Predicting Cognitive Decline in Stroke Patients Using Deep Learning DOI Open Access
Sang-Hee Kim

The Open Public Health Journal, Год журнала: 2025, Номер 18(1)

Опубликована: Июнь 2, 2025

Introduction Cognitive decline is a common outcome after stroke, often diminishing survivors’ quality of life. While early detection post-stroke cognitive impairment (PSCI) crucial for intervention, conventional diagnostic methods are time-consuming and resource-intensive. Methods We retrospectively analyzed data from 1,500 stroke patients, whom 450 (30%) developed within six months. A hybrid CNN-LSTM model was used to extract spatial temporal features MRI data. Model performance compared with Random Forest XGBoost, feature importance assessed using SHAP. Results The achieved an accuracy 88.5% AUC 0.92, outperforming (AUC: 0.85) XGBoost 0.87). Key predictors included NIHSS score, age, white matter hyperintensities, hippocampal atrophy. Multimodal integration enhanced predictive performance. Discussion These findings confirm the effectiveness deep learning models in predicting by integrating imaging clinical model’s ability identify structural brain changes key variables offers potential utility detection. However, further validation prospective cohorts needed establish generalizability. Conclusion proposed accurately predicts multimodal inputs. This approach may assist risk stratification individualized care planning. Further prospective, multicenter studies warranted.

Язык: Английский

Predicting Cognitive Decline in Stroke Patients Using Deep Learning DOI Open Access
Sang-Hee Kim

The Open Public Health Journal, Год журнала: 2025, Номер 18(1)

Опубликована: Июнь 2, 2025

Introduction Cognitive decline is a common outcome after stroke, often diminishing survivors’ quality of life. While early detection post-stroke cognitive impairment (PSCI) crucial for intervention, conventional diagnostic methods are time-consuming and resource-intensive. Methods We retrospectively analyzed data from 1,500 stroke patients, whom 450 (30%) developed within six months. A hybrid CNN-LSTM model was used to extract spatial temporal features MRI data. Model performance compared with Random Forest XGBoost, feature importance assessed using SHAP. Results The achieved an accuracy 88.5% AUC 0.92, outperforming (AUC: 0.85) XGBoost 0.87). Key predictors included NIHSS score, age, white matter hyperintensities, hippocampal atrophy. Multimodal integration enhanced predictive performance. Discussion These findings confirm the effectiveness deep learning models in predicting by integrating imaging clinical model’s ability identify structural brain changes key variables offers potential utility detection. However, further validation prospective cohorts needed establish generalizability. Conclusion proposed accurately predicts multimodal inputs. This approach may assist risk stratification individualized care planning. Further prospective, multicenter studies warranted.

Язык: Английский

Процитировано

0