Predicting total nucleic acid yield and dissection boundaries for histology slides
Inventors
Yip, Stephen • Ho, Irvin • Sha, Lingdao • Osinski, Boleslaw • Khan, Aly Azeem • Kruger, Andrew J. • Carlson, Michael • Greenwald, Abel • Willis, Caleb • Westley, Andrew • Jones, Ryan • Mahon, Brett
Assignees
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-12165236-B2
Publication Date
2024-12-10
Expiration Date
Abstract
A method for qualifying a specimen prepared on one or more hematoxylin and eosin (H&E) slides by assessing an expected yield of nucleic acids for tumor cells and providing associated unstained slides for subsequent nucleic acid analysis is provided.
Core Innovation
The invention relates to predicting an expected yield of nucleic acid from tumor cells within a dissection boundary on a histology slide. The method receives a digital image of the histology slide at an image-based nucleic acid yield prediction system, identifies tumor cells using a trained cell segmentation model, and defines a dissection boundary corresponding to the identified tumor cells.
The trained cell segmentation model is a pixel-resolution three-dimensional classification model that classifies a cell interior and a cell border. After defining the dissection boundary, the method predicts the expected yield of nucleic acid for the tumor cells within the dissection boundary by providing the digital image to a machine learning model trained on images of a plurality of histology slides having labeled dissection boundaries and labeled total nucleic yield.
The machine learning model is trained on a plurality of histology slides associated with imaging features. The computing device and computer system are configured to implement the same prediction process, including receiving the digital histology image, identifying tumor cells using the trained pixel-resolution three-dimensional classification model, defining the dissection boundary, and predicting the expected nucleic acid yield within that boundary using the machine learning model trained on labeled dissection boundaries and labeled total nucleic yield.
Claims Coverage
The independent claims are clm-00001, clm-00019, and clm-00020. Across these independent claims, the inventive coverage centers on two inventive features: tumor-cell identification using a trained pixel-resolution three-dimensional classification model, and prediction of expected nucleic acid yield within a defined dissection boundary using a machine learning model trained on labeled dissection boundaries and labeled total nucleic yield images from multiple histology slides associated with imaging features.
Pixel-resolution three-dimensional tumor-cell segmentation to define dissection boundary
The method/system identifies tumor cells of the histology slide from a digital image using a trained cell segmentation model, where the trained cell segmentation model is a pixel-resolution three-dimensional classification model trained to classify a cell interior and a cell border, and defines a dissection boundary corresponding to the identified tumor cells.
Machine learning expected nucleic acid yield prediction using labeled dissection boundaries and labeled total nucleic yield
The method/system predicts the expected yield of nucleic acid for the tumor cells within the dissection boundary by providing the digital image to a machine learning model trained on images of a plurality of histology slides having labeled dissection boundaries and labeled total nucleic yield, where the plurality of histology slides are associated with imaging features.
Computing device implementation of image-based nucleic acid yield prediction
A computing device with one or more memories and one or more processors configured to receive a digital image, identify tumor cells using the trained pixel-resolution three-dimensional classification model, define a dissection boundary, and predict expected nucleic acid yield using the machine learning model trained on images with labeled dissection boundaries and labeled total nucleic yield.
Computer system implementation of image-based nucleic acid yield prediction
A computer system with one or more processors configured to receive a digital image, identify tumor cells using the trained pixel-resolution three-dimensional classification model, define a dissection boundary, and predict expected nucleic acid yield using the machine learning model trained on images with labeled dissection boundaries and labeled total nucleic yield.
The independent claims consistently require tumor-cell identification via a pixel-resolution three-dimensional classification model that classifies cell interior and cell border, and expected nucleic acid yield prediction within a defined dissection boundary using a machine learning model trained on labeled dissection boundaries and labeled total nucleic yield from a plurality of histology slides associated with imaging features.
Stated Advantages
Predicts expected nucleic acid yield of tumor cells within a dissection boundary for providing associated unstained slides for sequencing only when summed predicted yield exceeds a predetermined yield threshold.
Uses a predetermined yield threshold tied to specification requirements, including explicit numeric ranges (50 ng–2000 ng), to support accept/reject decisions for slide sufficiency for genetic sequencing.
Documented Applications
H&E histology slide qualification for sequencing by digitizing H&E slides, predicting expected nucleic acid yield in a defined dissection boundary, and providing associated unstained slides for sequencing only when predicted yield exceeds a predetermined threshold.
Integration into pathology workflows, including reporting and overlay maps based on predicted yield and corresponding masks.
Interested in licensing this patent?