CONFERENCE PROCEEDINGS
Advanced Search
Home > Proceedings > Volume 12817 > Article
Presentation
13 March 2024 A temporal correlation-driven AI model powered by large-scale clinical data for endoscopic detection of bladder tumors
Mark Laurie, Md Tauhidul Islam, Okyaz Eminaga, Eugene Shkolyar, Xiao Jia, Timothy Lee, Hubert Lau, Lei Xing, Joseph C. Liao
Author Affiliations +
Proceedings Volume 12817, Advanced Photonics in Urology 2024; 128170H (2024) https://doi.org/10.1117/12.3003179
Event: SPIE BiOS, 2024, San Francisco, California, United States
Abstract
Cystoscopy is a common diagnostic procedure for hematuria evaluation, bladder cancer surveillance, and guidance for resection of suspicious bladder lesions. Similar to other endoscopic procedures, standard white light cystoscopy is limited by diagnostic accuracy, image quality, and operator dependency. Artificial intelligence is emerging as a potential tool for clinical decision support for endoscopy and endoscopic surgery. However, current deep learning methods for cystoscopy are based on highly curated individual images and do not consider the temporal correlation among the frames within a video sequence, thereby limiting their performance in real world clinical settings. Herein we propose SEQ-3D, a sequential deep learning model for cystoscopic video analysis that considers both the short- and long-range relationships among the frames and the spatial information in each frame. Model validation was performed using a benchmark cystoscopy video dataset derived from 60 patients and 163 pathologically confirmed bladder regions of interest (ROI) consisted of representative cancerous bladder tumors and cancer-mimicking benign lesions. The full-length videos (216,870 frames) were annotated by expert clinicians and divided into distinct frame sequences defined as scenes. SEQ-3D outperforms existing sequential deep learning methods and achieved a per-ROI accuracy of 100%, per-scene sensitivity of 93.1%, and per-scene specificity of 83.3%, exhibiting balanced performance at detecting a wide variety of bladder ROI. Our findings are promising for deployment in prospective clinical settings and can be extrapolated to endoscopic applications in other organ systems.
Conference Presentation
(2024) Published by SPIE. Downloading of the abstract is permitted for personal use only.
Citation Download Citation
Mark Laurie, Md Tauhidul Islam, Okyaz Eminaga, Eugene Shkolyar, Xiao Jia, Timothy Lee, Hubert Lau, Lei Xing, and Joseph C. Liao "A temporal correlation-driven AI model powered by large-scale clinical data for endoscopic detection of bladder tumors", Proc. SPIE 12817, Advanced Photonics in Urology 2024, 128170H (13 March 2024); https://doi.org/10.1117/12.3003179
ACCESS THE FULL ARTICLE
ORGANIZATIONAL
Sign in with credentials provided by your organization.
INSTITUTIONAL
Select your institution to access the SPIE Digital Library.
SELECT YOUR INSTITUTION
PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.
PERSONAL SIGN IN
No SPIE Account? Create one
PURCHASE THIS CONTENT
SUBSCRIBE TO DIGITAL LIBRARY
50 downloads per 1-year subscription
Members: $195
Non-members: $335 ADD TO CART
25 downloads per 1 - year subscription
Members: $145
Non-members: $250 ADD TO CART
PURCHASE SINGLE ARTICLE
Includes PDF, HTML & Video, when available
Members: $17.00
Non-members: $21.00 ADD TO CART
PROCEEDINGS
 PRESENTATION
WATCH
PRESENTATION SAVE TO MY LIBRARY
GET CITATION
Advertisement
Advertisement
KEYWORDS
Video processing
Cancer detection
Data modeling
Deep learning
Tumors
Bladder
Tumor growth modeling
RELATED CONTENT
Subscribe to Digital Library
Receive Erratum Email Alert
Back to Top