Artificial Intelligence Is Widely Used in Medical Diagnostics

Artificial Intelligence is being increasingly used for tasks such as – just to name a few – 

improving the speed and accuracy of diagnosing diseases, developing drugs and new treatments, helping practitioners, tracking timely prescriptions, and planning treatment procedures along with advising patients on a healthy lifestyle.

Duke University computer engineers and radiologists have developed an Artificial Intelligence platform to identify potentially cancerous lesions on mammograms to determine whether a patient should undergo an invasive biopsy. Unlike many of its predecessors, this algorithm is interpretable, meaning it shows doctors exactly how it has come to its conclusions. Researchers have trained AI to detect and evaluate lesions in the same way that a real radiologist has been trained to do.

The solution is designed to help physicians make more informed decisions in healthcare and radiology. “If a computer can help make important medical decisions, doctors need to be sure that the AI ​​is basing its findings on something reasonable,” said Joseph Lo, Professor of Radiology at Duke University. “We need algorithms that not only work, but also explain what they do and show examples of what they base their conclusions on. Thus, regardless of whether the doctor agrees with the result or not, AI helps to make better decisions.”

"Our idea was to create a system that would report that this particular part of a potential cancerous lesion is very similar to another one that doctors had previously discovered," said Alina Barnett, Ph.D., and the first author of the study. “Without these explicit details, practitioners will lose time and faith in the system if there is no way to understand why it sometimes makes mistakes.”

The researchers trained the new AI using 1,136 images taken from 484 patients in the Duke University Health System. First, they taught the AI ​​to find suspicious lesions and ignore all healthy tissue and other irrelevant data. They then hired radiologists to painstakingly label the images and train the AI ​​to focus on the edges of lesions where potential tumors meet healthy surrounding tissue and compare those edges to edges on images with known cancerous and benign outcomes. Radial lines or fuzzy edges, known medically as mass boundaries, are the best predictor of breast cancer and are the first thing radiologists look for. This is because cancer cells replicate and multiply so rapidly that not all edges of a developing tumor are easily seen on mammograms.

“This is a unique way to teach AI how to look at medical images,” Barnett said. “Other AIs are not trying to emulate radiologists; they come up with their own methods of answering a question, which are often useless or, in some cases, dependent on faulty reasoning processes.” Moving forward, the team is working on adding other physical characteristics that the AI ​​needs to consider when making decisions. The volume of investments in Artificial Intelligence (AI) in the field of medicine is growing from year to year, even despite COVID restrictions.

There are also promising products in Kazakhstan, which have already been rightfully appreciated by experts from all over the world. One such project is Cerebra – an Artificial Intelligence in the field of neuroradiology for automated diagnosis of ischemic and hemorrhagic stroke. The technology makes it possible to see the slightest signs of a stroke on a non-contrast CT image to differentiate the patient's condition within this pathology, as well as to determine the possible ways to apply proper treatment. Already today the product is used in several health centers in Kazakhstan. The project has attracted more than $1 million from specialized investors and is ready to be scaled up to foreign markets. Forus Data's centralized medical archive of PACS radiological images was presented by the City Council of Almaty at the Digital Almaty forum in 2021. Artificial intelligence views about 300 x-ray images per day and identifies the 14 most common symptoms of lung diseases with an accuracy of 84%. Moreover, processing and saving the result occur automatically, without the participation of a doctor.

Last year, on October 29th, the startup was named "The Innovative Discovery of the Year" within the largest IT forum in the "Digital Bridge" nomination. The project also won the 1st place in the "Best in Medicine" category in Rome. The company's value as of the end of 2021 is $9 million. The project is a unique solution for managers, doctors, and patients. Self-learning Artificial Intelligence analyzes the treatment protocols filled out by doctors, determines the most frequently encountered protocol, and compares it with international standards. The system then offers to automatically fill in the anamnesis, which significantly reduces the time of admission of the patient. Interest in the project has already been shown by venture capital companies from near and far abroad, including the United States: Startup Health, Ycombinator, and Morgan Stanley Innovation Lab.