Summary:
UCLA researchers in the Department of Electrical and Computer Engineering have developed an AI-guided, smart histopathological image viewer to aid clinicians in the identification of regions of interest.
Background:
Digital pathology has transformed traditional glass histology slides into high-resolution digital slide images, priming the field of pathology for a machine intelligence-based revolution. Artificial Intelligence (AI) has become increasingly capable of data-driven decision making, including assisting clinicians in making medical diagnoses. However, for higher-stake tasks like tumor detection and rare disease diagnosis, AI is not yet ready to replace human pathologists as any erroneous diagnosis is potentially a matter of life and death. Pathologists today still must manually examine histopathological images, which is a laborious process because each patient’s case often consists of multiple ultra-high-resolution images. Ultimately, a successful AI pathology tool needs to bolster a practicing physician’s repertoire, rather than attempt to replace it.
Innovation:
UCLA researchers have developed a smart histopathological image viewer called NaviPath. NaviPath, guided by artificial intelligence, recommends regions of interest likely to contain tumor cells on a digital histology slide for a pathologist to further investigate. As a pathologist proceeds through his or her standard histology examination workflow, the viewer displays regions of interest overlaid onto the image, as well as the criteria that were used to identify the regions; criteria that are weighted by the examining pathologist. The pathologist can then examine the proposed region in depth as the NaviPath viewer displays the factors that went into its decision matrix. The displayed decision-making removes blind trust from the physician’s point of view. Even with the AI’s guidance, the final medical decision is made by the pathologist, ensuring software accountability. This tool leverages AI’s medical diagnostic capabilities to inform medical professionals, saving both time and effort in a clinical setting.
Potential Applications:
● Tumor detection and grading
● Rare disease identification
● Drug response predictions
● Education and medical learning tool
Advantages:
● Reports multiple AI-computed pathology criteria.
● Presents traceable evidence for each report.
● Allows pathologists to perform diagnoses in routine practice workflows.
● Establishes symbiotic relationship between AI and clinician
State of Development:
The inventors have developed a software, NaviPath, which has been technically evaluated and tested by 12 clinical pathologists.
Related Papers:
1. Gu, H., Liang, Y., Xu, Y., Williams, C.K., Magaki, S., Khanlou, N., Vinters, H., Chen, Z., Ni, S., Yang, C. and Yan, W., 2023. Improving workflow integration with XPath: Design and evaluation of a human-AI diagnosis system in pathology. ACM Transactions on Computer-Human Interaction, 30(2), pp.1-37.
2. Gu, H., Huang, J., Hung, L. and Chen, X.A., 2021. Lessons learned from designing an AI-enabled diagnosis tool for pathologists. Proceedings of the ACM on Human-Computer Interaction, 5(CSCW1), pp.1-25.
Reference:
UCLA Case No. 2023-103
Lead Inventor:
Xiang "Anthony" Chen, UCLA Professor of Electrical and Computer Engineering