A Hybrid Preprocessor DE-ABC for Efficient Skin-Lesion Segmentation with Improved Contrast.

Area of Science:

• Medical Imaging
• Computer Vision
• Artificial Intelligence

Background:

• Autonomous algorithms are crucial for medical imaging tasks like segmentation and classification.
• Model performance heavily relies on the quality of imaging datasets.
• Existing deep learning and machine learning techniques still leave room for improvement in image analysis.

Purpose of the Study:

• To introduce a novel hybrid meta-heuristic preprocessor, DE-ABC, for optimizing contrast enhancement in medical images.
• To evaluate the efficiency of the proposed DE-ABC preprocessor in improving segmentation tasks.
• To validate the preprocessor's performance on publicly available skin lesion datasets.

Main Methods:

• Developed a hybrid meta-heuristic preprocessor (DE-ABC) to optimize contrast enhancement transformations.
• Applied the DE-ABC preprocessor to publicly available skin lesion datasets (PH2, ISIC-2016, ISIC-2017, ISIC-2018).
• Validated performance using Jaccard and Dice coefficients against state-of-the-art segmentation algorithms.

Main Results:

• The DE-ABC preprocessor demonstrated improved segmentation performance.
• The Dice coefficient saw a maximum improvement from 93.56% to 94.09% after contrast stretching.
• Cross-comparisons confirmed that DE-ABC-enhanced datasets lead to more efficient segmentation algorithms.

Conclusions:

• The proposed DE-ABC preprocessor effectively enhances medical image quality for segmentation tasks.
• Optimized contrast stretching using DE-ABC improves the accuracy and efficiency of autonomous segmentation models.
• This research highlights the significant impact of preprocessing techniques on medical image analysis performance.