همایش ملی بیوانفورماتیک ایران

صفحه اصلی / 4th international edition and 13th Iranian Conference on Bioinformatics

PTE-MED: AI-based Early Detection of Pulmonary Embolism

نویسندگان :

Toktam Dehghani¹ Maryam Panahi²

1- Department of Medical Biotechnology and Nanotechnology, Faculty of Medicine, Mashhad University of Medical Sciences, Mashhad, Iran 2- Department of Medical Biotechnology and Nanotechnology, Faculty of Medicine, Mashhad University of Medical Sciences, Mashhad, Iran

کلمات کلیدی :

Pulmonary Embolism،Artificial Intelligence،Machine Learning،Emergency Medicine

چکیده :

Timely diagnosis of pulmonary embolism (PE) is a significant challenge in clinical medicine, mainly due to the condition's non-specific symptoms. The PTE-MED artificial intelligence system has been developed to accurately predict the likelihood of PE by analyzing clinical data. Current research indicates that over 50% of suspected PE cases undergoing CT angiography yield negative imaging results (Li et al., 2021). This not only results in unnecessary exposure to contrast agents and radiation but also poses serious risks for vulnerable populations, including patients with renal conditions and pregnant women. The PTE-MED system employs advanced machine-learning algorithms to analyze critical variables such as age, gender, medical history, and clinical symptoms. Imaging results from CT angiography are also incorporated as vital inputs for predictive modeling (Valente Silva et al., 2023). This approach enables PTE-MED to provide early predictions regarding the probability of PE, generating interpretable results for individual patients through AI-driven analytical tools. By supporting healthcare professionals in making informed decisions, PTE-MED has the potential to enhance the management of this complex and urgent medical condition. To improve accessibility for healthcare providers, a mobile application named PTE-MED is being developed. This application will allow physicians and specialists to input patient symptoms and medical history and subsequently receive predictive insights about the likelihood of PE. Preliminary modeling results demonstrate that the CatBoost model achieves an Area Under the Curve (AUC) of 0.768, an accuracy of 71.1%, a precision of 74.0%, a recall of 71.0%, and an F1 score of 72.0%. In conclusion, this system assists healthcare providers in making better-informed treatment decisions by increasing the accuracy of predictions, addressing a key concern for emergency physicians, surgeons, cardiologists, infectious disease specialists, and obstetricians. The PTE-MED artificial intelligence system not only improves diagnostic accuracy but also potentially reduces the financial and temporal burdens associated with unnecessary diagnostic procedures. By implementing this system, healthcare providers can mitigate the risks associated with invasive diagnostic methods and contribute to enhanced public health outcomes.