Related Experiment Video
Updated: Jun 11, 2025
14:08
Automated Midline Shift and Intracranial Pressure Estimation based on Brain CT Images
Published on: April 13, 2013
42.5K
Automated ventricular segmentation and shunt failure detection using convolutional neural networks
Kevin T Huang1,2, Jack McNulty3,4,5, Helweh Hussein4
1Harvard Medical School, 25 Shattuck St, Boston, MA, 02115, USA. khuang@bwh.harvard.edu.
Scientific Reports
|September 28, 2024
Summary
Computer vision algorithms can accurately detect ventriculomegaly, a sign of adult hydrocephalus shunt failure. This technology shows high reliability in predicting the need for shunt revision, improving diagnosis.
Area of Science:
- Neurosurgery
- Medical Imaging
- Artificial Intelligence
Background:
- Adult hydrocephalus is primarily treated with ventricular shunts.
- Shunt malfunction is a frequent complication, posing diagnostic challenges.
Purpose of the Study:
- To assess the feasibility of a computer vision algorithm for automatically detecting ventriculomegaly in adult hydrocephalus patients with shunts.
- To evaluate the algorithm's accuracy in predicting shunt failure.
Main Methods:
- Retrospective analysis of CT scans from 191 adult hydrocephalus patients over eight years.
- Training a machine learning algorithm to identify ventricles and detect ventriculomegaly.
- Comparing algorithmic performance to human reviewers using Dice scores and ventricular volume calculations.
Main Results:
- The algorithm achieved an average Dice score of 0.809 ± 0.094.
- Computer-derived ventricular volumes were not significantly different from human assessments.
- The algorithm correctly identified ventriculomegaly in all test cases, predicting shunt revision needs with 92.3% accuracy.
Conclusions:
- Automated algorithms can reliably and accurately detect ventriculomegaly in adult hydrocephalus shunt malfunction.
- This technology presents a feasible solution for improving the diagnosis of shunt failure.
