Methods and systems for determining health of an organ based on medical image data using artificial intelligence

A method for determining health of an organ based on medical image data, comprising:

• generating a foundation model, utilizing one of more processors, by:
• retrieving, from one or more databases, a set of images related to a plurality of organs, the set of images comprising CT scans related to respective thymus;
• generating a set of pre-processed images by applying pre-processing to the set of images, the pre processing comprising standardizing the set of images to an isotropic resolution and adjusting scan intensities for consistency;
• training the foundation model to predict embedding vectors, using artificial intelligence, by;
• producing embedding vectors for each image of the set of pre-processed images by utilizing a deep convolutional neural network architecture;
• and
• refining the embedding vectors by applying a self-supervised learning framework to the embedding vectors associated with the set of pre-processed images;
• generating a task-specific classifying model based, wherein generating the task specific classifying model comprising generating a Quality Control (QC) model and generating a score model, the QC model providing binary predictions with probabilities differentiating between normal thymus degradation and abnormalities, wherein the score model evaluating a level of fat and soft tissue in the thymus, by:
• retrieving, from the one or more databases, a second set of images associated with another plurality of organs and associated medical data, the associated medical data comprising labels indicating health of organs;
• generating, utilizing the artificial intelligence, embedding vectors for each image of the respective second set of images by applying the foundation model,
• generating, utilizing the artificial intelligence, the QC model based on embedding vectors of each respective image of the second set of images and associated medical data;
• and
• generating, utilizing the artificial intelligence, the score model based on embedding vectors of each respective image of a chosen set of the second set of images and associated medical data;
• evaluating health of a potential organ based on medical image data, comprising:
• receiving medical image data of the potential organ, the potential organ comprising a thymus, the medical image data comprising a computed tomography scan;
• and
• generating a finalized image by applying pre-processing, using one or more processors, to the received medical image data;
• generating an embedding vector by applying the foundation model on the finalized image;
• and
• generating a classification output by applying the task-specific classification model on the embedding vector, wherein the task-specific classification model comprises of a Quality Control (QC) model and a score model, wherein generating the classification output comprises generating the classification output based on an inflammation label output by the QC model and based on the level of fat and soft tissue label output by the score model based on the embedding vector;
• outputting the health score based on the generated classification output;
• and
• applying a clinical solution based on the health score, comprising:
• applying chemotherapy and immunotherapy in response to a low health score, the low health score comprising less than 33 percent of a possible highest score of the health score;
• and
• applying immunotherapy in response to a high health score, the high health score comprising greater than 68 percent of a possible highest score of the health score.