Abstract:
DocAI, an amazing AI-based medical diagnosis system has the capabilities to great extent to determine the diseases with the help of natural language processing (NLP) and image processing. On the other hand, Vision Transformers (ViT), which let clinically significant information to be extracted from various modes of medical imaging like X-rays, MRIs, CT scans etc. System relies on the cutting-edge technology of Vision Transformers (ViT) which allows the unaided interaction of the medical personnel and the individuals with the illness to be managed customized language model under the purview of Transformers BioGPT-Large. This means that the system is in a place to understand and describe further the questions by the doctors regarding the images. The main purpose of DocAI is to relieve the human workload of providing the combination of multi-faced AI models, fast and accurate diagnoses for doctors. With the fusion of the two technologies, DocAI helps not only in the judgement of medical staff, but also the suggestions of an automatic clinical diagnosis to patients. It further allows the doctors to have access to the collection of more than 149k medical images along with doctor's questions and its corresponding answers, thus providing the doctors and patients with diagnostic related information and interpretations based on both the visual and textual content of the images and the questions/answers. The combination of Vision Transformers (ViT) to scrutinize images and BioGPT for handling texts gives DocAI the ability to tap into the evidence based medical literature and offer the medical practitioners the decision support that is necessary and relevant. Such decision support is very helpful in the areas which are deprived of limited and scarce resources through having the least number of trained medical personnel.
From the point of performance perspective, the accuracy of DocAI systems both in image diagnosis and text query interpretation is 87%. We measure precision, accuracy, hit rate and F1 score to determine if the system can meet high standards of performance. The different features of DocAI will revolutionize the medical diagnosis in remote places or under-served areas allowing a quick and reliable diagnosis. This will make better healthcare to the patient and doctors alike. Keywords: Vision Transformer, Biomedical Generative pre-trained transformer, Natural Language processing.
