A transformer-based approach to automate disease prediction from patient descriptions

Deep learning techniques have revolutionized healthcare by enabling automated disease diagnosis and treatment. However, medical diagnosis can still be time-consuming, invasive, and costly. In this paper, we propose a transformer-based approach for disease prediction using textual symptom descriptions, aiming to provide automated and accurate diagnoses. A pre-trained transformer model is applied on a dataset of symptom descriptions and disease labels for fine-tuning. The proposed approach leverages the model’s ability to extract contextual information and long-term dependencies from text, enhancing performance in disease prediction. The dataset, collected from online sources in collaboration with healthcare professionals, includes multiple languages (English, Hindi, and Hinglish). Experimental results demonstrate the effectiveness of the proposed approach, which has the potential to significantly reduce the time and cost of medical diagnosis. In English, DiagnoAI exhibited outstanding performance with a recall of 0.966, while for Hindi, it achieved an accuracy of 96.29 with a recall of 0.972, precision of 0.961, and F1-score of 0.968, demonstrating its robust diagnostic capabilities across multiple languages.