Interactive Explainable Deep Learning Model for Hepatocellular Carcinoma Diagnosis at Gadoxetic Acid-enhanced MRI: A...

Area of Science:

• Radiology and Medical Imaging
• Artificial Intelligence in Medicine
• Oncology Diagnostics

Background:

• Hepatocellular carcinoma (HCC) diagnosis relies on advanced imaging techniques.
• Gadoxetic acid-enhanced MRI provides detailed liver lesion characterization.
• Improving diagnostic accuracy and efficiency in HCC detection remains a clinical need.

Purpose of the Study:

• To develop and validate an artificial intelligence (AI) model for HCC diagnosis using MRI.
• To assess the AI model's performance in classifying focal liver lesions (FLLs).
• To evaluate the impact of AI assistance on radiologist diagnostic performance.

Main Methods:

• Retrospective analysis of gadoxetic acid-enhanced MRI scans from patients with FLLs.
• Development of a deep learning-based AI model for lesion classification (HCC vs. non-HCC).
• External validation of the AI model and assessment of its influence on radiologist LI-RADS categorization.

Main Results:

• The AI model achieved high diagnostic performance, with an area under the ROC curve of 0.97 in the external testing set.
• AI model demonstrated superior sensitivity (91.6%) compared to LI-RADS category 5 (74.8%) for HCC diagnosis.
• AI assistance significantly improved radiologist sensitivity and accuracy in classifying LR-5 lesions.

Conclusions:

• The developed AI model accurately diagnoses HCC based on MRI findings.
• AI-assisted reading improves radiologists' diagnostic performance in identifying HCC.
• This AI tool shows potential for enhancing clinical decision-making in liver lesion assessment.