Multi-object active shape model construction for abdomen segmentation: preliminary results.

Area of Science:

• Medical imaging analysis
• Computational anatomy
• Biomedical engineering

Background:

• Automatic segmentation of abdominal organs is crucial for medical applications.
• Active shape models (ASMs) are common but typically focus on individual organs.
• Exploiting inter-organ relationships can enhance segmentation accuracy.

Purpose of the Study:

• To develop a holistic framework for automatic multi-object ASM construction.
• To enable coupled shape modeling for co-segmentation of abdominal organs.
• To evaluate a novel coupled shape/separate pose approach for liver and spleen segmentation.

Main Methods:

• Introduced a flexible framework for automatic multi-object ASM construction.
• Employed coupled shape modeling for inter-organ relationship exploitation.
• Utilized a new coupled shape/separate pose approach for liver and spleen co-segmentation.

Main Results:

• Achieved feasible segmentation accuracies for abdominal organs.
• Demonstrated that pose decoupling significantly improves segmentation results.
• Observed that the proposed method slightly outperforms standard single-object ASM approaches on average.

Conclusions:

• A holistic, multi-organ ASM approach is effective for abdominal organ segmentation.
• Considering inter-organ relationships and decoupling pose enhances segmentation performance.
• The developed framework offers a promising direction for complex anatomical structure segmentation.