Ensembles improve accuracy and confidence in brain scans for dementia and epilepsy
Evaluation of Uncertainty-Aware Multi-Software Ensembles for Hippocampal Segmentation
Computer Vision and Pattern Recognition
Summary
It can be hard for computers to accurately highlight a tiny part of the brain called the hippocampus, which is important for conditions like Alzheimer's and epilepsy. The authors tested combining five different computer programs to see if working together would do a better job. They found that using three of these programs together gave good results without needing too much computing power. The combined approach also helped spot when the computer might be unsure, so doctors can double-check those tricky cases. This makes it easier to trust the computer results and could help in real medical settings.
What this means in practice
- •For hospital imaging teams: Improve clinical brain scan analysis by highlighting uncertain hippocampal segmentation results for expert review, enhancing diagnostic confidence in dementia and epilepsy cases.
- •For medical device developers: Build more reliable MRI analysis tools by integrating ensemble-based uncertainty measures to flag ambiguous hippocampal segmentations automatically.$Commercial implications: Enables development of advanced imaging software tools aimed at hospitals and diagnostic centers desiring higher accuracy and reliability.
Authors
Gabriel Oliveira-Stahl, Anna Schroder, James Moggridge, Hamza A. Salhab, Caroline Micallef, Josephine Barnes, M. Jorge Cardoso, Carole H. Sudre, Matthew Grech-Sollars
Abstract
Accurate hippocampal segmentation can be a useful tool for diagnosing and monitoring neurological conditions such as Alzheimer's disease and epilepsy. While numerous automated segmentation methods exist, their clinical adoption remains limited. Reliable uncertainty assessment can enhance trust and facilitate clinical translation. This study evaluates ensembles of five heterogeneous hippocampal segmentation methods - InnerEye, ASHS, FastSurfer, HippoSeg, and FreeSurfer - across two dementia datasets and one epilepsy dataset. The sub-ensemble containing InnerEye, FastSurfer, and HippoSeg emerged as both accurate and efficient, highlighting the feasibility of balancing computational cost and performance. Additionally, ensemble-derived uncertainty quantification with sample variance, mutual information, and predictive entropy is shown to reduce inaccurate segmentations by flagging low-confidence cases, potentially providing a mechanism for automatically escalating ambiguous cases for expert assessment. Lastly, this study explores the downstream clinical utility of ensemble-derived uncertainty, finding a significant correlation with motion artefact severity and a modest contribution to Alzheimer's disease classification.