Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Spatiotemporal Disentanglement of Arteriovenous Malformations in Digital Subtraction Angiography

Kathleen Baur, Xin Xiong, Erickson Torio, Rose Du, Parikshit Juvekar, Reuben Dorent, Alexandra Golby, Sarah Frisken, Nazim Haouchine

TL;DR

This work addresses the challenge of disentangling arterial, nidal, and venous vascular phases in Digital Subtraction Angiography (DSA) for AVMs by fusing an unsupervised Independent Component Analysis (ICA) approach with a supervised CNN-based vessel segmentation. The ICA component separates temporal flow phases from DSA sequences, while a U-Net–based network provides precise vascular masks; a runtime fusion yields phase-constrained, color-coded overlays that enhance interpretation. Experimental results on AVM datasets show strong segmentation performance (Dice ≈ 0.944) and demonstrate clinically relevant improvements in workflow, despite some errors in a subset of visualizations. The method offers a practical avenue to improve AVM visualization and treatment planning, with future work aimed at broader user studies and higher-fidelity imaging data.

Abstract

Although Digital Subtraction Angiography (DSA) is the most important imaging for visualizing cerebrovascular anatomy, its interpretation by clinicians remains difficult. This is particularly true when treating arteriovenous malformations (AVMs), where entangled vasculature connecting arteries and veins needs to be carefully identified.The presented method aims to enhance DSA image series by highlighting critical information via automatic classification of vessels using a combination of two learning models: An unsupervised machine learning method based on Independent Component Analysis that decomposes the phases of flow and a convolutional neural network that automatically delineates the vessels in image space. The proposed method was tested on clinical DSA images series and demonstrated efficient differentiation between arteries and veins that provides a viable solution to enhance visualizations for clinical use.

Spatiotemporal Disentanglement of Arteriovenous Malformations in Digital Subtraction Angiography

TL;DR

This work addresses the challenge of disentangling arterial, nidal, and venous vascular phases in Digital Subtraction Angiography (DSA) for AVMs by fusing an unsupervised Independent Component Analysis (ICA) approach with a supervised CNN-based vessel segmentation. The ICA component separates temporal flow phases from DSA sequences, while a U-Net–based network provides precise vascular masks; a runtime fusion yields phase-constrained, color-coded overlays that enhance interpretation. Experimental results on AVM datasets show strong segmentation performance (Dice ≈ 0.944) and demonstrate clinically relevant improvements in workflow, despite some errors in a subset of visualizations. The method offers a practical avenue to improve AVM visualization and treatment planning, with future work aimed at broader user studies and higher-fidelity imaging data.

Abstract

Although Digital Subtraction Angiography (DSA) is the most important imaging for visualizing cerebrovascular anatomy, its interpretation by clinicians remains difficult. This is particularly true when treating arteriovenous malformations (AVMs), where entangled vasculature connecting arteries and veins needs to be carefully identified.The presented method aims to enhance DSA image series by highlighting critical information via automatic classification of vessels using a combination of two learning models: An unsupervised machine learning method based on Independent Component Analysis that decomposes the phases of flow and a convolutional neural network that automatically delineates the vessels in image space. The proposed method was tested on clinical DSA images series and demonstrated efficient differentiation between arteries and veins that provides a viable solution to enhance visualizations for clinical use.
Paper Structure (7 sections, 1 equation, 3 figures)

This paper contains 7 sections, 1 equation, 3 figures.

Table of Contents

  1. METHODS
  2. Temporal Flow Decomposition
  3. Vessel Classification
  4. Spatiotemporal re-composition of DSA Series
  5. RESULTS
  6. CONCLUSION
  7. ACKNOWLEDGEMENT

Figures (3)

  • Figure 1: Overview of applied techniques: In order to separate phases of vascular flow in a DSA image series $\mathbf{x}$ we use the ICA unmixing function $g$ to decompose $\mathbf{x}$ into arterial, venous, and nidal sources $\tilde{\mathbf{s}}$ and the segmentation model $h$ to produce a binary mask that will be used to build the final visualization.
  • Figure 2: The sources $\tilde{\mathbf{s}} = g(\mathbf{x})$ that result from applying ICA decomposition to two DSA image series (A and B) provide information for phase decomposition. The phases can be visualized as two or three distinct images.
  • Figure 3: Examples of phase-constrained color-coded DSA image series (A and B) showing the progressive appearance of classified arteries and veins which feed or drain the nidus. Arteries are shown in red, veins in blue, nidus and capillaries in green.