Improved EfficientNetB4 Attention Model for Multi-Disease Detection In Healthcare

As the requirements are continuously growing for disease diagnosis, professionals and researchers are seeking better solutions. An efficient and successful illness diagnosis system is required to fulfill the needs of the present. Machine learning and deep learning are expected to play an important role in future clinical events, such as the diagnosis of disease. Researchers have proposed many single-disease and a few multi-disease diagnosis systems. However, most of the existing multi-disease diagnosis techniques are geared toward using electronic health records instead of image scanning for identification. In health care, the best interventions for the maximization of patient outcomes depend on early and accurate diagnoses of several diseases in a single model. Therefore, a novel and advanced multi-disease diagnostic system, called EfficientNetB4-Improved Channel Attention (ENB4-ICA), was proposed that allows an automated diagnosis, appending deep learning and transfer learning with a channel attention mechanism. This would allow pre-surgical evaluation in healthcare, support yearly medical check-ups, and provide for early intervention, especially in remote and rural regions having limited access to healthcare services. For the implementation and comparative analysis, a dataset of four different diseases, namely eyes, lungs, brain, and skin, comprising 28 classes, was collected. Additionally, the comparison of the proposed channel attention model is conducted with existing state-of-the-art deep learning models. The proposed ENB4-ICA model achieved a training accuracy of 97.75% and a testing accuracy of 93.75%, outperforming the existing state-of-the-art deep learning models. Healthcare providers may be able to save time and improve service quality by using the suggested innovative method.