Screening of pulmonary embolism using large language model

The paper is an AI-based early pulmonary embolism (PE) screening system that evaluates structural protein data and analyzes it with the help of large language models (LLMs). PE is a severe disorder as it is difficult to diagnose it early and properly. The classic diagnostic methods may be effective, yet they may be time-consuming and may have a high level of error. In addressing this problem contact maps and detailed embeddings based on SeqVec and Spike2Vec were utilized in examining the protein structures of PDB. Efficient model training was done with dimensionality reduction and Support Vector Machine (SVM) classification. Structural and sequential features were combined to be able to properly classify proteins into PE-positive and PE-negative. The experimental findings indicated that the model had an accuracy of 90.4% and F1-score of 84.6%, precision of 83.2% and recall of 81.8%. This system enhances PE diagnosis and promotes real time clinical decision making. It may result in improved patient outcomes by enhancing faster and reliable screening.