[Supporting system for CT diagnosis referring to previous cases] | JoVE Visualize

[Supporting system for CT diagnosis referring to previous cases]

¹Department of Radiology, Keio University, School of Medicine.

Nihon Igaku Hoshasen Gakkai Zasshi. Nippon Acta Radiologica|December 25, 1993

View abstract on PubMed

Summary

This study introduces a Case-Based Reasoning (CBR) system to aid radiologists in CT image interpretation. The system improves diagnostic accuracy by retrieving similar past cases, preventing misdiagnoses from preconceptions or knowledge gaps.

Area of Science:

Medical Imaging Analysis
Artificial Intelligence in Radiology
Clinical Decision Support Systems

Context:

Radiologists rely on experience and past cases for accurate diagnoses.
Interpreting complex CT scans requires extensive knowledge and can be prone to errors.
Existing diagnostic methods may be limited by individual radiologist experience and potential biases.

Purpose:

To develop and evaluate a Case-Based Reasoning (CBR) system for assisting CT image interpretation.
To enable rapid retrieval and comparison of similar past cases based on translated Japanese reports and CT images.
To assess the system's effectiveness in improving diagnostic accuracy and reducing misdiagnoses.

Summary:

A CBR system was developed to interpret CT images by translating Japanese findings into descriptive units and comparing them with a case base of 1,060 past cases.
The system achieved a 68.1% matching ratio for CT diagnoses with the most similar cases, which improved with a larger case base.
In a clinical study, radiologists using the system achieved a 95.6% success rate in identifying correct diagnoses among the top three matches and a 97.8% overall correct diagnosis rate.

Impact:

The developed CBR system demonstrates significant potential to enhance diagnostic accuracy in CT image interpretation.
It serves as a valuable tool to mitigate misdiagnoses stemming from radiologist preconceptions or limited knowledge.
The system's efficiency, with searches completed in under a second, supports its integration into clinical workflows for real-time diagnostic assistance.