Advancing stroke imaging analysis with interpretable AI and effective connectivity models

Stroke is a leading cause of death and disability worldwide, making early diagnosis and intervention critical. In a recent study published in IEEE Access, our team introduced a groundbreaking end-to-end approach to stroke imaging analysis, combining effective connectivity modeling with interpretable artificial intelligence (AI). This innovation has the potential to transform clinical workflows by enhancing both the accuracy and transparency of stroke diagnoses, highlighting information and flow changes in areas that should be targeted by therapies such as stem cells.

Traditionally, stroke diagnosis relies on imaging modalities such as CT and MRI, alongside clinician expertise. However, these methods face challenges in speed, reproducibility, and the identification of complex patterns in imaging data. Our study addresses these gaps by leveraging effective connectivity models, which analyze the directional influence of one brain region on another, alongside interpretable AI algorithms. Together, these tools not only improve the precision of stroke localization but also shed light on the underlying neural pathways affected by stroke.

We developed an end-to-end framework that processes stroke imaging data using advanced machine learning techniques, such as feature extraction and deep neural networks, while maintaining interpretability. One of the key innovations in our study is the integration of explainability metrics, enabling clinicians to trust and understand the AI’s decision-making process. This feature is crucial for adoption in medical practice, where patient outcomes depend on informed decision-making.

To validate our model, we evaluated it on a large dataset of stroke patients, achieving state-of-the-art performance in identifying stroke regions, predicting patient outcomes, and understanding effective connectivity disruptions. By visualizing these disruptions, our framework provides clinicians with actionable insights previously inaccessible through conventional methods.

The implications of this work are far-reaching. It offers a pathway to personalized treatment plans by identifying stroke subtypes and predicting individual recovery trajectories. Moreover, its reliance on interpretable AI ensures compliance with ethical and legal standards for medical AI systems.

By integrating effective connectivity and interpretable AI, we aim to support clinicians in making faster, more reliable decisions while maintaining transparency in the process. The next steps involve validation on larger cohorts and assessing the usefulness of this approach for stem cell therapies for stroke.

This research represents a significant step forward in the application of AI to medical imaging, particularly for time-sensitive conditions like stroke. By combining cutting-edge technology with a focus on interpretability, our framework has the potential to redefine how strokes are diagnosed and treated in modern health care.

This story is part of Science X Dialog, where researchers can report findings from their published research articles. Visit this page for information about Science X Dialog and how to participate.

Alessandro Crimi received the degree in engineering from the University of Palermo, the Ph.D. degree in machine learning applied for medical imaging from the University of Copenhagen, and the M.B.A. degree in healthcare management from the University of Basel. He was a Postdoctoral Researcher with the French Institute for Research in Computer Science (INRIA), Technical School of Switzerland (ETH-Zurich), Italian Institute for Technology (IIT), and University Hospital of Zurich. He is currently a Professor with the AGH University of
Krakow.