abilities, such as baseline or historical timed up-and-go (TUG) measurements.

[0071] A trained machine learning algorithm is then accessed with the computer system, as indicated at step 104. As described, the trained machine learning algorithm may include a staged algorithm that includes one or more trained machine learning models (e.g., random forest models, other ensemble learning models, neural networks, and so forth). Accessing the trained algorithm may include accessing model parameters -17- QB\175340.00034\100682361.5Quarles Ref. 175340.00034that have been optimized or otherwise estimated by training the model on training data for each stage. As non-limiting example, when using a random forest model, model parameters may include split features and thresholds, tree structures, and leaf-node target means or medians; when using a neural network, model parameters may include weights, biases, or both. Accessing the trained algorithm may also include accessing model hyperparameters for each stage, such as number of trees, maximum depth, minimum samples per split, and so forth. In general, retrieving the model can also include retrieving, constructing, or otherwise accessing the particular model architecture to be implemented. For instance, data pertaining to the layers in a neural network architecture (e.g., number of layers, type of layers, ordering of layers, connections between layers, hyperparameters for layers) may be retrieved, selected, constructed, or otherwise accessed.

[0072] In general, the staged model is trained, or has been trained, on training data in order to predict disease progression for one or more analysis levels. In implementations in which the staged model includes one or more random forest models, each random forest model generally includes an ensemble of decision trees that operate in parallel. Each decision tree has internal decision nodes or rules that are defined by selected split features and corresponding thresholds, a hierarchical tree topology connecting those nodes, and leaf nodes storing output values based on training data. The trees operate in parallel on input data to produce individual predictions, which are aggregated (e.g., by majority voting or averaging) to yield a final model output. During training, each tree is constructed using bootstrap-sampled data and feature-subsampled split evaluations, producing diversified tree structures that collectively form the ensemble, preventing overfitting of the data.

[0073] In implementations in which the staged model includes one or more neural -18- QB\175340.00034\100682361.5Quarles Ref. 175340.00034networks, the artificial neural network generally includes an input layer, one or more hidden layers (or nodes), and an output layer. The number (and the type) of inputs provided to the artificial neural network may vary based on the particular task for the artificial neural network. Each connection between nodes may be assigned a weight parameter, bias value, or both. The nodes within the hidden layers are generally associated with an activation function that defines how the hidden layer processes the input received from a previous layer. The hidden layers may perform different functions, such as convolutional hidden layers that reduce the dimensionality of inputs; max pooling layers that reduce a group of inputs to the maximum value; averaging layers; batch normalization; and so forth. The last hidden layer in the artificial neural network is connected to the output layer in which each node provides a prediction value (e.g., future lean muscle volume, future fat fraction, future functional level, and so forth) for a particular analysis level.

[0074] Each model is trained to generate a disease progression prediction for a specified prediction level or task. For example, the model output can include disease progression predictions at a regional level, muscle level, or functional level. As nonlimiting examples, the predictions may include predictions of changes in region-level or muscle-level fat fraction, a region-level or muscle-level lean muscle volume, or a functional metric (e.g., TUG measurement). In this way, the multi-stage model provides a more wholistic and personalized prediction of the patient’s disease progression.

[0075] The MRI and clinical data are then input to the trained staged machine learning model, generating output as disease progression prediction, as indicated at step 106. For example, the disease progression prediction may include predictions of changes in region-level or muscle-level fat fraction, a region-level or muscle-level lean muscle volume, a functional metric (e.g., TUG measurement), or a combination thereof. The -19- QB\175340.00034\100682361.5Quarles Ref. 175340.00034quantified disease progression prediction can provide physicians or other clinicians with information to guide treatment (e.g., physical therapy, muscle focus areas) and provide metrics to evaluate new treatments in research or clinical trials. For example, the predictions can be used as a baseline measure representing the patient in an untreated state and compared to the patient’s realized progression with treatment to evaluate treatment efficacy.

[0076] The disease progression prediction data generated by inputting the MRI and clinical data to the trained model can then be displayed to a user, stored for later use or further processing, or both, as indicated at step 108. For example, an output module can be used to generate a report that includes a personalized disease progression for the patient based on the output of the staged machine learning model. This disease progression report may include a prediction of regional-level changes, a prediction of muscle-level changes, a prediction of functional-level changes, or a combination thereof.

[0077] Referring now to FIG. 2, a flowchart is illustrated as setting forth the steps of an example method for training a staged machine learning model on training data. The model is trained to receive MRI and clinical data as input data in order to generate disease progression prediction data as output data, where the disease progression prediction data are indicative of a prediction of a future progression in a muscular disease or condition.

[0078] In general, the trained model can implement a series of one or more machine learning models (e.g., ensemble learning models, neural networks, and so forth) in stages to predict disease progression at various physiological or functional levels. For instance, the model can implement three random forest models or the like.

[0079] The method includes accessing training data with a computer system, as indicated at step 202. Accessing the training data may include retrieving such data from -20- QB\175340.00034\100682361.5Quarles Ref. 175340.00034a memory or other suitable data storage device or medium (e.g., electronic health records system). Alternatively, accessing the training data may include acquiring such data with an MRI system and transferring or otherwise communicating the data to the computer system.

[0080] In general, the training data can include MRI data across various time points for a group of patients or subjects. As previously described, the MRI data may include raw MRI data that can be processed to produce MRI images, MRI images that can be processed (e.g., segmented) to generate musculoskeletal metric data, or musculoskeletal metric data generated based on MRI images. Additionally, the training data may include clinical data, as previously described. In some embodiments, the training data may include MRI and clinical that have been labeled based on future disease progression metrics (e.g., changes in fat fraction, lean muscle volume, or TUG). In this way, data can be paired through time for each patient or subject within the training data set such that future predictions can be made based on baseline measurements.

[0081] The staged model is trained on the training data at each stage, as indicated at step 204. In implementations that include one or more random forest models, a model stage can be trained by training an ensemble of decision trees over multiple bootstrap-resampled subsets of labeled training data and aggregating the predictions across the subsets. For each tree, the feature space is recursively partitioned by selecting, at each node, a randomly chosen subset of candidate features (e.g., region fat fraction, fat fraction kurtosis, and so forth) and identifying a split or threshold that minimizes a node impurity measure (e.g., variance) for each feature. Trees can be grown according to predefined stopping criteria.

[0082] In implementations that include one or more neural networks, a model stage can be trained by optimizing network parameters (e.g., weights, biases, or both)-21- QB\175340.00034\100682361.5Quarles Ref. 175340.00034based on minimizing a loss function. As one non-limiting example, the loss function may be a mean squared error loss function.

[0083] The staged model is then stored for later use, as indicated at step 206. Storing the model may include storing model architecture (e.g., tree structure, layer structure) and model parameters (e.g., split features and thresholds, weights, biases, and so forth), which have been determined by training the model on the training data.

[0084] FIG. 3A schematically illustrates an example staged machine learning model that can be used to generate disease progression predictions. As previously described, the staged machine learning model 300 is trained to predict disease progression based on a combination of MRI data and clinical data. For example, the MRI data may include MRI images that can be segmented to generate musculoskeletal metrics based on the MRI data (e.g., fat fraction, muscle volume, bone cross-sectional area, femur cross-sectional area, and so forth). In this way, the machine learning algorithm can receive images as input data and generate musculoskeletal metrics as a step within the algorithm or can directly receive musculoskeletal metrics as input data.

[0085] The staged model 300 may include a series of one or more machine learning models, such as ensemble learning based models (e.g., random forest models). In some implementations, the staged model 300 may include multiple (e.g., two, three, or more) stages in which each stage predicts disease progression at a particular analysis level (e.g., regional-level, muscle-level, functional level). Such staged model 300 may be referred to as a multi-stage model. In other implementations, the staged model 300 may include a single stage trained to predict disease progression at an analysis level of interest. In other implementations, the staged model 300 may include multiple stages and be applied using a single stage to predict disease progression at a desired analysis level.

[0086] As a non-limiting example, each stage of a multi-stage model can be trained -22- QB\175340.00034\100682361.5Quarles Ref. 175340.00034to generate disease progression predictions at three analysis levels, including regional-level, muscle-level, and functional-level. For each stage, the input data may include the output of the previous stage, a subset of the output of the previous stage, or an aggregated output of the previous stage, providing multi-level analysis of disease progression. For example, the output of stage 1 may be aggregated (e.g., averaged) across the muscle to provide an input for stage 2.

[0087] The multi-stage machine learning model 300 includes a first stage model 302 that is configured to predict regional-level disease progression 304 as an output based on the MRI data or the MRI data and the clinical data 350. For example, the predicted regional-level disease progression 304 may include a prediction of a change in muscle composition (e.g., fat fraction, lean muscle cross-sectional area or volume, and so forth) in several pre-defined regions within each muscle. In some implementations, the regions can be defined as cross-sectional slices along the length of the muscle.

[0088] The multi-stage machine learning model 300 also includes a second stage model 306 that is configured to predict muscle-level disease progression 308 as an output based on the output 304 of the first stage model 302 and the clinical data 350. For example, the predicted muscle-level disease progression 308 may include a prediction of a change in muscle composition (e.g., fat fraction, lean muscle volume, and so forth) for one or more individual muscles.

[0089] As a non-limiting example, the second stage model 306 uses clinical data 350 that includes D4Z4 repeat length as input data. In some implementations, the second stage model 306 may also include the MRI data 350 as an input. For example, the second stage model 306 may also include femur cross-sectional area measurements based on the MRI data 350 as input data.

[0090] In some implementations, the second stage model 306 may include two or -23- QB\175340.00034\100682361.5Quarles Ref. 175340.00034more parallel random forest models. For example, a first trained random forest model can be used to predict regional-level changes in fat fraction, and a second trained random forest model can be used to predict regional-level changes in lean muscle cross-sectional area.

[0091] The multi-stage machine learning model 300 also includes a third stage model 310 that is configured to predict functional-level disease progression 312 as an output based on the output 308 of the second stage model 306. For example, the predicted functional-level disease progression 312 may include a prediction of a change in a function measurement for each patient, such as TUG test performance (e.g., TUG total time, gait speed, number of steps), anatomical-based severity scores, completion time of functional test (e.g., 10-meter walk, sit-to-stand), and so forth).

[0092] In some implementations, input to the third stage model 310 may include predicted muscle-level progression 308 for a subset of muscles of interest. For example, the input may include predicted changes in lean muscle volume for a functional muscle groups (e.g., trunk extensors, hip extensors, and knee extensors). In some implementations, the third stage model 310 may also use MRI data, clinical data, or both 350 as input data. For example, input to the third stage model 310 may include baseline TUG performance or another baseline functional performance measure.

[0093] As another non-limiting example, the model may include a single stage trained to generate disease progression predictions at one analysis level. In this case, the input data to a model stage (e.g., 302, 306, or 310) may include MRI data, clinical data, or a combination thereof, rather than relying on inputs generated by a previous stage. In this way, the model 300 may omit one or more of the model stages 302, 306, and 310 and corresponding outputs 304, 308, and 312. Thus, the staged machine learning model 300 may include one or more model stages that include at least one of a first stage model 302-24- QB\175340.00034\100682361.5Quarles Ref. 175340.00034that is configured to predict regional-level changes in muscle composition 304 based on at least the MRI data 350; a second stage model 306 that is configured to predict musclelevel changes in muscle composition 308 based on at least the MRI data 350; or a third stage model 310 configured to predict changes in functional outcomes 312 based on at least one of the MRI data 350, outputs from the first stage model, or outputs from the second stage model.

[0094] FIG. 3B provides a block diagram illustrating a system 360 that can be used to provide a personalized disease progression prediction for a patient. The system 360 includes a data collection system 362 that collects MRI data and clinical data from the patient. In some implementations, the data collection system 362 may be configured to access the MRI data and clinical data from an MRI system 366, a memory (e.g., electronic health record system), or a combination thereof. The MRI data accessed may include MRI images (e.g., Dixon images, whole-body Dixon images), segmented MRI images, or musculoskeletal metrics generated based on segmented MRI images. In some configurations, the data collection system 362 may be configured to control the MRI system to acquire the MRI data from the patient. In some implementations, the data collection system 362 may be configured to segment the MRI images to delineate individual muscles. For example, the data collection system 362 can use an artificial intelligence (Al) -based segmentation algorithm to segment the muscles of interest. In some implementations, this segmentation may be confirmed or corrected manually by a radiologist or other trained personnel. The data collection system 362 may also generate musculoskeletal metrics (e.g., fat fraction, muscle volume, and so forth) based on the segmented images.

[0095] The system 360 also includes a staged machine learning model 300 that can predict progression of disease (e.g., FSHD or other muscular condition) on one or -25- QB\175340.00034\100682361.5Quarles Ref. 175340.00034more scales, as previously described.

[0096] The system 360 also includes an output module 364 that generates a report of the personalized disease progression prediction for each patient based on the output of the staged machine learning model 300.

[0097] Examples

[0098] Example 1

[0099] The present example provides a non-limiting example implementation of a multi-scale machine learning framework. In the example study described below, the machine learning algorithm leveraged whole-body magnetic resonance imaging (MRI) and clinical data to predict regional, muscle, joint, and functional progression in FSHD. Using a combined dataset of over 100 patients from seven studies, MRI-derived metrics, including fat fraction, lean muscle volume, and fat spatial heterogeneity at baseline, were integrated with demographic and functional measures. A three-stage random forest model was developed to predict annualized changes in muscle composition and a functional outcome (timed up-and-go (TUG)).

[0100] All model stages revealed strong predictive performance in separate holdout datasets. Stage 2 models predicted fat fraction change with a root mean square error (RMSE) of 2.16% and lean volume change with a RMSE of 8.1ml. The stage 3 model predicted changes in TUG with a RMSE of 0.6 seconds. This study demonstrates the random forest models incorporating individual muscle data can effectively predict MRI disease progression and functional performance of complex tasks, addressing the heterogeneity and nonlinearity inherent in FSHD. Given linked structure-function data represents an important pairing for trial success, these models can provide an improved approach to characterize these relationships. As many neuromuscular diseases are characterized by muscle and pattern heterogeneity, like FSHD, such approaches may be -26- QB\175340.00034\100682361.5Quarles Ref. 175340.00034broadly generalizable to several neuromuscular disorders.

[0101] Methods

[0102] We leveraged multiple FSHD retrospective datasets, combined into a so-called "data lake.” The datasets included MRI scans collected from >100 patients with FSHD from seven different studies: Wellstone cohort (n = 34, P50 AR065139, J. Chamberlain PI, S. Tapscott co-PI); Kennedy Krieger Institute (KKI) cohort (n = 30) (1K23NS091379, D. Leung PI), a subset of the Fulcrum Phase II DUX4 placebo cohort (n = 20, NCT05397470), FSHD Global Research Foundation Registry cohort (n = 28), and MOVE Plus patients (n = 24) (Statland PI, funded by Avidity Biosciences, FSHD Canada). All subjects were adults and provided informed consent. All experimental protocols used to collect the data were in accordance with relevant guidelines / regulations and were approved by institutional review boards (IRB) or ethics committees. Subjects provided informed consent for data collection and aggregation. The scans available varied in coverage and acquisition method, as summarized in Table 1.

[0103] Table 1: Data set information-27- QB\175340.00034\100682361.5Quarles Ref. 175340.00034

[0104] An Al-based method was used to segment the boundaries of up to 118 muscles (depending on scan coverage) across the whole body from the Dixon water images. Trained segmentation engineers manually reviewed and edited for accuracy (referred to as "vetting”) utilizing the manual editing tools (e.g., paint, draw, erase) in 3D Slicer. This process required between one and three hours, depending on the level of fatty replacement (with higher fatty replacement muscles requiring more user interaction time for contour verification).

[0105] Since disease progression is often heterogeneous within each muscle in FSHD, the Stage 1 model focused on slice-level, regionally sensitive level metrics, which may be referred to as musculoskeletal metrics. First, muscle and fat quantities were computed within the cross-section of each muscle in each axial slice. Those metrics were represented as a function of superior-inferior position along the muscle. Slice-level measurements are particularly sensitive to localized changes in fat distribution, which often occur within discrete anatomical zones. To summarize regional variation along the length of each muscle, the muscle was divided into eight longitudinal intervals along the -28- QB\175340.00034\100682361.5Quarles Ref. 175340.00034superior-inferior direction spanning 10% increments, including 10-20%, 20-30%, 30-40%, 40-50%, 50-60%, 60-70%, 70- 80%, and 80-90%. Within each of these regions, average slice-level metrics were calculated, such as cross-sectional area (CSA), lean muscle CSA, fat fraction (FF), and spatial heterogeneity metrics, including fat variation, kurtosis, skewness, and Moran’s index. This approach enabled assessment of how fat infiltration and structural changes varied along the muscle’s length in a standardized and interpretable way. Fat variation (in units of %) for each region was calculated as the standard deviation of fat fraction across all pixels in the boundary volume. Kurtosis and skewness (expressed as percentages) for each region were calculated based on the distribution of pixel fat fraction values within the boundary cross-section. Moran’s index (normalized) for each region assesses whether nearby or neighboring areas are more similar (or dissimilar) than what would be expected under random spatial distribution.

[0106] To complement the slice-level analysis, muscle-level aggregate metrics were computed by collapsing values across all slices within each muscle. These metrics provide an overall summary of muscle condition. Fat fraction (FF, %) was calculated using the ratio of fat to the combined fat and water signal intensities. Boundary volume (BV, ml) was determined as the total segmented muscle volume. Lean muscle volume (LMV, ml) was calculated by subtracting fat volume from the total boundary volume using the formula: LMV = BV - (FF x BV).

[0107] The volume, average cross-sectional area, and fat fraction of each femur bone was determined. The CSA and volume were used to standardize muscle volume metrics and fat fraction was incorporated because prior work has demonstrated a correlation between trabecular fat fraction and bone mineral density measured by quantitative computed tomography.

[0108] The baseline muscle fat fraction and lean muscle volume data were -29- QB\175340.00034\100682361.5Quarles Ref. 175340.00034analyzed to examine trends across the FSHD population and to establish normative ranges for metrics within the FSHD population. Analysis included: fat fraction for each muscle, fat variation for each muscle, and lean volume (adjusted for body size) for each muscle. In addition to standard descriptive statistics, heatmap visualizations of muscle involvement, with elements organized by the average fat fraction across muscles (within each patient) and TUG times were used to characterize the relationships between patterns of muscle involvement and overall functional and global fat replacement. The relationships between total volume and the product of femur cross-sectional area and lean muscle volume across patients with FSHD were determined, to provide a size adjusted value.

[0109] To integrate contextual information about muscle involvement into the predictive model, two derived metrics were developed based on the relative ranking of muscles by disease involvement. First, the average fat fraction and standard deviation for each muscle across all patients at baseline was calculated. Using these values, a fat fraction z-score for each muscle in each patient was computed to standardize its involvement relative to the cohort distribution. Next, all muscles were ranked based on their cohort-wide average fat fraction, from lowest (least typically involved) to highest (most typically involved). This ranking allowed definition of two contextual metrics for each muscle in each patient. The higher-ranking muscle score for a given muscle was calculated as the average fat fraction z-score of all muscles in that patient ranked higher (i.e., more typically involved muscles with lower ranking numbers). The lower-ranking muscle score was similarly calculated as the average fat fraction z-score of all muscles in that patient ranked lower (i.e., less typically involved muscles with higher ranking numbers). Biologically, a higher higher-ranking muscle score indicates that muscles normally involved earlier in the disease process are already highly infiltrated in this -30- QB\175340.00034\100682361.5Quarles Ref. 175340.00034patient, suggesting a higher likelihood that the current muscle will also progress. In contrast, a higher lower-ranking muscle score suggests that muscle is usually spared until later stages are already highly involved, indicating the disease may have reached a plateau phase, and the current muscle is therefore less likely to progress.

[0110] The D4Z4 repeat length and timed up-and-go (TUG) was included in the baseline description of each patient. For a subset of patients, TUG measurements were also available at follow up time points. All subjects that had longitudinal MRI measures and thus were simulated with the disease progression model were confirmed FSHD Type 1. Disease severity metrics were not the same across all cohorts and thus were not used in the model. Individual patient age and sex were not included in the example implementation of the model because: (1) preliminary analysis indicated that they were not features of high significance, and (2) excluding these features allowed for maximal use of the data because age and sex were not known for all patients.

[0111] The random forest algorithm was chosen for its ability to handle complex, non-linear relationships and its resistance to overfitting. Prior research has established the suitability of this approach for predicting disease progression. To incorporate both regional and total muscle metrics, a multi-stage random forest training and validation approach was developed, as illustrated in FIG. 4. The first two models are designed to predict individual muscle progression over time, with the third model designed to incorporate the output from the first two model stages to predict change in TUG performance. In the first stage, which includes one unique random forest model, training was performed on a subset of the data to predict the changes in fat fraction and lean cross-sectional area at the regional level. This allows the model to capture how the local tissue structure influences progression of fat infiltration and atrophy in a specific region. This model was trained on 10% of the full data set, which included 6330 muscle regions. The -31- QB\175340.00034\100682361.5Quarles Ref. 175340.00034data used for the stage 1 model training were not included in any other training or testing datasets. This approach ensured that each stage was trained and tested on different subsets of data, minimizing the risk of overfitting from incorporating outputs of earlier model stages. The second stage, which includes a two parallel-running random forest model, was trained using another subset of the data (that was not included in the prior training datasets) to predict the whole-muscle level changes in fat fraction and lean muscle volume. Each of these samples was first run through the stage 1 model to predict regionalized changes; then the cross-sectional area weighted average of the regional fat fraction changes was calculated and input into the stage 2 model. This model was trained on 70% of the full data set, which included 4513 muscles, and the model was tested on 833 muscles. Cases with incomplete inputs for key variables were excluded from model training and testing, and no imputation was performed, ensuring that only complete data were used to maintain prediction reliability. For data points where fat fraction decreased or lean muscle volume increased over time, these changes were set to zero. As a result, the model focuses on predicting progression (i.e., degeneration) and does not capture potential improvements. Therefore, a prediction of no change indicates either stabilization or improvement. The outputs of the stage two models were combined to calculate functional group changes by calculating the volume-weighted average of the individual fat fractions at the joint level and summing the lean muscle volumes, normalized by the product of height and mass. Finally, the right and left group-level predicted changes were averaged and incorporated into the stage 3 random forest model to predict change in TUG measurements. The stage 3 model was trained with 26 data points and tested with 4 data points (only a subset of patients included had both coverage of major lower limb muscle groups as well as multiple measurement of TUG).

[0112] FIG. 4 further illustrates the multi-scale disease progression model. Three -32- QB\175340.00034\100682361.5Quarles Ref. 175340.00034separate stages, which examined specific scales of muscle involvement, were developed. Outputs from each stage were used to determine inputs to the next stage. The pipeline begins with MRI-based segmentation and analysis of muscle and bone from whole-body mDixon MRI scans at baseline. Stage 1 predicts regional fat fraction changes using subregion data (eight regions per muscle). Stage 2 aggregates regional predictions to musclelevel, combining them with patient data to predict changes in muscle fat fraction (FF) and lean muscle volume (LMV). Stage 3 integrates muscle-level predictions with baseline timed up-and-go (TUG) measurements to predict changes in functional performance at follow-up. This framework captures the progression of disease from local muscle regions to functional-level outcomes.

[0113] For each stage, a random forest model consisting of around 500 trees was trained on the training dataset. Models were implemented in MATLAB (Mathworks, Natick, MA, USA), usingthe combination oftemplateTree and fitrensemble functions. Grid search was conducted to optimize key hyperparameters such as the number of trees, maximum depth, and minimum samples per split. The out-of-bag method was used to prevent overfitting. Shapley Additive exPlanations (SHAP) analysis was used to identify the key variables contributing to predicted outcome. The final list of features used in each model is provided and described in Table 2.

[0114] Table 2: Model parameters-33- QB\175340.00034\100682361.5Quarles Ref. 175340.00034-34- QB\175340.00034\100682361.5Quarles Ref. 175340.00034

[0115] Each model was evaluated by assessing the ability to predict annualized change in fat fraction for each region (for Stage 1), muscle-level fat fraction or lean muscle volume (for Stage 2) for both the training and testing data samples. The Stage 3 model was evaluated by assessing the ability to predict change in timed up-and-go. The performance accuracy metrics included: mean error (ME), root-mean-square error (RMSE), as well as Pearson’s correlation coefficient between predicted and actual fat fraction or lean muscle volume change (r) and its p-value. Feature importance scores were extracted from the random forest model using SHAP analysis to identify the most influential predictors of fat fraction or lean muscle volume changes.

[0116] Results

[0117] Fat fraction varied substantially across muscles and patients, as demonstrated in FIG. 5. FIG. 5 provides a visualization of 3D segmentations of muscles in patients with FSHD of varying disease severity. Fat fraction % is provided as a colormap on each muscle in the models to illustrate the unique pattern of involvement in each patient and heterogeneous nature of FSHD muscle involvement. These examples highlight regional differences in muscle degeneration, with some muscles showing early or severe fat replacement while others remain relatively spared. This visualization underscores both the widespread and individualized progression patterns of FSHD.-35- QB\175340.00034\100682361.5Quarles Ref. 175340.00034

[0118] On average, the semimembranosus (hamstring) muscle was the most affected in the cohort while the popliteus muscle was least affected. Heatmap visualization of fat fraction across all muscles (organized average fat fraction across individuals) and patients at baseline (organized by TUG) reveal an overall pattern of muscle involvement across disease severities, as shown in FIG. 6.

[0119] FIG. 6 is a visualization of muscle involvement at baseline in patients with FSHD, ordered by timed up-and-go measurement at baseline. The heatmap shows baseline muscle-level fat fraction (%) for each subject, ordered by TUG time (seconds) at baseline (lower plot). Muscles are arranged from most commonly involved (top) to least involved (bottom) across the cohort. The pattern reveals a relationship between higher TUG times and increased fat infiltration in multiple muscles, illustrating disease progression and heterogeneity in FSHD.

[0120] As TUG time increases, more muscles have fat fraction levels in the high range, consistent with the progressive nature of the disease. The order of muscles is not regional: the muscles in the ‘earlier’ group are in both upper and lower regions of the body; similarly, muscles in the ‘later’ group are muscles from all regions as well. Substantial heterogeneity across patients and muscles is most apparent in the patients in the middle range of involvement.

[0121] The region-specific random forest model (Stage 1) showed a significant correlation with measured regional fat fraction per muscle (r = 0.43, p < 0.001), with a mean error of -0.32% and an RMSE of 3.5% in the testing dataset, as shown in FIGS. 7A-7G. SHAP analysis revealed the parameters: Regional Fat Fraction Kurtosis, Regional Fat Fraction Variation, Difference from FF Neighbors, and Relative CSA had the greatest influences on prediction of fat fraction change at the regional level.

[0122] FIGS. 7A-7G provide a set of graphs illustrating a stage 1 model for -36- QB\175340.00034\100682361.5Quarles Ref. 175340.00034Regional-Level Fat Fraction Change Prediction. Each muscle is separated into 8 separate regions longitudinally, and the model predicts the change in fat fraction of the region based on measurements at the regional level. The output of this stage is combined into a singular prediction (by averaging the predictions across the regions within each muscle) that serves as a variable input to the Stage 2 models. FIGS. 7A and 7D show the predicted vs. measured regional fat fraction change (%) for training (FIG. 7 A) and testing (FIG. 7D) datasets. Strong correlation in training (r = 0.89, p < 0.001) and moderate in testing (r = 0.43, p < 0.001) demonstrate predictive performance. FIGS. 7B and 7E show Bland-Altman plots comparing measured and modeled changes. Training shows minimal error (mean = 0.00%, RMSE = 1.62%), while testing shows modest bias (mean = 0.32%, RMSE = 3.50%). FIGS. 7C and 7F provide histograms of prediction errors (model-measured) with centered distributions in training (n = 6330) and testing (n = 7218). FIG. 7G shows a SHAP beeswarm plot showing the contribution of key features to fat fraction change predictions. Each dot represents a regional sample, with SHAP values indicating feature impact and colors reflecting feature values. Top predictors include fat fraction kurtosis, skewness, variation, neighbor differences, and CSA metrics.

[0123] At the muscle level (Stage 2), the random forest model predicted fat fraction change with a significant correlation to measured changes (r = 0.5, p < 0.001), a mean error of -0.3%, and an RMSE of 2.2%, as shown in FIGS.8A-8G. Regional Prediction (output from Stage 1 model), femur cross-sectional area, higher-ranking and lower-ranking muscle scores, and muscle had the greatest influence on prediction of musclelevel fat fraction change.

[0124] FIGS.8A-8G provide a set of graphs illustrating a stage 2 model for Muscle-Level Fat Fraction Change Prediction. FIGS. 8A and 8D plot the predicted vs. measured muscle-level fat fraction change (%) for training (FIG. 8A) and testing (FIG. 8D) datasets.-37- QB\175340.00034\100682361.5Quarles Ref. 175340.00034Strong training correlation (r = 0.95, p < 0.001) and moderate testing correlation (r = 0.50, p < 0.001) demonstrate good model performance. FIGS. 8B and 8E show Bland-Altman plots comparing measured and modeled changes. Training error is minimal (mean = 0.04%, RMSE = 0.93%) while testing shows a slight negative bias (mean = -0.31%, RMSE = 2.16%). FIGS. 8C and 8F provide histograms of prediction errors (model-measured) for training (n = 4513) and testing (n = 833) samples, centered around zero. FIG.8G shows a SHAP beeswarm plot showing contributions of key features to muscle-level fat fraction change prediction. Each dot represents a muscle sample, with SHAP values indicating feature impact and color representing feature value. Top predictors include regional prediction (Stage 1 output), D4Z4 repeat length, baseline fat fractions (muscle and femur), femur CSA, ranking scores, fat fraction ranges, and muscle identity.

[0125] For predicting lean muscle volume change, the Stage 2 model showed significant correlation with measured changes (r = 0.57, p < 0.001), a mean error of -0.8 ml, and an RMSE of 8.3 ml, as shown in FIGS. 9A-9G. Femur cross-sectional area, D4Z4 repeat length, LMV to femur CSA z-score, higher-ranking muscle scores, and muscle had the greatest influences on prediction of muscle-level lean muscle volume change.

[0126] FIGS. 9A-9G provide a set of graphs illustrating a stage 2 model for Lean Muscle Volume Changes Prediction. FIGS. 9A and 9D show the predicted vs. measured lean muscle volume change (ml) for training (FIG.9A) and testing (FIG.9D) datasets. The model showed strong correlation in training (r = 0.95, p < 0.001) and moderate correlation in testing (r = 0.57, p < 0.001). FIGS. 9B and 9E show Bland-Altman plots showing measured minus modeled differences against their average for training and testing datasets. Training error is minimal (mean = 0.02 ml, RMSE = 3.05 ml), while testing shows a small negative bias (mean = -0.80 ml, RMSE = 8.13 ml). FIGS. 9C and 9F -38- QB\175340.00034\100682361.5Quarles Ref. 175340.00034show histograms of prediction errors (model-measured) for training (n = 4513) and testing (n = 833) samples, with distributions centered around zero. FIG. 9G provides a SHAP beeswarm plot illustrating the impact of features on lean muscle volume change predictions. Each dot represents a muscle sample; the x-axis shows SHAP values (feature contributions), and the color indicates feature value. Key predictors include Stage 1 regional predictions, LMV to bone z-score, D4Z4 repeat length, baseline femur fat fraction, femur CSA, ranking scores, and fat fraction range across regions.

[0127] Combining the predicted individual fat fractions and lean muscle volumes to calculate functional group-level changes in fat fraction and lean muscle volume accurately predicted variability in functional-group level changes in the testing dataset, as demonstrated in FIGS. 10A-10B. For group-level fat fraction change prediction, all RMSE values were less than 2.5%, with the exception of the trunk flexors (5.18%). Group-level lean muscle volume change predictions showed RMSE values ranging from 3.3 ml (hip external rotators) to 26.7 ml (knee extensors).

[0128] FIGS. 10A-10B provide a set of graphs illustrating the evaluation of the output of stage 2 when determining group-level fat fraction (FIG. 10A) and lean muscle volume (FIG. 10B), comparing model prediction and measured values for training (solid circles) and testing (open circles) datasets. FIG. 10A shows the predicted vs. measured fat fraction (FF) changes (%) for various muscle functional groups across the body. Each scatter plot shows predictions for a specific functional group, including shoulder abductors, hip abductors, trunk extensors, hip flexors, knee extensors, and ankle dorsiflexors, among others. Filled circles represent training data, and open circles represent testing data. Model performance metrics, including mean error and RMSE (root mean square error), are reported for the testing data of each group. The diagonal line indicates perfect prediction. FIG. 10B shows the predicted vs. measured lean muscle -39- QB\175340.00034\100682361.5Quarles Ref. 175340.00034volume (LMV) changes (ml) for the same functional groups. Model accuracy, including mean error and RMSE, is reported for the testing dataset in each group. The plots demonstrate the model’s ability to predict group-level changes in both FF and LMV across multiple muscle groups, reflecting functional variations in disease progression.

[0129] The functional-level model (Stage 3) predicted change in TUG from the first to the second time point with an RMSE of 0.75 s in the testing dataset, as shown in FIGS. E8A-E8B. SHAP analyses revealed that the TUG at baseline had the largest influence on the predicted change in TUG, followed by the normalized change in lean muscle volume of the Trunk Extensors, Hip Extensors, and Knee Extensors muscle functional groups.

[0130] FIGS. E8A-E8B show performance of the Stage 3 model in the prediction of change in timed up-and-go (TUG). FIG. 8A plots predicted vs. measured change in TUG (seconds) for the model. Filled circles represent training data (n = 26) and open circles represent testing data (n = 4). The diagonal line indicates perfect prediction. The test dataset shows a mean error of 0.60 seconds and an RMSE of 0.75 seconds. FIG. E8B provides a SHAP beeswarm plot illustrating the contribution of key features to the model’s TUG prediction. Each point represents a sample, with horizontal position indicating the SHAP value (impact on prediction) and color representing the feature value. The most influential predictors include TUG at baseline, and normalized changes in lean muscle volume (LV) of knee extensors, hip abductors, hip extensors, hip flexors, and trunk extensors. This analysis highlights the factors driving functional performance prediction in the model.

[0131] Discussion

[0132] A major challenge for drug development in neuromuscular disease is how to measure the effects of treatment in chronic progressive disorders that are marked by high variability in severity at baseline and in rates of progression. Clinical trials in FSHD -40- QB\175340.00034\100682361.5Quarles Ref. 175340.00034currently adopt a "one-size-fits-all” approach to evaluating drug efficacy. Strategies that rely on composite metrics from MRI or performance tests risk regressing small changes to the mean, thereby reducing sensitivity in detecting meaningful differences and prolonging the required trial duration. Conversely, using a single task as a reporter is typically only effective if participants are pre-selected based on muscle involvement relevant to that task. Otherwise, a substantial proportion of participants, potentially unevenly distributed across placebo and treatment groups, may fail to exhibit measurable change in task performance. In this study, a novel ‘digital twin’ approach was presented to predict personalized disease progression at multiple length scales: muscle region, muscle, group, and functional levels. The developed method integrates a threetiered machine-learning model, each component designed to capture progression at a distinct scale. This framework enables the prediction of which muscles, and to what extent, will undergo changes over time during the natural course of the disease, offering a more precise and individualized approach to monitoring FSHD progression in clinical trials. Furthermore, the feature importance analyses provide insight into which factors best predict disease progression. In these models, metrics of fat fraction variability (within regions and whole muscle) were most associated with fat progression changes over one year, with bone metrics playing an important role in lean muscle volume decline over one year.

[0133] Given the heterogeneity and nonlinearity of the disease, each patient has a specific set of muscles that are most likely to change ata given time. As might be expected in a heterogeneous disease, some muscle groups changed more than others in each patient, which has implications for the functional metrics that best identify disease progression in a particular patient. For example, the muscles that best predict TUG - the trunk extensors, hip extensors, and knee extensors - only changed in a subset of patients;-41- QB\175340.00034\100682361.5Quarles Ref. 175340.00034therefore, TUG only changed in a subset of patients as well. Not surprisingly, the TUG at the baseline timepoint demonstrated a strong relationship with predicted change in TUG, as well. Taken together, these results suggest that the disease progression model provides a new approach for clinical trials that allows for patient-specific selection of functional metrics that are most likely to detect deviations from each individual’s disease progression and / or selection of qualified patients if specific functional metrics are of interest.

[0134] Analysis of the models provided insight into the parameters that influenced model prediction of disease progression. At the regional (Stage 1 model) level, the initial fat fraction, the fat variation, and the kurtosis had the greatest influence on predicted fat fraction changes, illustrating that progression patterns are associated with the heterogeneity of fat within the muscle. At the whole muscle (Stage 2 model) level, the predicted fat fraction change from the Stage 1 model, the muscle, the D4Z4 repeat length, the associated muscle group, femur size, femur fat fraction, and metric of overall involvement influenced change in fat fraction prediction. The explanations for some of these predictive features are intuitive and consistent with current knowledge, while others are novel. The fact that ‘muscle’ was a key feature contributing to prediction accuracy indicates that each muscle’s identity influences its likelihood of progression, supporting the notion of muscle-specific progression patterns that has already been demonstrated. Similarly, the fact that ‘D4Z4 length’ is associated with faster progression is also expected. However, the mechanisms by which bone size and composition affect disease progression represent additional complexities yet to be fully unraveled. It is contemplated that these bone metrics (e.g., femur CSA and fat fraction) reflect unmodeled confounders such as sex, body size, or age, which were not included in the example implementation of the model. In other implementations, these variables can be -42- QB\175340.00034\100682361.5Quarles Ref. 175340.00034incorporated into the model to clarify these associations and their potential biological relevance. Overall, these results demonstrate that the complex nature of disease progression can be handled within iterative model building, allowing insights into primary predictors and the individualized nature of FSHD.

[0135] The model presented in this study furthers work in the field demonstrating general features of disease progression as determined by muscle and fat measurements on MRL For example, other studies have shown that a small percentage (~ 5%) of fat-affected muscles also exhibit signal elevation on short tau-inversion-recoveiy (STIR+) sequences. This STIR+ signal can be indicative of intramuscular edema or inflammation. While this biomarker has been suggested to foreshadow faster progression in the literature, these studies did not discriminate between muscle fat fraction at baseline, fat pattern, or muscle identity. Notably, the stability of the STIR+ signal, even when challenged with immunosuppression and steroid treatment, combined with long imaging times required by STIR sequences, highlights a limitation of STIR as a dynamic biomarker for disease progression. STIR imaging was not consistently available across the cohorts included in this study, further limiting its potential utility in this analysis. However, in some implementations, STIR imaging data, as well as complementary imaging markers such as water T2 (wT2) can be incorporated into the model to help clarify how these biomarkers independently or synergistically predict disease progression.

[0136] This work represents a significant advancement in the use of machine learning and imaging to address the challenges of disease heterogeneity in FSHD. By integrating advanced MRI-derived muscle metrics with clinical and biological data, the proposed multi-scale framework reveals that individual muscle progression over a year interval is predictable, given a comprehensive integration of personalized data at baseline. The digital twin model serves as a powerful benchmark for assessing untreated -43- QB\175340.00034\100682361.5Quarles Ref. 175340.00034progression, enhancing the precision and efficiency of tracking natural history changes over time. This advancement has broad potential applications in clinical trial design, including the use of personalized digital twins as surrogate placebos and leveraging the progression model’s identified features to better isolate ‘at-risk’ muscles in traditional trial designs. While incorporating additional biomarkers and expanding datasets will further refine predictive capabilities, this study lays the foundation for applying personalized machine learning and imaging approaches to address variability inherent in neuromuscular diseases and to enable more precise, efficient, and patient-tailored clinical trials.

[0137] Example Systems

[0138] Referring particularly now to Fig. 11, an example of an MRI system 1100 that can implement the methods described herein is illustrated. The MRI system 1100 includes an operator workstation 1102 that may include a display 1104, one or more input devices 1106 (e.g., a keyboard, a mouse), and a processor 1108. The processor 1108 may include a commercially available programmable machine running a commercially available operating system. The operator workstation 1102 provides an operator interface that facilitates entering scan parameters into the MRI system 1100. The operator workstation 1102 may be coupled to different servers, including, for example, a pulse sequence server 1110, a data acquisition server 1112, a data processing server 1114, and a data store server 1116. The operator workstation 1102 and the servers 1110, 1112, 1114, and 1116 may be connected via a communication system 1140, which may include wired or wireless network connections.

[0139] The MRI system 1100 also includes a magnet assembly 1124 that includes a polarizing magnet 1126. The polarizing magnet 1126 produces a main magnetic field (Bo). The magnet assembly 1124 may also include shimming coils that can shape Bo. The .44.QB\175340.00034\100682361.5Quarles Ref. 175340.00034MRI system 1100 may optionally include a whole-body RF coil 1128 and a gradient system 1118 that controls a gradient coil assembly 1122.

[0140] The pulse sequence server 1110 functions in response to instructions provided by the operator workstation 1102 to operate a gradient system 1118 and a radiofrequency ("RF”) system 1120. Gradient waveforms for performing a prescribed scan are produced and applied to the gradient system 1118, which then excited gradient coils in an assembly 1122 to produce the magnetic field gradients (e.g., Gx, Gy, and Gzthat can be used for spatially encoding magnetic resonance signals. The gradient coil assembly 1122 forms part of a magnet assembly 1124 that includes a polarizing magnet 1126 and a whole-body RF coil 1128.

[0141] RF waveforms are applied by the RF system 1120 to the RF coil 1128, or a separate local coil to perform the prescribed magnetic resonance pulse sequence. Responsive magnetic resonance signals detected by the RF coil 1128, or a separate local coil, are received by the RF system 1120. The responsive magnetic resonance signals may be amplified, demodulated, filtered, and digitized under direction of commands produced by the pulse sequence server 1110. The RF system 1120 includes an RF transmitter for producing a wide variety of RF pulses used in MRI pulse sequences. The RF transmitter is responsive to the prescribed scan and direction from the pulse sequence server 1110 to produce RF pulses of the desired frequency, phase, and pulse amplitude waveform. The generated RF pulses may be applied to the whole-body RF coil 1128 or to one or more local coils or coil arrays.

[0142] The RF system 1120 also includes one or more RF receiver channels. An RF receiver channel includes an RF preamplifier that amplifies the magnetic resonance signal received by the coil 1128 to which it is connected, and a detector that detects and digitizes the I and Q quadrature components of the received magnetic resonance signal.-45- QB\175340.00034\100682361.5Quarles Ref. 175340.00034The magnitude of the received magnetic resonance signal may, therefore, be determined at a sampled point by the square root of the sum of the squares of the I and Q components:M = 7U2+ Q2)

[0143] and the phase of the received magnetic resonance signal may also be determined according to the following relationship:<

[0144] The pulse sequence server 1110 may receive patient data from a physiological acquisition controller 1130. By way of example, the physiological acquisition controller 1130 may receive signals from a number of different sensors connected to the patient, including electrocardiograph ("ECG”) signals from electrodes, or respiratory signals from a respiratory bellows or other respiratory monitoring devices. These signals may be used by the pulse sequence server 1110 to synchronize, or "gate,” the performance of the scan with the subject’s heartbeat or respiration.

[0145] The pulse sequence server 1110 may also connect to a scan room interface circuit 1132 that receives signals from various sensors associated with the condition of the patient and the magnet system. Through the scan room interface circuit 1132, a patient positioning system 1134 can receive commands to move the patient to desired positions during the scan.

[0146] The digitized magnetic resonance signal samples produced by the RF system 1120 are received by the data acquisition server 1112. The data acquisition server 1112 operates in response to instructions downloaded from the operator workstation 1102 to receive the real-time magnetic resonance data and provide buffer storage, so that data are not lost by data overrun. In some scans, the data acquisition server 1112 passes-46- QB\175340.00034\100682361.5Quarles Ref. 175340.00034the acquired magnetic resonance data to the data processor server 1114. In scans that require information derived from acquired magnetic resonance data to control the further performance of the scan, the data acquisition server 1112 may be programmed to produce such information and convey it to the pulse sequence server 1110. For example, during pre-scans, magnetic resonance data may be acquired and used to calibrate the pulse sequence performed by the pulse sequence server 1110. As another example, navigator signals may be acquired and used to adjust the operating parameters of the RF system 1120 or the gradient system 1118, or to control the view order in which k-space is sampled. In still another example, the data acquisition server 1112 may also process magnetic resonance signals used to detect the arrival of a contrast agent in a magnetic resonance angiography ("MRA”) scan. For example, the data acquisition server 1112 may acquire magnetic resonance data and processes it in real-time to produce information that is used to control the scan.

[0147] The data processing server 1114 receives magnetic resonance data from the data acquisition server 1112 and processes the magnetic resonance data in accordance with instructions provided by the operator workstation 1102. Such processing may include, for example, reconstructing two-dimensional or three-dimensional images by performing a Fourier transformation of raw k-space data, performing other image reconstruction algorithms (e.g., iterative or backprojection reconstruction algorithms), applying filters to raw k-space data or to reconstructed images, generating functional magnetic resonance images, or calculating motion or flow images.

[0148] Images reconstructed by the data processing server 1114 are conveyed back to the operator workstation 1102 for storage. Real-time images may be stored in a data base memory cache, from which they may be output to operator display 1102 or a -47- QB\175340.00034\100682361.5Quarles Ref. 175340.00034display 1136. Batch mode images or selected real time images may be stored in a host database on disc storage 1138. When such images have been reconstructed and transferred to storage, the data processing server 1114 may notify the data store server 1116 on the operator workstation 1102. The operator workstation 1102 may be used by an operator to archive the images, produce films, or send the images via a network to other facilities.

[0149] The MRI system 1100 may also include one or more networked workstations 1142. For example, a networked workstation 1142 may include a display 1144, one or more input devices 1146 (e.g., a keyboard, a mouse), and a processor 1148. The networked workstation 1142 may be located within the same facility as the operator workstation 1102, or in a different facility, such as a different healthcare institution or clinic.

[0150] The networked workstation 1142 may gain remote access to the data processing server 1114 or data store server 1116 via the communication system 1140. Accordingly, multiple networked workstations 1142 may have access to the data processing server 1114 and the data store server 1116. In this manner, magnetic resonance data, reconstructed images, or other data may be exchanged between the data processing server 1114 or the data store server 1116 and the networked workstations 1142, such that the data or images may be remotely processed by a networked workstation 1142.

[0151] Referring now to FIG. 12, an example of an MRI system 1200 is shown, which may be used in accordance with some aspects of the systems and methods described in the present disclosure. As shown in FIG. 12, a computing device 1250 can receive one or more types of data (e.g., signal evolution data, k-space data, receiver coil sensitivity data) from data source 1202. In some configurations, computing device 1250-48- QB\175340.00034\100682361.5Quarles Ref. 175340.00034can execute at least a portion of a disease progression prediction system 1204 to reconstruct images from magnetic resonance data (e.g., k-space data), which may be acquired using a Dixon or STIR+ technique. In some configurations, the disease progression prediction system 1204 can implement an automated pipeline to provide MRI images, segmented MRI images, musculoskeletal metrics based on MRI data, etc.

[0152] Additionally or alternatively, in some configurations, the computing device 1250 can communicate information about data received from the data source 1202 to a server 1252 over a communication network 1254, which can execute at least a portion of the disease progression prediction system 1204. In such configurations, the server 1252 can return information to the computing device 1250 (and / or any other suitable computing device) indicative of an output of the disease progression prediction system 1204.

[0153] In some configurations, computing device 1250 and / or server 1252 can be any suitable computing device or combination of devices, such as a desktop computer, a laptop computer, a smartphone, a tablet computer, a wearable computer, a server computer, a virtual machine being executed by a physical computing device, and so on. The computing device 1250 and / or server 1252 can also reconstruct images from the data.

[0154] In some configurations, data source 1202 can be any suitable source of data (e.g., measurement data, images reconstructed from measurement data, processed image data), such as an MRI system, another computing device (e.g., a server storing measurement data, images reconstructed from measurement data, processed image data), and so on. In some configurations, data source 1202 can be local to computing device 1250. For example, data source 1202 can be incorporated with computing device 1250 (e.g., computing device 1250 can be configured as part of a device for measuring,-49- QB\175340.00034\100682361.5Quarles Ref. 175340.00034recording, estimating, acquiring, or otherwise collecting or storing data). As another example, data source 1202 can be connected to computing device 1250 by a cable, a direct wireless link, and so on. Additionally or alternatively, in some configurations, data source 1202 can be located locally and / or remotely from computing device 1250, and can communicate data to computing device 1250 (and / or server 1252) via a communication network (e.g., communication network 1254).

[0155] In some configurations, communication network 1254 can be any suitable communication network or combination of communication networks. For example, communication network 1254 can include a Wi-Fi network (which can include one or more wireless routers, one or more switches, etc.), a peer-to-peer network (e.g., a Bluetooth network), a cellular network (e.g., a 3G network, a 4G network, etc., complying with any suitable standard, such as CDMA, GSM, LTE, LTE Advanced, WiMAX, etc.), other types of wireless network, a wired network, and so on. In some configurations, communication network 1254 can be a local area network, a wide area network, a public network (e.g., the Internet), a private or semi-private network (e.g., a corporate or university intranet), any other suitable type of network, or any suitable combination of networks. Communications links shown in FIG. 12 can each be any suitable communications link or combination of communications links, such as wired links, fiber optic links, Wi-Fi links, Bluetooth links, cellular links, and so on.

[0156] Referring now to FIG. 13, an example of hardware 1300 that can be used to implement data source 1202, computing device 1250, and server 1252 in accordance with some configurations of the systems and methods described in the present disclosure is shown.

[0157] As shown in FIG. 13, in some configurations, computing device 1250 can include a processor 1302, a display 1304, one or more inputs 1306, one or more -50- QB\175340.00034\100682361.5Quarles Ref. 175340.00034communication systems 1308, and / or memory 1310. In some configurations, processor 1302 can be any suitable hardware processor or combination of processors, such as a central processing unit ("CPU”), a graphics processing unit ("GPU”), and so on. In some configurations, display 1304 can include any suitable display devices, such as a liquid crystal display ("LCD”) screen, a light- emitting diode ("LED”) display, an organic LED ("OLED”) display, an electrophoretic display (e.g., an "e-ink” display), a computer monitor, a touchscreen, a television, and so on. In some configurations, inputs 1306 can include any suitable input devices and / or sensors that can be used to receive user input, such as a keyboard, a mouse, a touchscreen, a microphone, and so on.

[0158] In some configurations, communications systems 1308 can include any suitable hardware, firmware, and / or software for communicating information over communication network 1254 and / or any other suitable communication networks. For example, communications systems 1308 can include one or more transceivers, one or more communication chips and / or chip sets, and so on. In a more particular example, communications systems 1308 can include hardware, firmware, and / or software that can be used to establish a Wi-Fi connection, a Bluetooth connection, a cellular connection, an Ethernet connection, and so on.

[0159] In some configurations, memory 1310 can include any suitable storage device or devices that can be used to store instructions, values, data, or the like, that can be used, for example, by processor 1302 to present content using display 1304, to communicate with server 1252 via communications system(s) 1308, and so on. Memory 1310 can include any suitable volatile memory, non-volatile memory, storage, or any suitable combination thereof. For example, memory 1310 can include random-access memory ("RAM”), read-only memory ("ROM”), electrically programmable ROM ("EPROM”), electrically erasable ROM ("EEPROM”), other forms of volatile memory, other -51- QB\175340.00034\100682361.5Quarles Ref. 175340.00034forms of non-volatile memory, one or more forms of semi-volatile memory, one or more flash drives, one or more hard disks, one or more solid state drives, one or more optical drives, and so on. In some configurations, memory 1310 can have encoded thereon, or otherwise stored therein, a computer program for controlling operation of computing device 1250. In such configurations, processor 1302 can execute at least a portion of the computer program to present content (e.g., images, user interfaces, graphics, tables), receive content from server 1252, transmit information to server 1252, and so on. For example, the processor 1302 and the memory 1310 can be configured to perform the methods described herein.

[0160] In some configurations, server 1252 can include a processor 1312, a display 1314, one or more inputs 1316, one or more communications systems 1318, and / or memory 1320. In some configurations, processor 1312 can be any suitable hardware processor or combination of processors, such as a CPU, a GPU, and so on. In some configurations, display 1314 can include any suitable display devices, such as an LCD screen, LED display, OLED display, electrophoretic display, a computer monitor, a touchscreen, a television, and so on. In some configurations, inputs 1316 can include any suitable input devices and / or sensors that can be used to receive user input, such as a keyboard, a mouse, a touchscreen, a microphone, and so on.

[0161] In some configurations, communications systems 1318 can include any suitable hardware, firmware, and / or software for communicating information over communication network 1254 and / or any other suitable communication networks. For example, communications systems 1318 can include one or more transceivers, one or more communication chips and / or chip sets, and so on. In a more particular example, communications systems 1318 can include hardware, firmware, and / or software that can be used to establish a Wi-Fi connection, a Bluetooth connection, a cellular connection,-52- QB\175340.00034\100682361.5Quarles Ref. 175340.00034an Ethernet connection, and so on.

[0162] In some configurations, memory 1320 can include any suitable storage device or devices that can be used to store instructions, values, data, or the like, that can be used, for example, by processor 1312 to present content using display 1314, to communicate with one or more computing devices 1250, and so on. Memory 1320 can include any suitable volatile memory, non-volatile memory, storage, or any suitable combination thereof. For example, memory 1320 can include RAM, ROM, EPROM, EEPROM, other types of volatile memory, other types of non-volatile memory, one or more types of semi-volatile memory, one or more flash drives, one or more hard disks, one or more solid state drives, one or more optical drives, and so on. In some configurations, memory 1320 can have encoded thereon a server program for controlling operation of server 1252. In such configurations, processor 1312 can execute at least a portion of the server program to transmit information and / or content (e.g., data, images, a user interface) to one or more computing devices 1250, receive information and / or content from one or more computing devices 1250, receive instructions from one or more devices (e.g., a personal computer, a laptop computer, a tablet computer, a smartphone), and so on.

[0163] In some configurations, the server 1252 is configured to perform the methods described in the present disclosure. For example, the processor 1312 and memory 1320 can be configured to perform the methods described herein.

[0164] In some configurations, data source 1202 can include a processor 1322, one or more data acquisition systems 1324, one or more communications systems 1326, and / or memory 1328. In some configurations, processor 1322 can be any suitable hardware processor or combination of processors, such as a CPU, a GPU, and so on. In some configurations, the one or more data acquisition systems 1324 are generally -53- QB\175340.00034\100682361.5Quarles Ref. 175340.00034configured to acquire data, images, or both, and can include an MRI system. Additionally or alternatively, in some configurations, the one or more data acquisition systems 1324 can include any suitable hardware, firmware, and / or software for coupling to and / or controlling operations of an MRI system. In some configurations, one or more portions of the data acquisition system(s) 1324 can be removable and / or replaceable.

[0165] Note that, although not shown, data source 1202 can include any suitable inputs and / or outputs. For example, data source 1202 can include input devices and / or sensors that can be used to receive user input, such as a keyboard, a mouse, a touchscreen, a microphone, a trackpad, a trackball, and so on. As another example, data source 1202 can include any suitable display devices, such as an LCD screen, an LED display, an OLED display, an electrophoretic display, a computer monitor, a touchscreen, a television, etc., one or more speakers, and so on.

[0166] In some configurations, communications systems 1326 can include any suitable hardware, firmware, and / or software for communicating information to computing device 1250 (and, in some configurations, over communication network 1254 and / or any other suitable communication networks). For example, communications systems 1326 can include one or more transceivers, one or more communication chips and / or chip sets, and so on. In a more particular example, communications systems 1326 can include hardware, firmware, and / or software that can be used to establish a wired connection using any suitable port and / or communication standard (e.g., VGA, DVI video, USB, RS-232, etc.), Wi-Fi connection, a Bluetooth connection, a cellular connection, an Ethernet connection, and so on.

[0167] In some configurations, memory 1328 can include any suitable storage device or devices that can be used to store instructions, values, data, or the like, that can be used, for example, by processor 1322 to control the one or more data acquisition -54- QB\175340.00034\100682361.5Quarles Ref. 175340.00034systems 1324, and / or receive data from the one or more data acquisition systems 1324; to generate images from data; present content (e.g., data, images, a user interface) using a display; communicate with one or more computing devices 1250; and so on. Memory 1328 can include any suitable volatile memory, non-volatile memory, storage, or any suitable combination thereof. For example, memory 1328 can include RAM, ROM, EPROM, EEPROM, other types of volatile memory, other types of non-volatile memory, one or more types of semi-volatile memory, one or more flash drives, one or more hard disks, one or more solid state drives, one or more optical drives, and so on. In some configurations, memory 1328 can have encoded thereon, or otherwise stored therein, a program for controlling operation of medical image data source 1202. In such configurations, processor 1322 can execute at least a portion of the program to generate images, transmit information and / or content (e.g., data, images, a user interface) to one or more computing devices 1250, receive information and / or content from one or more computing devices 1250, receive instructions from one or more devices (e.g., a personal computer, a laptop computer, a tablet computer, a smartphone, etc.), and so on.

[0168] In some configurations, any suitable computer-readable media can be used for storing instructions for performing the functions and / or processes described herein. For example, in some configurations, computer-readable media can be transitory or non-transitory. For example, non-transitory computer-readable media can include media such as magnetic media (e.g., hard disks, floppy disks), optical media (e.g., compact discs, digital video discs, Blu-ray discs), semiconductor media (e.g., RAM, flash memory, EPROM, EEPROM), any suitable media that is not fleeting or devoid of any semblance of permanence during transmission, and / or any suitable tangible media. As another example, transitory computer-readable media can include signals on networks, in wires, conductors, optical fibers, circuits, or any suitable media that is fleeting and devoid of any -55- QB\175340.00034\100682361.5Quarles Ref. 175340.00034semblance of permanence during transmission, and / or any suitable intangible media.

[0169] As used herein in the context of computer implementation, unless otherwise specified or limited, the terms "component," "system," "module," "controller," "framework," and the like are intended to encompass part or all of computer-related systems that include hardware, software, a combination of hardware and software, or software in execution. For example, a component may be, but is not limited to being, a processor device, a process being executed (or executable) by a processor device, an object, an executable, a thread of execution, a computer program, or a computer. By way of illustration, both an application running on a computer and the computer can be a component. One or more components (or system, module, and so on) may reside within a process or thread of execution, may be localized on one computer, may be distributed between two or more computers or other processor devices, or may be included within another component (or system, module, and so on).

[0170] In some implementations, devices or systems disclosed herein can be utilized or installed using methods embodying aspects of the disclosure. Correspondingly, description herein of particular features, capabilities, or intended purposes of a device or system is generally intended to inherently include disclosure of a method of using such features for the intended purposes, a method of implementing such capabilities, and a method of installing disclosed (or otherwise known) components to support these purposes or capabilities. Similarly, unless otherwise indicated or limited, discussion herein of any method of manufacturing or using a particular device or system, including installing the device or system, is intended to inherently include disclosure, as embodiments of the disclosure, of the utilized features and implemented capabilities of such device or system.

[0171] As used herein, the phrase "at least one of A, B, and C" means at least one -56- QB\175340.00034\100682361.5Quarles Ref. 175340.00034of A, at least one of B, and / or at least one of C, or any one of A, B, or C or combination of A, B, or C. A, B, and C are elements of a list, and A, B, and C may be anything contained in the Specification.

[0172] The present disclosure has described one or more preferred embodiments, and it should be appreciated that many equivalents, alternatives, variations, and modifications, aside from those expressly stated, are possible and within the scope of the invention.-57- QB\175340.00034\100682361.5

Claims

Quarles Ref. 175340.00034CLAIMS1. A system for generating a report of disease progression, comprising: a data collection module configured to collect magnetic resonance imaging (MRI) data and clinical data from a patient;a multi-stage machine learning model comprising:a first stage model configured to predict regional-level changes in muscle composition based on at least the MRI data;a second stage model configured to predict muscle-level changes in muscle composition based on outputs from the first stage model and the clinical data from the patient;a third stage model configured to predict changes in functional outcomes based on outputs from the second stage model; andan output module configured to generate a report including a personalized disease progression prediction for the patient based on the multi-stage machine learning model.

2. The system of claim 1, wherein the first stage model is configured to predict regional-level changes in fat fraction and lean cross-sectional area.

3. The system of claim 1, wherein the second stage model comprises two parallel random forest models for predicting fat fraction changes and lean muscle volume changes, respectively.

4. The system of claim 1, wherein the third stage model is configured to predict changes in a timed up-and-go (TUG) test performance.

5. The system of claim 1, wherein the MRI data comprises whole-body Dixon imaging data.

6. The system of claim 5, wherein the data collection module is further configured control an MRI system to acquire the MRI data for segmentation of individual muscles from the whole-body Dixon imaging data.-58- QB\175340.00034\100682361.5Quarles Ref. 175340.000347. The system of claim 1, wherein the data collection module performs an Al-based segmentation.

8. A computer implemented method for generating a report of disease progression for a patient, comprising:controlling a magnetic resonance imaging (MRI) system to acquire MRI data; providing the MRI data into a first stage machine learning model to predict regional-level changes in muscle composition;providing outputs from the first stage model into a second stage machine learning model to predict muscle-level changes in muscle composition;providing outputs from the second stage model into a third stage machine learning model to predict changes in functional outcomes; andgenerating a report on personalized disease progression prediction for the patient based on outputs from the third stage model.

9. The method of claim 8, wherein predicting regional-level changes in muscle composition comprises predicting changes in fat fraction and lean cross-sectional area for multiple regions within individual muscles; and wherein predicting muscle-level changes in muscle composition comprises predicting changes in fat fraction and lean muscle volume for at least one individual muscle.

10. The method of claim 9, wherein the first stage machine learning model comprises a random forest model trained on regional-level muscle composition data.

11. The method of claim 8, wherein the second stage machine learning model comprises two parallel random forest models for predicting fat fraction changes and lean muscle volume changes, respectively.

12. The method of claim 11, wherein inputs to the second stage machine learning model include D4Z4 repeat length and femur cross-sectional area measurements.-59- QB\175340.00034\100682361.5Quarles Ref. 175340.0003413. The method of claim 8, wherein predicting changes in functional outcomes comprises predicting changes in timed up-and-go (TUG) test performance.

14. The method of claim 13, wherein inputs to the third stage machine learning model include baseline functional performance and predicted changes in lean muscle volume for a functional muscle group.

15. A non-transitory computer-readable medium storing instructions that, when executed by a processor, cause the processor to perform operations for generating reports predicting disease progression, the operations comprising:receiving magnetic resonance imaging (MRI) data and clinical data from a patient;applying a multi-stage machine learning model to the received data, wherein the multistage model comprises:a first stage for predicting regional-level changes in muscle composition; a second stage for predicting muscle-level changes in muscle composition; a third stage for predicting changes in functional outcomes; and outputting a personalized disease progression prediction for the patient based on results from the multi-stage machine learning model.

16. The non-transitory computer-readable medium of claim 15, wherein the first stage of the multi-stage machine learning model comprises a random forest model trained to predict changes in fat fraction and lean cross-sectional area for multiple regions within individual muscles.

17. The non-transitory computer-readable medium of claim 16, wherein the second stage of the multi-stage machine learning model comprises two parallel random forest models for predicting fat fraction changes and lean muscle volume changes, respectively.

18. The non-transitory computer-readable medium of claim 17, wherein inputs to the second stage of the multi-stage machine learning model include D4Z4 repeat length and femur cross-sectional area measurements.-60- QB\175340.00034\100682361.5Quarles Ref. 175340.0003419. The non-transitory computer-readable medium of claim 18, wherein the third stage of the multi-stage machine learning model is configured to predict changes in timed up-and-go (TUG) performance.

20. The non-transitory computer-readable medium of claim 19, wherein inputs to the third stage of the multi-stage machine learning model include baseline functional performance and predicted changes in lean muscle volume for a functional muscle group.

21. A system for generating a report of disease progression, comprising: a data collection module configured to collect magnetic resonance imaging (MRI) data and clinical data from a patient;a staged machine learning model comprising one or more model stages, the model stages comprising at least one of:a first stage model configured to predict regional-level changes in muscle composition based on at least the MRI data;a second stage model configured to predict muscle-level changes in muscle composition based on at least the MRI data; ora third stage model configured to predict changes in functional outcomes based on at least one of the MRI data, outputs from the first stage model, or outputs from the second stage model; andan output module configured to generate a report including a personalized disease progression prediction for the patient based on the staged machine learning model.

22. The system of claim 21, wherein the second stage model is configured to predict muscle-level changes in muscle composition based further on at least one of outputs from the first stage model or clinical data from the patient.-61- QB\175340.00034\100682361.5

Images