MDL-IWS: Multi-view Deep Learning with Iterative Watershed for Pulmonary Fissure Segmentation

Area of Science:

• Medical Imaging
• Computer-Aided Diagnosis
• Pulmonary Medicine

Background:

• Accurate pulmonary fissure segmentation is crucial for localizing lung lesions within specific lobar territories.
• Precise localization aids in diagnosis and treatment planning for various lung conditions.

Purpose of the Study:

• To propose a novel coarse-to-fine fissure segmentation approach using a Multi-View Deep Learning driven Iterative WaterShed Algorithm (MDL-IWS).
• To enhance the accuracy of fissure segmentation and reduce false positives for improved clinical utility.

Main Methods:

• A Multi-View Deep Learning (MDL) model generates an initial coarse fissure segmentation.
• An Iterative Watershed Algorithm (IWS) refines the segmentation.
• Surface fitting and a weight map are employed within IWS to improve accuracy and reduce over-segmentation.

Main Results:

• The MDL-IWS approach achieves fine segmentation of fissure surfaces.
• Surface fitting significantly reduces false positives, yielding a more accurate fissure volume of interest (VOI).
• Experimental results on the LOLA11 dataset demonstrate superior performance compared to existing state-of-the-art methods.

Conclusions:

• The proposed MDL-IWS method offers an effective solution for pulmonary fissure segmentation.
• This technique shows significant potential for improving lung lesion localization, diagnosis, and treatment planning.
• The approach demonstrates robust performance and outperforms current competitors on a public dataset.