Systems and methods for artificial intelligence-based image analysis for cancer assessment

Inventors

Anand, Aseem Undvall • Sjöstrand, Karl Vilhelm • Richter, Jens Filip Andreas

Assignees

Exini Diagnostics AB • Progenies Pharmacenticals Inc • Progenics Pharmaceuticals Inc

Abstract

Presented herein are systems and methods that provide for automated analysis of medical images to determine a predicted disease status (e.g., prostate cancer status) and/or a value corresponding to predicted risk of the disease status for a subject. The approaches described herein leverage artificial intelligence (AI) to analyze intensities of voxels in a functional image, such as a PET image, and determine a risk and/or likelihood that a subject's disease, e.g., cancer, is aggressive. The approaches described herein can provide predictions of whether a subject that presents a localized disease has and/or will develop aggressive disease, such as metastatic cancer. These predictions are generated in a fully automated fashion and can be used alone, or in combination with other cancer diagnostic metrics (e.g., to corroborate predictions and assessments or highlight potential errors). As such, they represent a valuable tool in support of improved cancer diagnosis and treatment.

Core Innovation

The invention determines a predicted disease status and/or a value corresponding to predicted risk of the disease status of a subject using automated analysis of intensities of voxels of a 3D functional image. A 3D anatomical image obtained using an anatomical imaging modality is used to identify a target volume of interest (VOI) corresponding to a prostate region, or a first volume corresponding to a target tissue VOI. The intensities of voxels of the 3D functional image identified as corresponding to the target VOI are used as input to a machine learning module.

The machine learning module receives the functional-image voxel intensities corresponding to the anatomical-image-defined VOI and determines the predicted disease status and/or a value corresponding to predicted risk. In some implementations, the VOI is used to restrict the functional-image information to the prostate region for the automated determination. The predicted disease status and/or predicted risk value is generated based on the automated analysis of the functional voxel intensity patterns within that VOI.

In further embodiments, the machine learning module can also receive one or more clinical variables selected from a defined group, including race/ethnicity; prostate specific antigen (PSA) level and/or velocity; hemoglobin level; lactate dehydrogenase level; albumin level; clinical T stage; biopsy Gleason score; and percentage positive core score. The determination is thus based on the intensities of voxels of the 3D functional image identified as corresponding to the target or first volume of the anatomical image together with the selected clinical variables. Additional refinements include outputting values corresponding to predicted risks and using threshold-based comparison to determine classification.

Claims Coverage

The document includes three independent claims that cover both methods and systems. Across these independent claims, four main inventive features are present: anatomical-to-VOI definition on a 3D anatomical image, aligned extraction of functional voxel intensities within the VOI, use of a machine learning module to output predicted disease status and/or predicted risk, and optional addition of specific clinical variables and threshold-based classification logic.

Overall claim coverage centers on predicting a predicted disease status and/or predicted risk by feeding machine learning with functional-image voxel intensities restricted or aligned to a prostate VOI defined on a 3D anatomical image, with optional incorporation of specified clinical variables as additional machine-learning inputs.

Stated Advantages

Documented Applications