OMM-RLC Optimised Multimodal Fusion For Recurrence Lung Cancer Detection

Lung cancer is one of the deadliest cancers, involving abnormal cell changes and uncontrollable growth leading to tumors. It’s a leading cause of cancer-related deaths and can spread throughout the body if untreated. An efficient detection system is crucial, and multimodal fusion techniques offer greater accuracy than single-data-type models. Our approach for diagnosing recurrent lung cancer combines clinical and imaging data from a radiogenomic dataset. We create two unique models specific to their datasets and combine them using a late fusion strategy. This includes data gathering, preprocessing, feature extraction, and identifying recurrent lung cancer. Models are individually trained and then combined, aiming to select the best model based on accuracy and mean absolute error (MAE). Deep Learning is effective for clinical data processing, and CNNs like ResNet50 have shown promising results in image classification over traditional CNNs like VGG-16. The late fusion technique enhances predictive outcomes for multi-source datasets. Using an ensemble of models with an averaging method achieves a $78.04 \%$ accuracy rate, surpassing individual models. Our project focuses on the often-neglected aspect of recurrent lung cancer, emphasizing the need for early detection techniques. By using advanced fusion methods, integrated various data sources and models, improving predictive precision and advancing cancer detection and patient care