AmygdalaGo-BOLT for boundary-aware segmentation of the human amygdala

Clinical Snapshot

Key Findings

• P Value: Not reported in the abstract

• Effect Size: Described qualitatively as 'strong agreement' and 'substantially improving efficiency and accuracy' — no quantitative effect size reported in the abstract

• Primary Outcome: Agreement between AmygdalaGo-BOLT automated segmentation and expert manual annotations of the amygdala in pediatric MRI, assessed across multiple imaging centres and age groups

• Nnt Or Sensitivity: No sensitivity, specificity, Dice similarity coefficient, Hausdorff distance, or equivalent segmentation accuracy metrics reported in the abstract; full-text review required for quantitative assessment

• Confidence Interval: Not reported in the abstract

Clinical Application

Deep learning inference is computationally feasible in research neuroimaging pipelines with GPU access. Integration into existing neuroimaging workflows (e.g., alongside FreeSurfer or FSL) would require software packaging, documentation, and institutional IT support. The model's compact encoder-decoder architecture suggests reasonable computational requirements for inference. Clinical deployment would require regulatory consideration and prospective validation. In Australia, amygdala segmentation is primarily relevant to research contexts (e.g., paediatric anxiety, epilepsy, autism spectrum disorder studies) and specialised clinical neuroimaging at tertiary centres. The TGA does not currently regulate research-use neuroimaging segmentation tools, but clinical diagnostic applications would require regulatory pathway consideration. The RACGP and relevant specialist colleges (RANZCP, AES) have no current guidelines specifically addressing automated amygdala segmentation. Australian research groups participating in large paediatric neuroimaging cohorts (e.g., through the Australian Imaging Biomarkers and Lifestyle study infrastructure or paediatric epilepsy networks) may find this tool relevant, but would need to validate performance on Australian scanner configurations and demographic populations before adoption. PBS implications are not directly applicable at this stage. Pediatric and adolescent populations (approximate age range: preschool through adolescence based on MeSH terms) undergoing structural brain MRI in research or clinical neuroimaging contexts; potentially applicable to adult populations with further validation

Abstract

Tracing the boundaries of the amygdala from brain images remains a major challenge in human neuroscience. Although large-scale neuroimaging studies increasingly collect thousands of scans to investigate structural development in children and adolescents, reliable segmentation of the amygdala is difficult due to its small size and complex morphology-particularly in pediatric populations. To address this, we developed AmygdalaGo-BOLT, a boundary-aware deep learning model specifically designed for amygdala segmentation. The model was trained and validated on 1,086 manually labeled pediatric MRI scans, with independent datasets used to assess generalizability. It integrates multiscale feature extraction, spatial priors, and self-attention mechanisms within a compact encoder-decoder architecture to enhance boundary detection. Across imaging centers and age groups, AmygdalaGo-BOLT demonstrates strong agreement with expert manual annotations, while substantially improving efficiency and accuracy relative to existing tools. This enables robust and scalable analysis of amygdala morphology in population neuroscience studies where manual tracing is impractical.

References

1. 1.Zhou, Q., Dong, B., Gao, P., Wei, J., Xiao, J., Wang, W., Liang, P., Lin, D., Lu, J., Zuo, X.-N., & He, H. (2026). AmygdalaGo-BOLT for boundary-aware segmentation of the human amygdala. Cell Reports Methods. https://doi.org/10.1016/j.crmeth.2026.101473