Multimodality Data Integration for Improved Skin Cancer Disease Classification and Prediction: An Analysis Using Trained Model on HAM-10000 Metadata & Squamous Cell Carcinoma (SCC) Images

Skin cancer diseases and conditions are common health issues that, if untreated, can sometimes lead to cancer. These problems range from minor skin issues to serious, life-threatening cancers, making it essential to detect and diagnose them accurately. In our study, we suggest a new method that combines different types of skin image data to improve the retrieval accuracy of skin disease classification and help predict when a disease may turn cancerous. Using the HAM_10000 dataset, which includes various skin lesion types, and one hundred high-quality images of SCC, we employ pre-trained deep learning models to identify detailed patterns, designs and features. By combining evidence from these varied causes, our method achieves high accuracy, with a classification rate of 92%. This approach shows promise for enhancing Dermatology and Oncology by offering more accurate and earlier diagnosis, benefiting patient care.