Assessment of endothelial cells and corneas at risk from ophthalmological images
Inventors
Wilson, David L. • Wu, Hao • Joseph, Naomi • Kolluru, Chaitanya • Benetz, Beth • Lass, Jonathan
Assignees
Case Western Reserve University
Publication Number
US-12230397-B2
Publication Date
2025-02-18
Expiration Date
2040-06-22
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Abstract
Embodiments discussed herein facilitate determining a prognosis for keratoplasty based on segmented endothelial cells. One example embodiment comprises a computer-readable medium storing computer-executable instructions that, when executed, cause a processor to perform operations, comprising: accessing an optical microscopy image comprising a set of corneal endothelial cells of a patient of a keratoplasty; segmenting, based at least in part on a first model, a plurality of corneal endothelial cells of the set of corneal endothelial cells; calculating one or more features based on the segmented plurality of corneal endothelial cells; and generating, via a second model trained based at least on the one or more features, a prognosis associated with the keratoplasty.
Core Innovation
The invention provides a system and method for determining a prognosis for keratoplasty based on the analysis of segmented corneal endothelial cells from optical microscopy images. The approach includes accessing an optical microscopy image of a keratoplasty patient's corneal endothelial cells, segmenting the cells using a first model such as a deep learning model, and extracting a range of cell features—including bright pigments, dark nuclei, and thickness of dark gaps between cells—from the segmented data. A second model, trained on these features, generates a prognosis related to the keratoplasty, such as assessing risk of graft rejection or failure.
The problem addressed by this invention is the significant rate of graft failure following keratoplasty (corneal transplantation), including both endothelial keratoplasty (EK) and penetrating keratoplasty (PK) procedures. Traditional evaluation metrics such as endothelial cell density (ECD), coefficient of variation of cell area (CV), and hexagonality (HEX) have shown limitations in predictive power, and current commercial systems often underperform in clinical settings. There is a clear need for improved predictive image analytics capable of identifying corneas at risk of post-surgical rejection or failure, facilitating early intervention and improving patient outcomes.
The core innovation lies in utilizing advanced, highly automated image analytics techniques—including deep learning and machine learning models—to segment endothelial cells and quantify novel and existing cellular features. These include graph-based features to assess cell arrangement, hand-crafted features to capture cell morphology and distribution, and quantitative metrics of structural integrity. The system supports both automated segmentation and semi-automated correction via a graphical user interface, allowing refinement of segmentation where necessary. By integrating these automated analyses with predictive models, the invention enables more accurate, efficient, and routine risk assessment from ophthalmological images.
Claims Coverage
The claims present three main inventive features covering prognosis generation for keratoplasty using segmented cell features, model training for segmentation and prognosis, and interactive correction of segmentation results.
Automated prognosis generation from segmented endothelial cell features
A system implementing: - Accessing an optical microscopy image comprising corneal endothelial cells of a keratoplasty patient. - Segmenting a plurality of corneal endothelial cells using a first model (such as a deep learning model). - Calculating one or more features from the segmented cells, specifically including bright pigments, dark nuclei, and thickness of dark gaps between cells. - Generating, via a second model using the calculated features, a prognosis associated with the keratoplasty at a time post-surgery. Features may further include statistical image measures (e.g., skewness, kurtosis, binned histogram), graph features describing cell arrangement, and traditional metrics such as endothelial cell density, coefficient of variation, and hexagonality.
Model training for segmentation and prognosis based on ground truth images
A method comprising: - Accessing a training set of optical microscopy images, each image associated with ground truth segmentation and known prognoses. - Training a deep learning model (for example, a convolutional neural network) to segment the plurality of corneal endothelial cells in such images. - Calculating features (including bright pigments, dark nuclei, and thickness of dark gaps between cells) from the segmented results. - Training a second model (machine learning or otherwise) to generate prognosis from the calculated features. May include preprocessing of training images and selection of features based on methods such as minimum redundancy maximum relevance.
Interactive correction of segmentation results via graphical user interface
A method or system incorporating: - Receiving from the segmentation model a result which may include segmentation errors. - Highlighting to a user, via a graphical user interface (GUI), those cells deemed to have been segmented incorrectly. - Enabling the user to manually erase or edit cell boundaries corresponding to highlighted cells. This workflow allows improved accuracy of segmentation by integrating user corrections in regions flagged as problematic by the automated model.
The inventive features combine automated deep learning-based image segmentation and feature extraction from endothelial cell images, prognosis generation using machine learning, and the option for interactive correction of segmentation, collectively supporting more accurate and efficient assessment of keratoplasty outcomes.
Stated Advantages
The invention enables early detection of keratoplasties at risk of failure and/or rejection, leading to earlier interventions that can reduce failure and rejection rates.
Saving the initial keratoplasty is important since there is a greater rejection rate in a second keratoplasty, and this invention helps achieve that benefit.
Embodiments reduce healthcare costs, patient discomfort, patient angst, and vision loss by facilitating routine, sophisticated risk analysis of corneal images.
Automated and semi-automated analysis allows for more consistent, objective, and faster evaluation compared to manual methods.
The invention enables analysis of features not easily or reliably perceived by human experts, resulting in more sensitive prediction of graft failure.
Documented Applications
Prediction and prognosis of keratoplasty (corneal transplant) outcomes, including assessment of risk for graft failure or rejection based on corneal endothelial cell images.
Routine, automated or semi-automated analysis of corneal endothelial cell images acquired by specular or confocal microscopy to determine corneas at risk, especially for transplanted tissue.
Training, deploying, and using machine learning or deep learning models for segmentation of corneal endothelial cells and outcome prediction in ophthalmological practices and research.
Assisting clinicians in tailoring patient monitoring and interventions following keratoplasty by providing objective, image-based prognostic information.
Interested in licensing this patent?