Fits Health

Fit And Well, Daily

Breaking News
Move Goals

AI improves kidney cancer detection rates

By Indigo Pemberton November 19, 2026
AI improves kidney cancer detection rates - kidney cancer
To evaluate its real-world value, the team organized a crossover trial with 21 pathologists from Asan.

Korean researchers discovered that AI tools can enhance pathologists’ detection of a specific, uncommon kidney tumor, yet doctors who trusted these systems the most were also more likely to be misled by their mistakes. The investigation, carried out at Asan Medical Center in Seoul, created a machine learning tool designed to spot fumarate hydratase-deficient renal cell carcinoma (FHdRCC), an aggressive and infrequent form of kidney cancer.

This AI system, based on the ResNeXt101 framework, was trained and tested using 335 digital pathology slides. It scanned tissue samples for patterns linked to FHdRCC, marked potential cases, and produced color-coded visuals pinpointing areas of concern.

Assessing Clinical Potential

To evaluate its real-world value, the team organized a crossover trial with 21 pathologists from Asan. Each specialist reviewed 30 difficult kidney cancer samples, including 15 FHdRCC cases, first without AI help, then with it, with a two-week gap between evaluations.

When using AI support, participants could consult the system’s predictions and highlighted tissue areas before finalizing their diagnoses. The study also tracked how pathologists reacted to both accurate and flawed AI suggestions, exploring whether these responses correlated with their experience or specialization.

Findings and Implications

The results showed AI assistance boosted diagnostic accuracy from 60% to 73.3% and improved precision from 66.7% to 77.8%. Pathologists’ agreement rates rose, and review times stayed roughly the same. However, the advantages, and risks, of AI assistance varied depending on how doctors engaged with its guidance.

The study revealed two distinct groups: 14 pathologists described as ‘receptive,’ who frequently followed AI advice, and others labeled ‘resistant,’ who approached recommendations more cautiously. The receptive group saw marked accuracy gains with AI but were also more vulnerable to incorrect suggestions.

Despite these differences, both groups achieved similar overall success rates when using AI. The researchers found no link between these tendencies and either years of practice or subspecialty focus, indicating that expertise alone doesn’t guarantee effective AI integration.

Broader Context and Future Directions

FHdRCC represents under 0.4% of kidney cancers but can be highly aggressive, emphasizing the importance of early detection. Its similarities to other kidney tumors often lead to delayed diagnoses, especially when initial suspicions are low and extra tests aren’t ordered. Professor Yong Mee Cho from Asan’s pathology department noted that AI could help doctors spot these rare cancers that might otherwise be missed and flag patients needing further confirmation.

Leave a Reply

Your email address will not be published. Required fields are marked *

© 2026 Fits Health. All rights reserved.