Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

End-to-end Stroke imaging analysis, using reservoir computing-based effective connectivity, and interpretable Artificial intelligence

Wojciech Ciezobka, Joan Falco-Roget, Cemal Koba, Alessandro Crimi

TL;DR

The study tackles the diagnostic challenge of stroke by transforming MRI-derived signals into directed, interpretable connectivity representations. It introduces an end-to-end pipeline that uses reservoir computing to estimate effective connectivity between 100 brain regions, converts these into directed graphs, and classifies them with graph CNNs while leveraging LIME-based explanations. The approach yields approximate AUC values around $0.69$ on a heterogeneous acute stroke dataset and identifies brain networks (e.g., dorsal/ventral attention) that drive classification, providing mechanistic insight into stroke-related network disruption. This framework enhances clinical interpretability and suggests a path toward robust, explainable stroke biomarkers that could generalize to other brain disorders.

Abstract

In this paper, we propose a reservoir computing-based and directed graph analysis pipeline. The goal of this pipeline is to define an efficient brain representation for connectivity in stroke data derived from magnetic resonance imaging. Ultimately, this representation is used within a directed graph convolutional architecture and investigated with explainable artificial intelligence (AI) tools. Stroke is one of the leading causes of mortality and morbidity worldwide, and it demands precise diagnostic tools for timely intervention and improved patient outcomes. Neuroimaging data, with their rich structural and functional information, provide a fertile ground for biomarker discovery. However, the complexity and variability of information flow in the brain requires advanced analysis, especially if we consider the case of disrupted networks as those given by the brain connectome of stroke patients. To address the needs given by this complex scenario we proposed an end-to-end pipeline. This pipeline begins with reservoir computing causality, to define effective connectivity of the brain. This allows directed graph network representations which have not been fully investigated so far by graph convolutional network classifiers. Indeed, the pipeline subsequently incorporates a classification module to categorize the effective connectivity (directed graphs) of brain networks of patients versus matched healthy control. The classification led to an area under the curve of 0.69 with the given heterogeneous dataset. Thanks to explainable tools, an interpretation of disrupted networks across the brain networks was possible. This elucidates the effective connectivity biomarker's contribution to stroke classification, fostering insights into disease mechanisms and treatment responses.

End-to-end Stroke imaging analysis, using reservoir computing-based effective connectivity, and interpretable Artificial intelligence

TL;DR

The study tackles the diagnostic challenge of stroke by transforming MRI-derived signals into directed, interpretable connectivity representations. It introduces an end-to-end pipeline that uses reservoir computing to estimate effective connectivity between 100 brain regions, converts these into directed graphs, and classifies them with graph CNNs while leveraging LIME-based explanations. The approach yields approximate AUC values around on a heterogeneous acute stroke dataset and identifies brain networks (e.g., dorsal/ventral attention) that drive classification, providing mechanistic insight into stroke-related network disruption. This framework enhances clinical interpretability and suggests a path toward robust, explainable stroke biomarkers that could generalize to other brain disorders.

Abstract

In this paper, we propose a reservoir computing-based and directed graph analysis pipeline. The goal of this pipeline is to define an efficient brain representation for connectivity in stroke data derived from magnetic resonance imaging. Ultimately, this representation is used within a directed graph convolutional architecture and investigated with explainable artificial intelligence (AI) tools. Stroke is one of the leading causes of mortality and morbidity worldwide, and it demands precise diagnostic tools for timely intervention and improved patient outcomes. Neuroimaging data, with their rich structural and functional information, provide a fertile ground for biomarker discovery. However, the complexity and variability of information flow in the brain requires advanced analysis, especially if we consider the case of disrupted networks as those given by the brain connectome of stroke patients. To address the needs given by this complex scenario we proposed an end-to-end pipeline. This pipeline begins with reservoir computing causality, to define effective connectivity of the brain. This allows directed graph network representations which have not been fully investigated so far by graph convolutional network classifiers. Indeed, the pipeline subsequently incorporates a classification module to categorize the effective connectivity (directed graphs) of brain networks of patients versus matched healthy control. The classification led to an area under the curve of 0.69 with the given heterogeneous dataset. Thanks to explainable tools, an interpretation of disrupted networks across the brain networks was possible. This elucidates the effective connectivity biomarker's contribution to stroke classification, fostering insights into disease mechanisms and treatment responses.
Paper Structure (12 sections, 11 equations, 6 figures, 4 tables)

This paper contains 12 sections, 11 equations, 6 figures, 4 tables.

Table of Contents

  1. Methods
  2. Data and preprocessing
  3. Reservoir computing
  4. Reservoir computing networks to map causal interactions in lesioned brains
  5. Graph convolutional neural networks
  6. Local Topology Profile
  7. Local Interpretable Machine-Agnostic Explanations
  8. Results
  9. Effective connectivity maps derived from Reservoir Computing
  10. Classification results
  11. Node and edge importance in stroke detection
  12. Discussion

Figures (6)

  • Figure 1: Overall pipeline of the study, where MRI data are preprocessed, used to define an effective connectivity representation, classified and the results are investigated by explainable AI tools.
  • Figure 2: Predictability scores from an the same chaotic system defined in ccm-originalwholebrainRCC. Solid lines show the predictability in Eq. \ref{['eq:prediction-skill']} between embeddings. Shaded regions show 1 standard error of the mean. Transparent lines show the predictability of the surrogate system, which is used to define the expected level of chance against which the hypotheses are tested. In this academic example, it can be said that strong asymmetric interactions between two time series exist at different temporal lags.
  • Figure 3: Working diagram of causality given by the reservoir computing (TOP) and graph convolutional architecture (BOTTOM).
  • Figure 4: Group averaged effective connectivity matrices for two different Times of Repetition. Top: -1 TR. Bottom: -2 TRs. The left column is the average of subjects suffering from a stroke located in the left hemisphere. The middle column is the average of subjects suffering from a stroke located in the right hemisphere. The right column is the average of the control group.
  • Figure 5: Global effective connectivity alterations between regions located in the same hemisphere (top) and between regions located in different hemispheres (bottom). Error bars depict 1 standard error of the mean. Insets show the average difference between left-left and right-right effective connectivity (top) and between left-right and right-left effective connectivity (bottom). Statistical significance was assessed via a two-sample t-test ('*' p<0.05). Global connectivities were obtained by averaging the weight value over the connections belonging to the corresponding hemispheres $H$.
  • ...and 1 more figures