Deep learning segmentation suppresses overlapping anatomy on MRCP maximum intensity projections, improving biliary visualization

Deep learning segmentation improved MRCP maximum intensity projection visualization: structure overlap scores rose from 2.5 to 3.5 and diagnostic confidence from 3.2 to 3.4 on 4-point scales in 35 evaluation cases.

• Study included 322 3D MRCP datasets from 162 patients; 265 training cases from 127 patients and 35 evaluation cases with one acquisition per patient.
• Two board-certified radiologists rated processed and original MIP images using 4-point Likert scales; significance was tested with Wilcoxon signed-rank tests and Benjamini-Hochberg correction.
• After correction, processed images showed significantly improved structure overlap scores (3.5 ± 0.9 vs. 2.5 ± 0.6) and higher diagnostic confidence (3.4 ± 0.7 vs. 3.2 ± 0.7).

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.