Archives of clinical neuropsychology : the official journal of the National Academy of Neuropsychologists · 2025
Get Your Brain in the Game: Using Machine Learning to Predict Recovery Timelines Following Sports-Related Concussion.
СтатьяНазвание и аннотация: Перевод в очереди
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Thomas Garrett A, Arnett Peter A
Аннотация
OBJECTIVE: This exploratory proof-of-concept study aimed to develop predictive models for return-to-play (RTP) timelines following sports-related concussion using machine learning techniques.
METHODS: Using data available through Federal Interagency Traumatic Brain Injury Research Informatics System (FITBIR) and the Concussion Assessment, Research and Education (CARE) Consortium, the sample included 971 college athletes with available RTP data. Data were split into training, testing, and validation sets. Random forest (RF) regression modeling was used to predict number of days to RTP based on individual factors, injury data, and symptom and cognitive performance data collected 24-48 hr post-concussion. Recursive feature elimination (RFE) was used to identify the features that were most strongly associated with RTP. We also conducted RF classification modeling with recovery type (typical [≤ 28 days] vs. prolonged [> 28 days]) as the outcome of interest.
RESULTS: RFE revealed optimal performance with 31 features, most of which were related to post-concussion symptomatology and cognitive performance. The RF regression model showed modest performance, accounting for 21% and 17% of the variance in testing and validation sets, respectively. The RF classification models also showed good performance across different levels of dataset balancing. The strongest classification model showed an accuracy of 89.04% with an F1 score of 0.56 on the testing set. On the validation set, the model showed 85.52% accuracy with an F1 score of 0.40. Receiver operator characteristics showed an AUC of 0.85.
CONCLUSIONS: Overall, these findings support the use of machine learning models in concussion management - particularly with predicting recovery timelines.
METHODS: Using data available through Federal Interagency Traumatic Brain Injury Research Informatics System (FITBIR) and the Concussion Assessment, Research and Education (CARE) Consortium, the sample included 971 college athletes with available RTP data. Data were split into training, testing, and validation sets. Random forest (RF) regression modeling was used to predict number of days to RTP based on individual factors, injury data, and symptom and cognitive performance data collected 24-48 hr post-concussion. Recursive feature elimination (RFE) was used to identify the features that were most strongly associated with RTP. We also conducted RF classification modeling with recovery type (typical [≤ 28 days] vs. prolonged [> 28 days]) as the outcome of interest.
RESULTS: RFE revealed optimal performance with 31 features, most of which were related to post-concussion symptomatology and cognitive performance. The RF regression model showed modest performance, accounting for 21% and 17% of the variance in testing and validation sets, respectively. The RF classification models also showed good performance across different levels of dataset balancing. The strongest classification model showed an accuracy of 89.04% with an F1 score of 0.56 on the testing set. On the validation set, the model showed 85.52% accuracy with an F1 score of 0.40. Receiver operator characteristics showed an AUC of 0.85.
CONCLUSIONS: Overall, these findings support the use of machine learning models in concussion management - particularly with predicting recovery timelines.