A Novel Image Segmentation Framework Using Adaptive Region‐Based Unsupervised Machine Learning Mechanism

INTRODUCTION: The segmentation of medical images plays a significant role in the healthcare sector. Accurate lesion segmentation and other pathologies from healthcare images are fraught with many challenges. Solely depending on the specialist's knowledge for the disease analysis is subject to bias and is time-consuming. METHODS: At first, essential medical images are collected from the standard resources. Subsequently, the collected images are used to obtain different image features like"Local Binary Pattern" (LBP), Local Weighting Pattern (LWP), and Red Green Blue (RGB) channel intensity. Next, the extracted features are given to the image segmentation using the Adaptive K-Region-based Clustering (AKRC) technique. Moreover, parameters in AKRC are tuned by using an Enhanced Golf Optimization Algorithm (EGOA). RESULT: The outcomes ensured that the implemented approach achieved 97% accuracy and 95% dice coefficient values. DATA CONCLUSION: Thus, the findings illustrate that the implemented model secures satisfactory solutions than the classical mechanisms. The proposed region-based unsupervised segmentation method enhances segmentation accuracy by integrating rich image features and adaptive clustering. Additionally, the EGOA fine-tunes the AKRC parameters automatically, eliminating manual tuning and improving generalization across different datasets. Together, the use of robust image features and adaptive optimization techniques leads to more accurate segmentation outcomes in complex medical imaging scenarios.