DL-Based Automatic Extraction of CSF Spaces Using 3D-T1WI for DESH Identification: Comparison with Evans Index and Callosal Angle

Motivation: To develop an automated detection method for DESH. Goal(s): To clarify the relationship between a automated CSF space analysis method and estabished indicators, mamely Evans index (EI) and corpus callosum angle (CA), assessing its potential as an alternative diagnostic tool. Approach: We performed automated CSF space analysis on 598 3D-T1WI, extracting four CSF regions. Correlations between these indices and EI and CA were analyzed. Results: The EI showed a strong correlation with total ventricular volume (r = 0.78). The CA demonstrated a moderate correlation with total ventricular volume and the index of inferior devided by superior subarachnoid sapace volume (r =-0.51 and -0.62). Impact: Since the proposed automated CSF space volumetry does not require specific section settings or manual adjustments, it offers a viable alternative to conventional indicators for identifying DESH. This approach accelerates the identification and understanding of iNPH.