Keywords
90-day mRS
Composite deep learning
ADASYN
5-fold cross-validation
Submitted : 2026-01-13
Accepted : 2026-01-28
Published : 2026-02-12
Abstract
To counteract small sample size, severe class imbalance and high feature redundancy in 90-day mRS prediction after stroke, this study proposes a four-stage pipeline— “ADASYN re-sampling → clinical + statistical feature screening → dimensionality reduction → 5-fold cross-validation” —and benchmark composite deep-learning architectures. ADASYN first balances the minority classes in the original feature space. Next, a tri-level filter (clinical domain knowledge, variance threshold, mutual information) removes clinically meaningless or redundant variables, after which PCA compresses the remaining features while preserving critical neurological signatures (e.g., brain-herniation history). Four hybrid CNN–RNN models are trained and compared under strict 5-fold cross-validation; the optimal ensemble yields stable, clinically interpretable probabilities that can support individualized rehabilitation planning.
References
Gao L, 2016, Expert Consensus on Integrated Chinese and Western Medicine Diagnosis and Treatment of Hypertensive Intracerebral Hemorrhage in the Acute Phase. Chinese General Practice, 19(30): 3641–3648.
Jiang B, Chen C, Tan H, et al., 2024, Correlation Analysis of Common Blood Biochemical Indicators with Neurological Function and Prognosis After Intravenous Alteplase Thrombolysis in Patients with Acute Cerebral Infarction. Chinese Journal of Geriatric Heart Brain and Vessel Diseases, 26(1): 63–66.
Obermeyer Z, Emanuel E, 2016, Predicting the Future: Big Data, Machine Learning, and Clinical Medicine. New England Journal of Medicine, 375(13): 1216–1219.
Wang L, 2023, Development and Validation of an Early Warning Model for New-Onset Stroke in Chinese Populations at High Risk of Stroke, thesis, China Academy of Chinese Medical Sciences.
Sun Z, Ji J, Ma Z, et al., 2023, Construction and Application of Stroke TCM Pattern Differentiation Model Based on Machine Learning. Journal of Hunan University of Chinese Medicine, 43(4): 694–699.
Cetinoglu Y, Koska I, Uluc M, et al., 2021, Detection and Vascular Territorial Classification of Stroke on Diffusion-Weighted MRI by Deep Learning. European Journal of Radiology, 145: 8.
Liu T, Fan W, Wu C, 2019, A Hybrid Machine Learning Approach to Cerebral Stroke Prediction Based on Imbalanced Medical Dataset. Artificial Intelligence in Medicine, 101: 101723.
Zhu A, Zhu M, Li T, et al., 2025, Prognosis Prediction and Key-Factor Exploration for Patients with Hemorrhagic Stroke. Journal of Qilu University of Technology, 39(2): 61–71.
Mao T, Li Y, Li M, et al., 2025, Review on Deep-Learning-Based Lesion Segmentation and Detection in Acute Ischemic Stroke. Computer Systems and Applications, 34(1): 11–25.
Wang H, Guo J, 2024, Establishment of a Machine-Learning-Based Clinical Intelligent Diagnosis and Prediction Model for Hemorrhagic Stroke. Heilongjiang Science, 15(10): 129–132.
Meng Y, Xu J, Qiu K, et al., 2022, Application of Deep Learning in Ischemic Stroke. Zhejiang Clinical Medical Journal, 24(12): 1892–1894.
Xu P, Fan C, Zhu R, et al., 2021, Prediction of Hospital Inpatients Based on Combined Prophet-LSTM-PSO Model. Journal of University of Shanghai for Science and Technology, 43(1): 68–72.
Liu B, 2024, Assisted Clinical Decision-Making for Post-Stroke Depression Based on Long Short-Term Memory Networks and Its Implementation, thesis, University of Electronic Science and Technology of China.
Li C, Yang P, Feng C, et al., 2025, Abnormal Gait Recognition in Stroke Patients Using Deep-Learning Fusion. Journal of Medical Biomechanics, 40(4): 955–962.
Zhang Y, Li F, Wang W, et al., 2025, Deep Extreme Learning Machine Model for Imbalanced Datasets. Journal of Chongqing University of Technology (Natural Science), 39(6): 108–116.