Saliency-CCE: Exploiting colour contextual extractor and saliency-based biomedical image segmentation.

Area of Science:

• Biomedical image analysis
• Computer-aided diagnosis
• Medical imaging

Background:

• Biomedical image segmentation is crucial for computer-aided diagnosis.
• Existing methods often focus on single tasks, neglecting multi-task potential.
• Integrating salient object detection (SOD) with segmentation can enhance performance.

Purpose of the Study:

• To propose a novel dual-task framework, Saliency-CCE, for white blood cell (WBC) and skin lesion (SL) saliency detection and segmentation.
• To enhance the accuracy and efficiency of biomedical image analysis through a unified approach.
• To address limitations of single-task driven segmentation methods.

Main Methods:

• Developed Saliency-CCE, a framework incorporating hair removal preprocessing for skin lesions.
• Introduced a Colour Contextual Extractor (CCE) module with Colour Channel Volume (CCV) and Colour Activation Mapping (CAM) blocks for SOD.
• Implemented an improved Adaptive Threshold (AT) paradigm for automatic segmentation of WBC and SL.
• Utilized ISIC-2016, ISIC-2017, and SCISC datasets for evaluation.

Main Results:

• Saliency-CCE demonstrated superior performance compared to state-of-the-art SOD and segmentation methods.
• The CCE module effectively extracted features for salient object detection.
• The AT strategy enabled accurate automatic segmentation from salient maps.
• The framework successfully integrated dual-task learning for improved biomedical image analysis.

Conclusions:

• The proposed Saliency-CCE framework offers a significant advancement in biomedical image segmentation and salient object detection.
• Dual-task learning, combining SOD and segmentation, proves effective for WBC and SL analysis.
• Saliency-CCE provides a robust and accurate solution for computer-aided diagnosis in dermatology and hematology.