研究目的
To segment the lesion and identify melanoma from dermoscopy images for accurate early detection of skin cancer.
研究成果
The proposed system effectively segments and classifies melanoma using image processing, with Otsu method performing better for protruded lesions and morphological operations for most cases but with some misclassifications. Future work should involve more features and neural networks for better accuracy.
研究不足
In Otsu method, melanoma images with lesions blended with skin are not correctly classified; morphological operations misclassify some non-melanoma as melanoma. Accuracy could be improved by extracting more features and using trained neural networks.
1：Experimental Design and Method Selection:
The study uses image processing algorithms including enhancement, segmentation via Otsu thresholding and morphological operations, feature extraction based on ABCD rule, and classification using Total Dermatoscopy Score (TDS).
2：Sample Selection and Data Sources:
170 dermoscopy images from the Dermatology database of Med-Node, consisting of 70 melanoma and 100 non-melanoma images from the University Medical Centre Groningen (UMCG).
3：List of Experimental Equipment and Materials:
No specific equipment or materials are mentioned; the work is computational using image processing techniques.
4：Experimental Procedures and Operational Workflow:
Steps include image enhancement (conversion to grayscale, bottom-hat filtering for hair removal), segmentation (Otsu thresholding and morphological operations to isolate lesions), feature extraction (calculating asymmetry, border irregularity, color, diameter), and classification (TDS calculation).
5：Data Analysis Methods:
Accuracy is calculated as (Correctly Classified Images / Total Images) * 100%.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
