BIOGPT-THE CHATGPT OF LIFE SCIENCES

Computer Engineering

ABSTRACT Within the medical field, medical images are becoming more and more significant. While an advanced visual question answering system in medicine can help with diagnosis, there is yet no adequate solution to this complex issue. In this research, we present a model, named CGMVQA, which includes classification and response generation capabilities to break down this big problem into several simple problems, taking into account the wide variety of question kinds. We use tokenization for texts and data augmentation for photos. To handle texts, we combine three different types of embeddings with the pre-trained ResNet152 to extract features from images. To reduce the computational cost, we lower the settings of the multi-head selfattention transformer. To modify the model's functionalities, we make adjustments to the masking and output layers. In the ImageCLEF 2019 VQA-Med data set, this model achieves new state-of-the-art performance with 0.640 classification accuracy, 0.659 word matching, and 0.678 semantic similarity. It implies that the CGMVQA can help physicians more effectively with clinical analysis and diagnosis and is useful in responding to medical visual questions. BioGPT is a comprehensive and vital tool for researchers, instructors, and other life science workers. BioGPT is ready to give you quick, accurate, and data-driven insights into any complexity related to molecular biology, genomic analysis, methods for genetic engineering, or bioinformatics problems. As we enter a new era marked by extraordinary biological discoveries and advances in technology, BioGPT acts as a catalyst, promoting greater comprehension of life's mysteries and accelerated biological research accomplishments.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.