JIITA, vol.9 no.2, p.1098-1103, DOI: 10.22664/ISITA.2025.9.2.1098
Diviya Krishnamoorthy, Radhakrishnan Palanikumar
Abstract. The segmentation of tumor regions in medical images plays a critical role in enhancing diagnostic accuracy, optimizing treatment planning, and tracking the progression or regression of diseases. This paper proposes a novel methodology that integrates K-means clustering with pre-processing and morphological operations to automate tumor segmentation in dermoscopic images. The proposed approach achieved high performance with a Dice Similarity Coefficient (DSC) of 91.4%, Jaccard Index of 84.2%, Sensitivity of 94.1%, and Specificity of 96.8%. These results, though based on assumed ground truth, indicate the approach’s strong potential for automated tumor detection. Future research will focus on validating the methodology using annotated datasets to enhance its applicability in clinical practice.
Keywords; Tumor Detection; Image Preprocessing; Feature Extraction; K-means Clustering; Morphological Operations
Fullpaper: