Accurate multilevel thresholding image segmentation via oppositional Snake Optimization algorithm: Real cases with...

Area of Science:

• Medical Imaging Analysis
• Artificial Intelligence in Medicine

Background:

• Liver diseases are a major global health concern, necessitating accurate diagnostic tools.
• Computer-aided diagnosis (CAD) systems require precise liver segmentation from CT scans for effective treatment.
• Challenges in liver segmentation include inconsistent organ presence and ambiguous boundaries.

Purpose of the Study:

• To develop an enhanced Snake Optimization (SO) algorithm integrated with opposition-based learning (OBL), named SO-OBL, for improved liver disease segmentation.
• To evaluate the performance of the SO-OBL algorithm in global optimization and multilevel image segmentation tasks.
• To create an advanced liver segmentation model for computer-aided diagnosis (CAD) systems.

Main Methods:

• An enhanced Snake Optimization (SO) algorithm incorporating opposition-based learning (OBL) was developed (SO-OBL).
• The SO-OBL algorithm was benchmarked against eleven state-of-the-art metaheuristic algorithms using CEC'2022 test functions.
• A liver disease segmentation model was constructed using the SO-OBL algorithm and an optimized multilevel thresholding technique (Otsu's function).

Main Results:

• The SO-OBL algorithm demonstrated superior performance in global optimization compared to existing metaheuristic algorithms.
• The liver segmentation model achieved high accuracy with FSIM = 0.947, SSIM = 0.941, and PSNR = 24.876.
• The model exhibited high efficiency with a short execution time of 0.281 seconds.

Conclusions:

• The proposed SO-OBL algorithm effectively addresses challenges in liver segmentation from CT scans.
• The developed segmentation model shows significant potential for accurate and efficient diagnosis in computer-aided diagnosis (CAD) systems.
• This research contributes to advancing medical image analysis for liver disease detection.