AI Imaging Software Detects Intracranial Hemorrhage

The software takes non-contrast CT images and processes them in the cloud, using proprietary algorithms to note changes and highlight areas so that physicians can immediately see potential areas of bleeding. The enhanced images are sent back to the doctor’s workstation, together with the original.

The algorithms were developed using deep learning, wherein a series of example images are fed to the computer to set benchmarks for what is considered a baseline reading. MedyMatch has secured billions of images from millions of cases via collaborations with hospitals such as Hadassah Medical Center (Jerusalem, Israel) and Massachusetts General Hospital (Boston, MA, USA). Analysis of whole series of images allows the software to learn what a bleed looks like. The process is expected to enable doctors to get an expert opinion within three to five minutes.

“Despite advances in medical imaging, the medical misdiagnosis error rate of around 30 percent rate has not changed for decades. There is a need in the marketplace to provide radiologists and physicians a second set of eyes to help them overcome any limitation that is preventing them from providing a correct patient diagnosis,” said Gabriel Polliack, MD, a member of MedyMatch’s medical advisory board. “The idea is great, and not only has significant clinical value, meaning improvement of patient outcomes, but it will have a direct impact on the cost of care.”

“Non-contrast CT remains the primary imaging modality for the initial evaluation of patients with suspected stroke for traumatic brain injury,” said Gene Saragnese, chairman and CEO of MedyMatch. “MedyMatch is bringing to market a new category of medical solutions that leverages deep learning, machine vision, and the full richness of 3D imaging and other relevant patient data.”

Deep learning is part of a family of machine learning methods based on learning data representations, as opposed to task specific algorithms. It involves artificial neural network (ANN) algorithms that use a cascade of many layers of nonlinear processing units for feature extraction and transformation, with each successive layer using the output from the previous layer as input to form a hierarchical representation.

Related Links:
MedyMatch Technology