HIP: Homotopic Individual Brain Parcellation with Dynamic Threshold Region Growing

The human brain is organized into distinct regions based on properties like cytoarchitecture and topography. Mapping these regions into atlas is essential for understanding the organization and function of the brain. However, group-level brain atlases fail to account for individual variability, while individual-level parcellations typically lack interhemispheric homotopic correspondence, limiting their applicability in individualized prediction and lateralization analyses. Here, we propose a homotopic individual brain parcellation (HIP) with dynamic threshold region growing. HIP leverages the reference atlas to guide the initial seed selection and incorporates a homotopic weight term to enhance interhemispheric correspondence. The dynamic threshold ensures smoother parcel boundaries while preserving high functional homogeneity. Results show that brain atlases generated by HIP are highly reproducible and outperform other individual-level parcellations in terms of functional homogeneity and clustering performance across three cortical atlases and one subcortical atlas. The ablation experiments validate the effectiveness and contribution of the homotopic weight and dynamic threshold. Overall, HIP offers a reliable and precise approach to capturing individual variability while maintaining interhemispheric homotopic correspondence.Clinical Relevance- HIP generates individualized brain atlases that capture individual variability and interhemispheric homotopic correspondence, providing valuable support for personalized brain imaging analysis and individualized clinical diagnosis.