KEY TAKEAWAYS
- The study aimed to design a multi-modal DNN-based model by multiple imaging and genomic profiling for diagnosis of patients with Melanoma.
- The designed model showed higher accuracy vs. traditional method further validation and clinical studies were suggested.
Advanced stage Melanoma known for its high mortality rates is particularly an aggressive type of skin cancer. Early and accurate detection of melanoma is crucial for improved patient outlooks. The traditional diagnosis was limited to biopsies and was thus not significantly reliable. This approach is designed to explore new and more effective diagnostic methods is essential for better identification and treatment melanoma.
Ajmeera Kiran and the team attempted to utilize the deep neural networks (DNNs) that integrate multiple types of imaging and genomic data to classify melanoma, potentially offering a more reliable diagnosis vs. currently practiced methods.
Researchers built a dataset comprising dermoscopic images, histopathological slides, gene expression profiles, and whole-exome sequencing data of more than 1,000 patients with melanoma and healthy controls. They used convolutional neural networks (CNNs) for image data analysis and Graph Neural Networks (GNNs) for genomic data analysis, trained and evaluated this framework on datasets retrieved from The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO).
The results demonstrated that the developed multi-modal DNN achieved higher accuracy than traditional medical approaches, with a mean accuracy of 92.5% and an AUC of 0.96. This further suggested its potential usage in detecting critical features of melanoma beyond traditional methods. AI and machine learning methods could potentially, offer larger datasets and assist dermatologists to make more accurate decisions and refine treatment strategies. However, larger-scale validation and additional clinical trials are needed to establish the effectiveness and feasibility of this novel diagnostic approach.
The study concluded that the combined DNN-based model integrated with multiple imaging and genomic profiling could potentially offer a better diagnostic tool vs. traditional and less reliable methods, subject to larger-scale validation and additional clinical trials.
The funding information was not provided.
Source: https://pubmed.ncbi.nlm.nih.gov/38881051/
Kiran A., Narayanasamy N., Ramesh J.V.N., et al. (2024). “A novel deep learning framework for accurate melanoma diagnosis integrating imaging and genomic data for improved patient outcomes.” Skin Res Technol. 2024 Jun;30(6):e13770. doi: 10.1111/srt.13770. PMID: 38881051; PMCID: PMC11180689.
