Cancer is the leading cause of mortality worldwide, making early and accurate detection a critical priority. Traditional detection methods, such as biopsies and laboratory tests, are often invasive, time-consuming, and prone to human error, creating a need for faster, more efficient, and precise diagnostic approaches. This study addresses these challenges by utilizing AI-driven cancer detection, specifically leveraging deep learning models to classify seven cancer types: brain, oral, breast, kidney, acute lymphocytic leukemia (ALL), lung and colon, and cervical cancer.
Model Performance Comparison
The research focuses on histopathological image analysis, employing advanced preprocessing techniques such as image segmentation, noise removal, and contour feature extraction to enhance tumor identification. Ten transfer learning models, including DenseNet121, InceptionV3, MobileNetV2, and ResNet152V2, were evaluated based on various performance metrics. Among them, DenseNet121 achieved the highest accuracy (99.94%), with minimal loss and the lowest Root Mean Square Error (RMSE), demonstrating its effectiveness in multi-cancer detection.
AI for Early Cancer Detection
By integrating AI and deep learning techniques, this study provides a non-invasive, highly accurate, and efficient alternative to traditional cancer detection. AI-driven models can analyze vast datasets, detect subtle patterns, and assist in early diagnosis, which is crucial in reducing cancer-related mortality and improving patient outcomes. The results highlight the potential of deep learning in medical imaging, offering a scalable solution for early detection, classification, and diagnosis of multiple cancer types.
Reference : https://www.nature.com/articles/s41598-024-75876-2