Related Experiment Video
Updated: Dec 15, 2025
05:41
Author Spotlight: Integrating Ultrasound Imaging with Biochemical Markers for Thyroid Disease Diagnosis
Published on: February 9, 2024
953
Interactive thyroid whole slide image diagnostic system using deep representation
Pingjun Chen1, Xiaoshuang Shi1, Yun Liang1
1J. Crayton Pruitt Family Department of Biomedical Engineering, University of Florida, Gainesville, FL, United States.
Computer Methods and Programs in Biomedicine
|July 8, 2020
Summary
This study introduces an interactive system for diagnosing thyroid frozen sections using whole slide images (WSIs). Pathologist-selected suspicious regions improve diagnostic accuracy and efficiency in computer-aided diagnosis.
Area of Science:
- Digital pathology
- Computer-aided diagnosis
- Thyroid histopathology
Background:
- Whole slide images (WSIs) present computational challenges for automated diagnosis.
- Suspicious regions in thyroid WSIs are identifiable, aiding targeted analysis.
Purpose of the Study:
- To develop an interactive WSI diagnostic system for thyroid frozen sections.
- To leverage pathologist-identified suspicious regions for improved computer-aided diagnosis.
Main Methods:
- Generating feature representations from suspicious regions using deep neural networks.
- Extracting and fusing patch features for comprehensive analysis.
- Evaluating region classification and retrieval using multiple classifiers and hashing methods.
Main Results:
- Achieved 96.1% cross-validated classification accuracy on 345 thyroid frozen sections.
- Obtained a retrieval mean average precision (MAP) of 0.972.
- Demonstrated system potential in practical thyroid frozen section diagnosis.
Conclusions:
- The interactive system reduces interference from irrelevant regions and lowers computational costs.
- Enables fine-grained, precise retrieval of suspicious regions.
- Fosters a collaborative relationship between pathologists and diagnostic systems.
