New Geometric Deep Learning Model for Detecting Stroke Lesions

Ischemic stroke, which occurs when a blood vessel in the brain gets blocked by a clot, is among the leading causes of death worldwide. Fortunately, surgeons now have access to advanced imaging techniques that allow them to visualize the interior of a patient's brain during a stroke. This helps them pinpoint the location of the clot and analyze the extent of damage to the brain tissue.

Computed tomography-perfusion (CT-P) is one of the most useful imaging modalities in the early stages of an acute stroke. However, it is challenging to accurately identify segmentation - the outline of stroke lesions - in a CT-P scan, and the final diagnosis depends greatly on the surgeon's expertise and ability. To address this issue, scientists have come up with various machine learning models that perform automatic segmentation of CT-P scans. Unfortunately, none of them has reached a level of performance suitable for clinical applications.

Against this backdrop, a team of researchers from Germany recently developed a new segmentation algorithm for stroke lesions. As reported in their study published in the Journal of Medical Imaging, the team built a geometric deep learning model called "Graph Fully-Convolutional Network" (GFCN). The internal operations performed by their geometric algorithm differ fundamentally from those of the more widely used Euclidean models. In their study, the researchers explored the benefits and limitations of this alternative approach.

A key advantage of the proposed model is that it can better learn and preserve important features inherent to brain topology. By using a graph-based neural network, the algorithm can detect complex inter-pixel relationships from different angles. This, in turn, enables it to detect stroke lesions more accurately.

In addition, the team adopted “pooling” and “unpooling” blocks in their network structure. Put simply, the pooling operations, also called "downsampling," reduce the overall size of the feature maps extracted by the network from input images. This reduces the computational complexity of the algorithm, enabling the model to extract the most salient features of the CT-P scans. In contrast, the unpooling operations (or "upsampling") revert the pooling operations to help properly localize the detected features in the original image based on contextual cues. By combining these two operations, the network structure can extract richer geometric information.

The team conducted a series of analyses to determine the effect of each component of GFCN on its segmentation performance. They then compared the performance of the proposed algorithm against the state-of-the-art models, all trained using the same public dataset. Interestingly, although their model used basic unpooling techniques and a simple input configuration, it performed better than the conventional models under most conditions.

Notably, GFCN-8s, with three pooling layers and eight-fold upsampling, achieved a Dice coefficient score - a metric indicating the overlap between the predicted and actual lesion areas - of 0.4553, which is significantly higher than other models. Moreover, the proposed model could adapt to irregular segmentation boundaries better than the state-of-the-art models.

Overall, the findings of this study showcase the potential of geometric deep learning for segmentation problems in medical imaging. Further research on similar strategies could pave the way for highly accurate models for automatic stroke diagnosis that could improve patient outcomes and save lives.

Iporre-Rivas A, Saur D, Rohr K, Scheuermann G, Gillmann C.
Stroke-GFCN: ischemic stroke lesion prediction with a fully convolutional graph network.
J Med Imaging (Bellingham). 2023 Jul;10(4):044502. doi: 10.1117/1.JMI.10.4.044502

Most Popular Now

Collective Intelligence can Help Reduce …

An estimated 250,000 people die from preventable medical errors in the U.S. each year. Many of these errors originate during the diagnostic process. A powerful way to increase diagnostic accuracy...

New Study Suggests ECG-AI can Detect Car…

Artificial intelligence (AI) from patient electrocardiograms (ECGs) may be an innovative solution to enhance heart disease risk assessment. Atherosclerotic cardiovascular disease - arteries narrowed or blocked by the accumulation of...

Software Created from 'Building Blo…

New 'building-block' approaches to the creation of digital tools which include data and artificial intelligence (AI) could play a key role in improving the running of hospital wards and disease...

How could Technology Better Support Pati…

The NHS exists to serve patients. But more could be done to make their experience a key focus when it comes to technology adoption, senior NHS delegates told a recent...

"Showtime" for Digital Health …

13 - 16 November 2023, Düsseldorf, Germany. A hundred start-ups and more than 120 high-calibre professional speakers: These are just the "naked" facts which this year's MEDICA CONNECTED HEALTHCARE FORUM will...

Artificial Intelligence: Unexpected Resu…

Artificial intelligence (AI) is on the rise. Until now, AI applications generally have "black box" character: How AI arrives at its results remains hidden. Prof. Dr. Jürgen Bajorath, a cheminformatics...

Philips Program Developing AI-Powered Ul…

Royal Philips (NYSE: PHG, AEX: PHIA), a global leader in health technology, today announced it has received a second round of funding from the Bill & Melinda Gates Foundation to...

CGM Continues to Drive Digitization in H…

CompuGroup Medical SE & Co. KGaA (CGM), one of the world's leading e-health providers, successfully progressed the digitization in healthcare during the first three quarters in 2023. CGM supports physicians...

Wolverhampton's New 10-Year EPR Dea…

The Royal Wolverhampton NHS Trust (RWT) has just signed a 10-year contract with System C for an integrated electronic patient record (EPR) system, which will replace the trust's in-house built...

Printed Robots with Bones, Ligaments, an…

3D printing is advancing rapidly, and the range of materials that can be used has expanded considerably. While the technology was previously limited to fast-curing plastics, it has now been...

Orchestrating the New World of AI in Hea…

Orion Health's UK and Ireland Customer Conference 2023 focused on the future potential and immediate, practical application of AI to healthcare - and gave delegates a first look at the...

Researchers Take New AI Approach to Anal…

Researchers at Karolinska Institutet and SciLifeLab in Sweden have combined artificial intelligence (AI) techniques used in satellite imaging and community ecology to interpret large amounts of data from tumour tissue...