One or more non-transitory computer-readable media storing instructions that, when executed by one or more processors of a system, program the one or more processors of the system to perform operations comprising:
- receiving medical records for a plurality of patients;
- generating a plurality of data representations based on the medical records for individual ones of the patients, each data representation representing at least a therapy performed and a quantified disability status of the individual patient, each of the data representations being generated as an image including a plurality of blocks arranged in rows and columns, each block corresponding to a value of a patient information parameter, and removing, from each data representation, a set of blocks corresponding to most recent patient information prior to using the plurality of data representations as training data for a machine learning model, the patient information parameters including, in addition to therapy performed and the quantified disability status of the individual patient, at least one of disease type, disability type, patient age, patient sex, a time series of quantified disability statuses, a number of physical therapy sessions during a period, a number of occupational therapy sessions per during a period, a number of speech therapy sessions during a period, or a number of specific training sessions during a period;
- grouping the data representations into a plurality of subsets based at least on an amount of improvement in a respective quantified disability status of the individual patient for a respective therapy performed in comparison to a quantified disability status of the individual patient prior to performance of the respective therapy;
- selecting, as at least a portion of the training data, a subset of the data representations for which the amount of improvement in the quantified disability status exceeds a threshold indicating a highest amount of improvement as compared to other subsets of the plurality of subsets;
- training the machine learning model using at least the plurality of the data representations in the selected subset as the training data;
- receiving patient information for a first patient, the patient information including a current quantified disability status of the first patient;
- generating at least a first data representation from the patient information for the first patient by appending at least one combination of therapy plans or options extracted from the data representations selected in an ascending order of a difference from the first data representation in the selected subset;
- performing classification of the first data representation with the data representations in the subset based at least on using the machine learning model to determine whether the first data representation matches at least one of the data representations in the subset within a threshold probability for selecting at least one therapy predicted to provide a highest amount of expected improvement at a highest probability in the current quantified disability status of the first patient, wherein the at least one combination of therapy plans or options includes at least one of physical therapy, occupational therapy, speech therapy, or specific training;
- and
- sending information related to the selected at least one combination of therapy plans or options, including the probability and the expected improvement of quantified disability status, to a computing device to cause, at least in part, the computing device to present a graphical user interface including the selected at least one combination of therapy plans or options.