Yanwei Jin 靳颜蔚

I am a PhD student in Bioinformatics and Computational Biology (BICB) at the University of Minnesota, where I go by Yana. My research develops and applies statistical, machine learning, and AI methods to improve clinical decision-making and public health outcomes.

My primary line of research, advised by Dr. Feng Xie (UMN Department of Surgery), focuses on sepsis in the emergency department, covering risk prediction, early identification, and treatment, using data from over 2 million emergency department visits. I also work with Dr. Xiao Zang (UMN School of Public Health) on drug overdose prevention, using simulation modeling and national mortality data.

I received my MS in Biostatistics from the University of Minnesota School of Public Health in 2026, and my Bachelor degree from Peking University in 2022. Before graduate school, I worked at the Hong Kong Centre for Cerebro-cardiovascular Health Engineering (COCHE) on cuff-less blood pressure estimation, advised by Prof. Yuan-Ting Zhang.

News

Publications

My research focuses on E-health, disease modeling, and the application of machine learning and statistical methods to improve healthcare outcomes.

US overdose deaths fell sharply in 2024, driven by a 38.3% drop in fentanyl-involved mortality. Deaths involving stimulants without opioids did not follow: cocaine without opioids rose 1.7% and methamphetamine without opioids rose 2.4%.

Since 2015, non-Hispanic Black adults aged 55 and older have had the fastest rising overdose mortality of any demographic group, from 11.0 per 100,000 in 2015 to 61.1 in 2023, driven largely by cocaine, both alone and co-involved with fentanyl.

Multimodal McBP-Net: hybrid CNN-LSTM for continuous beat-to-beat BP estimation from PPG, ECG, IPG, and skin temperature. Achieves MAE of 4.19 and 2.98 mmHg for SBP/DBP. Published in IEEE JBHI.

MedThink-Bench: 500-question benchmark for assessing LLMs’ medical reasoning with expert rationales. Our LLM-w-Ref framework showed smaller models (MedGemma-27B) can surpass larger proprietary ones (OpenAI-o3).

Model Training Method, Physiological Indicator Detection Method, Apparatus, and Electronic Device

Patent for a model training method, physiological indicator detection method, apparatus, and electronic device. This patent covers innovative approaches for training machine learning models and detecting physiological indicators using wearable devices and sensor technologies. Patent Number: CN119441860A, Application Number: CN202411352734.6, granted by the China National Intellectual Property Administration.

Cold pressor test effects on heart rate and HRV in 22 subjects. Cold exposure significantly increased HR (p<0.001) and decreased HRV.

Deep RNN for continuous BP estimation using PPG and ECG. Achieved MAE of 2.514 and 1.383 mmHg for SBP/DBP on 660 subjects, attaining grade A per BHS standards.

Systematic review of 21 studies (42,855 participants) on public risk perception during COVID-19. Found 51.1% had high-level risk perception.

Qualitative study on medical staff risk perception evolution during public health emergencies. Identified three stages: Vigilance, Observation, and Maintenance.

Modified version of template from here