Publications by authors named "Niels Turley"
Purpose: Population level tracking of post-stroke functional outcomes is critical to guide interventions that reduce the burden of stroke-related disability. However, functional outcomes are often missing or documented in unstructured notes. We developed a natural language processing (NLP) model that reads electronic health records (EHR) notes to automatically determine the modified Rankin Scale (mRS).
View Article and Find Full Text PDFArticle Synopsis
- Unstructured and structured data in electronic health records (EHR) can provide valuable insights for research, but extracting this information can be challenging; researchers introduced an automated model to identify patients with Alzheimer's Disease, related dementias (ADRD), and mild cognitive impairment (MCI).
- The study involved a sample of 3,626 outpatient adults, using medical notes and diagnoses from chart reviews to develop a logistic regression model that predicts MCI/ADRD diagnoses with high performance metrics.
- The model demonstrated impressive accuracy (99.88%) and other metrics (like AUROC of 0.98), showing that automated EHR phenotyping could effectively facilitate large-scale research on MCI/ADRD.
Article Synopsis
- Atrial fibrillation (AF) is often undetected due to its asymptomatic nature, presenting a significant risk for stroke and heart failure, making early prediction and management essential.
- The study focused on analyzing 18,782 single-lead ECG recordings from 13,609 patients undergoing polysomnography (PSG) to identify individuals at high risk for developing AF, using both hand-crafted features and deep learning methods for prediction.
- By employing advanced feature extraction techniques, the researchers aimed to enhance AF detection using PSG data, ultimately improving patient outcomes through early intervention.
Article Synopsis
- Atrial fibrillation (AF) is often unnoticed but poses significant risks for stroke and heart failure, making early detection and management vital, especially since many AF patients also suffer from obstructive sleep apnea (OSA).
- The study analyzed over 18,000 ECG recordings from patients at Massachusetts General Hospital to find indicators of AF by leveraging data from standard sleep assessments that included ECG monitoring.
- A deep learning approach was used to enhance the prediction model, extracting features from the ECG data to forecast individuals who are at high risk of developing AF in the future.