Rapid Diagnosis of Stroke-Causing Blockages Using Deep Learning Model

Artificial Intelligence (AI) is changing the way healthcare industry functions. It is helping doctors diagnose patients more accurately. AI makes predictions about patients' future health and recommends better treatment. Henceforth, healthcare organizations of all sizes, types, and specialties are becoming increasingly interested in employing AI for better patient care and improved efficiency.

Medicine is one of the fastest-growing and important application areas with unique challenges. Over a relatively short period of time, the availability of AI has exploded, leaving providers, payers and stakeholders with a variety of tools, technologies and strategies from where they can choose. Understanding how data is ingested, analyzed and returned to end-user can have a big impact on expectations for accuracy and relatability. In order to effectively choose technologies that will help develop algorithms, healthcare organisations should feel confident that they have a firm grasp on the different flavours of artificial intelligence and its use cases.

Deep learning is a good place to start with. Deep learning is a machine learning technology that teaches computers to do what comes naturally to humans. In deep learning, a computer model learns to perform classification tasks directly from images, texts or sound. Deep learning models are trained by using a large set of labelled data and neural architectures that contain many layers. In a recent study published by Radiology, a sophisticated deep learning model helps rapidly detect blockages in the arteries that supply blood to the head, potentially speeding the inset of life-saving treatment.

Deep learning model detects blockages in arteries

Dr Matthew T. Stib, a radiologist resident at the Warren Alpert Medical School at Brown University in Providence, Rhode Island along with his colleagues explored the use of deep learning to help provide rapid detection of large vessels occlusions on CT angiography (CTA) and reduces time to treatment. The team worked under the direction of Ryan A. McTaggart, a neuroradiologist specialised in interventional neuroradiology and proponent of decreasing the time to treatment for large vessel occlusion.

Large vessel occlusions are blockages in the arteries that supply oxygenated blood to the brain. These occlusions account for a significant proportion of ischemic strokes, the most common type of stroke. Prompt diagnosis is critical in order to begin recanalization, or opening of the blocked artery, through a treatment known as endovascular therapy.

Every minute matters while diagnosing blockages as it reduces the time to recanalization which extends the patient's disability-free life by a week. CTA is a three-minute exam that provides detailed views of the blood vessel. It is used to detect occlusions. Radiologists are highly accurate while identifying large vessel occlusions on CTA, but they are not always available. As time matters the most in treatment, a delay could unravel further consequences.

The research team from Brown's computer science department developed a deep learning model from scratch. They used a large sample of CTA examinations for patients with suspected acute ischemic stroke to train the algorithm to recognize the appearance of large vessel occlusions and distinguish it from other conditions. Preprocessing of the CTA exams included the creation of maximum intensity projection images to emphasize the contrast-enhanced vasculature. The researchers also used multiphase CTA, a newer approach that provides more comprehensive information than the single-phase technique.

The deep learning model scored 100% result on multiphase CTA examinations by detecting all 31 large vessel occlusions of 62 patients. It is a statistically significant improvement over 77% sensitivity rate of single-phase CTA. The use of multiphase CTA triggered the improvement in performance.

The study is the first to use multiphase CTA to look at occlusions in both the arteries of the front, or anterior, part of the head and neck and those in the back, or posterior. Posterior circulation occlusions have not been discussed much in machine learning literature. They are less common but have pretty profound clinical consequences if missed. It's important to have an algorithm that detects all categories of occlusion, both anterior and posterior.

The next step in the research is to validate the results using the algorithm in real-time and see if it can improve outcomes for patients. If the results hold up, then the deep learning model could be a useful asset in medical centres or hospitals that don't have the expertise for reading large vessel occlusion CTA images. This algorithm is not replacing the ability of radiologists to do their job; rather, it's trying to speed up the time to diagnosis. So if the radiologist isn't around or there is a large workflow that is preventing someone from looking at the exam results quickly, there will be an alert that says an occlusion may be present and someone should look at this. That's where the value is in this kind of a model.