Cancer mapping using machine learning
Inventors
Natarajan, Shyam • Priester, Alan Martin • Shubert, Joshua Dylan • Bong, Jeremy Tet Kong • Pusey, Brittany Nan
Assignees
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-12711620-B2
Publication Date
2026-08-18
Expiration Date
Abstract
A method for mapping cancer includes inputting data elements from medical images, biopsy, and biopsy pathology labels into a machine learning model estimating the likelihood of clinically significant cancer in a patient and outputting, via the machine learning model, an estimate of the clinically significant cancer likelihood at each voxel of a three-dimensional image.
Core Innovation
The device and methods map cancer by applying a machine learning algorithm or model to data elements from medical images of a patient, optionally including biopsy pathology data or prostate-specific antigen. The system estimates clinically significant cancer likelihood at each voxel of a three-dimensional image. Outputs include a lesion contour representing lesion size or margin and an encapsulation confidence score representing the likelihood that the contour encompasses all clinically significant prostate cancer.
The displayed cancer estimation map overlays a color-coded heat map on the image and includes a curve relating encapsulation confidence score to lesion size. A point on the curve can be visually manipulated to change the represented lesion size and confidence score, altering the lesion contour. Displays further show lesion contours, patient information, tool-placement controls, biopsy core locations, and an interventional instrument relative to the image.
Claims Coverage
The three independent claims are claims 1, 12, and 16, covering three inventive features across a cancer-mapping device and two cancer-mapping methods.
Machine-learning device for cancer mapping
Claim 1 covers a device in which a machine learning algorithm receives data elements from a medical image and produces an estimate of clinically significant cancer likelihood at each voxel of a three-dimensional image, a lesion contour representing lesion size, and an encapsulation confidence score representing the estimated likelihood that the contour encompasses all clinically significant prostate cancer in the patient.
Cancer mapping with an adjustable lesion contour
Claim 12 covers inputting data elements from medical images into a machine learning model and generating voxel-level estimates of clinically significant cancer likelihood in a three-dimensional image, outputting an adjustable lesion contour representing at least one of cancer lesion size or a margin determination, and an encapsulation confidence score representing the probability that all clinically significant prostate cancer is within the specified contour.
Interactive cancer estimation map and confidence curve
Claim 16 covers displaying voxel-level cancer likelihood as a color-coded cancer estimation map overlying an image, including a lesion contour, and a curve representing encapsulation confidence score versus lesion size, wherein visually manipulating a point on the curve changes the represented lesion size and confidence score and alters the lesion contour.
The three independent claims cover machine-learning cancer mapping from medical-image data to generate voxel-level likelihood estimates, lesion contours, and encapsulation confidence scores, as well as an interactive display linking lesion size and confidence score to alter the contour upon manipulation.
Stated Advantages
Supports patient-specific treatment planning.
Aims to reduce unnecessary treatment of healthy tissue.
Documented Applications
Mapping clinically significant cancer likelihood in a patient from medical-image data, including voxel-level likelihood estimates, lesion-contour presentation, and encapsulation confidence.
Displaying cancer estimation maps, lesion contours, biopsy core locations, and interventional instrument position relative to prostate imaging in procedure-planning displays.
Prostate MRI review displaying anatomical segmentations, region-of-interest and biopsy-core overlays, a cancer estimation map, and an adjustable lesion contour with confidence and lesion-size information.
Interventional-tool placement review and presentation of a downloadable fusion-biopsy navigation plan.
Supporting patient-specific treatment planning.
Interested in licensing this patent?