ஜர்னல் ஆஃப் கிளினிக்கல் & எக்ஸ்பெரிமென்டல் டெர்மட்டாலஜி ரிசர்ச்
திறந்த அணுகல்

சுருக்கம்

Deep Learning and Rule-Based Hybrid Approach to Improve the Accuracy of Early Detection of Skin Cancer

Varin Senthil*

Melanoma is the most serious type of skin cancer cell in the world. As per the American Cancer Society, there are 99,780 new melanoma patients diagnosed in 2022 of which 7,650 are fatal. Melanoma can develop anywhere on the body and is often a mole that changes size, shape or color. Melanoma shows color variations and an irregular border which are melanoma warning signs. Early diagnosis of the melanoma potentially increasing the chances of cure before cancer spreads. Recent development in Artificial Intelligence (AI) has wide applications in healthcare. Machine Learning (ML) is a subfield of AI has significant development in computer vision, through which statistical models and algorithms can be formed and continuously learn from data to improve perform of a desired task. However, there are possibilities for error and it depend on a lot of factors: E.g. number of samples, quality of the pictures, computer environment, and the machine learning may impact the algorithms. While, in the deep learning approach the knowledge is developed based on sample picture, there are rule based traditional approach which has a business logic built in the approach. Rules could be based on test results, skin colors, etc. This paper presents a comprehensive hybrid approach combining the machine learning methodology along with the color pigment analysis to improve the accuracy and a discussion on how it can be implemented in the field of melanoma diagnosis. We reviewed the latest research and key discoveries in computer vision machine learning, limitations, color pigment-based melanoma detection.