Summary:
Researchers from UCLA's Departments of Electrical and Computer Engineering and Pathology and Laboratory Medicine, along with a researcher from KUMC's Department of Pathology and Laboratory Medicine, have developed an innovative workflow for pathology, integrating artificial intelligence (AI) diagnostics with interactive reporting.
Background:
The digitization of medical data has led to the development of whole slide image viewers, enhancing pathologists' diagnostic efficiency and accuracy. Concurrently, significant progress has been made in creating computer vision models for this domain. However, reluctance to deploy these models on actual patient data stems from their "black box" nature, prioritizing accuracy over interpretability. Traditional slide viewers facilitate annotation and abnormal region detection but do not integrate advanced computer vision models. There is a need to harness computer vision advances in a manner that gains pathologists' trust. This requires integrating models into a slide viewer with findings that can be interacted with, enabling scalable diagnoses and annotations, reducing slide analysis costs and time, and generating more training data to improve existing models.
Innovation:
Dr. Chen and his research team have developed an interactive and hierarchical reporting tool for histopathological image analysis, which combines AI diagnostics optimized for explainability with the pathologist's expertise in detecting abnormal tissue samples. This tool enhances traditional slide viewers with a top-down approach, presenting the diagnosis, critical criteria, samples meeting each criterion's conditions, and flagged abnormal regions within each sample. Users can modify features at each stage, granting ultimate diagnostic control to the pathologist and fostering trust.
Demonstration Video: Improving Workflow Integration with xPath: Design and Evaluation of a Human-AI Diagnosis System in Pathology
Potential Applications:
● Training pathologists to identify regions of interest
● Pathology diagnostics, such as mitosis detection in cancer
● Development of new diagnostics for grading cancer
● Health record outlier detection and case sorting
Advantages:
● Hierarchical design
● Comprehensive diagnosis explanations
● Flexible model decision making
● Collaborative human-AI approach
Development to Date:
A paper on the technology has been published. They conducted a study with pathologists to compare a traditional computational workflow with an AI-assisted workflow on H&E and Ki-67 slides.
Related Papers:
Improving Workflow Integration with xPath: Design and Evaluation of a Human-AI Diagnosis System in Pathology
Reference:
UCLA Case No. 2023-090
Lead Inventor:
Xiang Chen