Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A dual-task mutual learning framework for predicting post-thrombectomy cerebral hemorrhage

Caiwen Jiang, Tianyu Wang, Xiaodan Xing, Mianxin Liu, Guang Yang, Zhongxiang Ding, Dinggang Shen

TL;DR

Ischemic stroke treated with thrombectomy carries a risk of postoperative cerebral hemorrhage, and current monitoring with multiple CT scans within 0-72 hours subjects patients to radiation and may delay detection. The authors present a dual-task mutual learning framework that, using only an initial CT, simultaneously generates a predicted follow-up CT and a prognostic label, facilitated by a transformer-based architecture with self-attention and an innovative interactive attention mechanism. Spatial alignment of initial and follow-up scans is performed to ensure reliable cross-time correlations, and a GradNorm-based loss balances the generation and classification tasks. Across a clinical dataset of 200 cases, the method achieves superior follow-up CT generation quality and prognostic accuracy compared to state-of-the-art baselines, indicating potential to accelerate screening and streamline post-thrombectomy care.

Abstract

Ischemic stroke is a severe condition caused by the blockage of brain blood vessels, and can lead to the death of brain tissue due to oxygen deprivation. Thrombectomy has become a common treatment choice for ischemic stroke due to its immediate effectiveness. But, it carries the risk of postoperative cerebral hemorrhage. Clinically, multiple CT scans within 0-72 hours post-surgery are used to monitor for hemorrhage. However, this approach exposes radiation dose to patients, and may delay the detection of cerebral hemorrhage. To address this dilemma, we propose a novel prediction framework for measuring postoperative cerebral hemorrhage using only the patient's initial CT scan. Specifically, we introduce a dual-task mutual learning framework to takes the initial CT scan as input and simultaneously estimates both the follow-up CT scan and prognostic label to predict the occurrence of postoperative cerebral hemorrhage. Our proposed framework incorporates two attention mechanisms, i.e., self-attention and interactive attention. Specifically, the self-attention mechanism allows the model to focus more on high-density areas in the image, which are critical for diagnosis (i.e., potential hemorrhage areas). The interactive attention mechanism further models the dependencies between the interrelated generation and classification tasks, enabling both tasks to perform better than the case when conducted individually. Validated on clinical data, our method can generate follow-up CT scans better than state-of-the-art methods, and achieves an accuracy of 86.37% in predicting follow-up prognostic labels. Thus, our work thus contributes to the timely screening of post-thrombectomy cerebral hemorrhage, and could significantly reform the clinical process of thrombectomy and other similar operations related to stroke.

A dual-task mutual learning framework for predicting post-thrombectomy cerebral hemorrhage

TL;DR

Ischemic stroke treated with thrombectomy carries a risk of postoperative cerebral hemorrhage, and current monitoring with multiple CT scans within 0-72 hours subjects patients to radiation and may delay detection. The authors present a dual-task mutual learning framework that, using only an initial CT, simultaneously generates a predicted follow-up CT and a prognostic label, facilitated by a transformer-based architecture with self-attention and an innovative interactive attention mechanism. Spatial alignment of initial and follow-up scans is performed to ensure reliable cross-time correlations, and a GradNorm-based loss balances the generation and classification tasks. Across a clinical dataset of 200 cases, the method achieves superior follow-up CT generation quality and prognostic accuracy compared to state-of-the-art baselines, indicating potential to accelerate screening and streamline post-thrombectomy care.

Abstract

Ischemic stroke is a severe condition caused by the blockage of brain blood vessels, and can lead to the death of brain tissue due to oxygen deprivation. Thrombectomy has become a common treatment choice for ischemic stroke due to its immediate effectiveness. But, it carries the risk of postoperative cerebral hemorrhage. Clinically, multiple CT scans within 0-72 hours post-surgery are used to monitor for hemorrhage. However, this approach exposes radiation dose to patients, and may delay the detection of cerebral hemorrhage. To address this dilemma, we propose a novel prediction framework for measuring postoperative cerebral hemorrhage using only the patient's initial CT scan. Specifically, we introduce a dual-task mutual learning framework to takes the initial CT scan as input and simultaneously estimates both the follow-up CT scan and prognostic label to predict the occurrence of postoperative cerebral hemorrhage. Our proposed framework incorporates two attention mechanisms, i.e., self-attention and interactive attention. Specifically, the self-attention mechanism allows the model to focus more on high-density areas in the image, which are critical for diagnosis (i.e., potential hemorrhage areas). The interactive attention mechanism further models the dependencies between the interrelated generation and classification tasks, enabling both tasks to perform better than the case when conducted individually. Validated on clinical data, our method can generate follow-up CT scans better than state-of-the-art methods, and achieves an accuracy of 86.37% in predicting follow-up prognostic labels. Thus, our work thus contributes to the timely screening of post-thrombectomy cerebral hemorrhage, and could significantly reform the clinical process of thrombectomy and other similar operations related to stroke.
Paper Structure (12 sections, 1 equation, 3 figures, 2 tables)

This paper contains 12 sections, 1 equation, 3 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Method
  3. Spatial Alignment of Image Pairs
  4. Model Architecture
  5. Self-Attention and Interactive Attention
  6. Experiments
  7. Dataset and Implementation
  8. Ablation Analysis
  9. Comparison with State-of-the-art Methods
  10. Conclusion
  11. Acknowledgments.
  12. Declaration of Competing Interest.

Figures (3)

  • Figure 1: Overview of our proposed dual-task interactive learning framework.
  • Figure 2: Left: Data preprocessing workflow for obtaining the spatially-aligned initial and follow-up brain images. Right: Details of the two attention mechanisms (i.e., self-attention, and interactive attention) involved in the proposed framework.
  • Figure 3: Visual comparison of follow-up scans produced by five different methods. From left to right are the input (initial scan), results by five other comparison methods (2nd-5th columns) and our method (6th column), and the ground truth (GT, i.e., the follow-up scan). The corresponding difference maps between the generated results and GT are shown in the 2nd and 4th rows, where darker color indicates larger differences. Red dotted boxes show the areas for detailed comparison.