Evaluating Performance of a Deep Learning Multilabel Segmentation Model to Quantify Acute and Chronic Brain Lesions at MRI after Stroke and Predict Prognosis.

Journal: Radiology. Artificial intelligence
PMID:

Abstract

Purpose To develop and evaluate a multilabel deep learning network to identify and quantify acute and chronic brain lesions at multisequence MRI after acute ischemic stroke (AIS) and assess relationships between clinical and model-extracted radiologic features of the lesions and patient prognosis. Materials and Methods This retrospective study included patients with AIS from multiple centers, who experienced stroke onset between September 2008 and October 2022 and underwent MRI as well as thrombolytic therapy and/or treatment with antiplatelets or anticoagulants. A SegResNet-based deep learning model was developed to segment core infarcts and white matter hyperintensity (WMH) burdens on diffusion-weighted and fluid-attenuated inversion recovery images. The model was trained, validated, and tested with manual labels (260, 60, and 40 patients in each dataset, respectively). Radiologic features extracted from the model, including regional infarct size and periventricular and deep WMH volumes and cluster numbers, combined with clinical variables, were used to predict favorable versus unfavorable patient outcomes at 7 days (modified Rankin Scale [mRS] score). Mediation analyses explored associations between radiologic features and AIS outcomes within different treatment groups. Results A total of 1008 patients (mean age, 67.0 years ± 11.8 [SD]; 686 male, 322 female) were included. The training and validation dataset comprised 702 patients with AIS, and the two external testing datasets included 206 and 100 patients, respectively. The prognostic model combining clinical and radiologic features achieved areas under the receiver operating characteristic curve of 0.81 (95% CI: 0.74, 0.88) and 0.77 (95% CI: 0.68, 0.86) for predicting 7-day outcomes in the two external testing datasets, respectively. Mediation analyses revealed that deep WMH in patients treated with thrombolysis had a significant direct effect (17.7%, = .01) and indirect effect (10.7%, = .01) on unfavorable outcomes, as indicated by higher mRS scores, which was not observed in patients treated with antiplatelets and/or anticoagulants. Conclusion The proposed deep learning model quantitatively analyzed radiologic features of acute and chronic brain lesions, and the extracted radiologic features combined with clinical variables predicted short-term AIS outcomes. WMH burden, particularly deep WMH, emerged as a risk factor for poor outcomes in patients treated with thrombolysis. MR-Diffusion Weighted Imaging, Thrombolysis, Head/Neck, Brain/Brain Stem, Stroke, Outcomes Analysis, Segmentation, Prognosis, Supervised Learning, Convolutional Neural Network (CNN), Support Vector Machines © RSNA, 2025.

Authors

  • Tianyu Tang
    Department of Radiology, Zhongda Hospital, School of Medicine, Southeast University, Nanjing, China.
  • Ying Cui
    Department of Medicine Chemistry, Logistics College of Chinese People's Armed Police Forces, Tianjin, 300309, China.
  • Chunqiang Lu
    Department of Radiology, Zhongda Hospital, School of Medicine, Southeast University, Nanjing, China.
  • Huiming Li
    State Key Laboratory of Pollution Control and Resources Reuse, School of the Environment, Nanjing University , Nanjing 210023, China.
  • Jiaying Zhou
    Nurturing Center of Jiangsu Province for State Laboratory of AI Imaging and Interventional Radiology, Department of Radiology, Zhongda Hospital, Medical School of Southeast University, 87 Dingjiaqiao Road, Nanjing 210009, China.
  • Xiaoyu Zhang
    First Department of Infectious Diseases, The First Affiliated Hospital of China Medical University, Shenyang, China.
  • Yujie Zhou
    School of Geographic and Oceanographic Sciences, Nanjing University, Nanjing 210023, China.
  • Ying Zhang
    Department of Nephrology, Nanchong Central Hospital Affiliated to North Sichuan Medical College, Nanchong, China.
  • Yi Zhang
    Department of Thyroid Surgery, China-Japan Union Hospital of Jilin University, Jilin University, Changchun, China.
  • Yuhao Xu
    Key Laboratory of Mechanism Theory and Equipment Design of Ministry of Education, School of Mechanical Engineering, Tianjin University, Tianjin, China.
  • Yuefeng Li
    School of Computer Science, Queensland University of Technology, Brisbane, QLD 4000, Australia.
  • Shenghong Ju