ARTIFICIAL INTELLIGENCE-DRIVEN 'MACHINE CLASSIFIERS' FOR ACCURATE DIAGNOSIS
As medical technologies advance, personalized medicine is increasingly driving the classification and diagnosis of not only commonly seen, but also never encountered types and subtypes of disease conditions. For pathologists and diagnosticians, this trend has increased the differential diagnostic demands, sample or test processing workflows and subspecialty training requirements.
Dr. Phedias Diamandis, a neuropathologist at the University Health Network (UHN) and an affiliate scientist at the Princess Margaret Cancer Centre is addressing just that.
Leveraging the power of deep convolutional neural networks (CNNs), Dr. Diamandis and his research team are developing diagnostic ‘machine classifiers’ – Artificial Intelligence (AI)–driven tools to enhance the capability of pathologists to quickly and accurately differentiate and diagnose neural cancers, including subtypes.
Although current classification methods that use probability distribution scores and binary approaches to arrive at disease classification are valid when the disease can be classified into fewer categories, Dr. Diamandis’ research is demonstrating how applying more transparent AI/CNN-driven approaches to visualizing and efficiently detecting anomalies can address the pressing ‘real world’ needs of both on-site and remotely-based pathologists.
This innovative approach would add incremental value to any pathologist trying to reach a definite diagnosis, when faced with multiple, misclassified, rarely seen, or even novel, classification possibilities. In turn, pathologists can substantially increase the overall diagnostic efficacy, enhance workflow cost-effectiveness, and ultimately, augment patient safety.
Learn more about licensing and collaboration opportunities around Dr. Phedias Diamandis’ unique technology https://flintbox.com/public/project/50442/.