Health and Research
10th March 2025
Rapid diagnosis of shunt failure is essential for timely and effective treatment. Shunt failure can usually be readily detected, but sometimes it can be trickier, [GV1] e.g. if a patient has non-typical shunt failure symptoms, or changes in brain/ventricle compliance, or if the shunt failure is slowly progressing or intermittent. There’s a need to improve the speed and reliability of diagnosis, particularly in less straightforward cases of shunt failure.
A team in Boston (US) has been testing whether AI computer vision can be used to improve shunt failure detection. They trained image analysis algorithms to evaluate past scans from real patients with hydrocephalus. The results were compared with traditional (human) clinical diagnosis. The AI system identified and estimated the volumes of the third and lateral ventricles from the scans and determined whether they were enlarged. In all test cases, the algorithm correctly detected ventricular enlargement (indicating shunt failure), and the need for shunt revision was correctly predicted 92% of the time.
It’s likely that AI will increasingly be integrated into different aspects of healthcare. While improvements still need to be made, this research suggests that shunt failure detection is an application that will be well suited to AI support.