JOIN OUR WHATSAPP CHANNEL
Scientists develop new AI model that identifies diseases; explains each diagnosis with a visual map
Scientists have created a new AI model that creates visual maps to diagnose illnesses, claim AI can be helpful in scenarios of doctor scarcity.
A new artificial intelligence (AI) model has been created by scientists that uses a visual map to describe each diagnosis to reliably identify tumors and disorders in medical photos.
How does the new AI model work?
According to the researchers, the model's distinct transparency makes it simple for physicians to follow the logic of the model, verify its accuracy twice, and explain the findings to patients. It was published in the journal IEEE Transactions on Medical Imaging. According to Sourya Sengupta, a graduate student at the US's Beckman Institute for Advanced Science and Technology, "The idea is to help catch cancer and disease in its earliest stages -- like an X on a map -- and understand how the decision was made."
Sengupta said, "In many developing countries, doctors are scarce, and a long line of patients. AI can be helpful in these scenarios." According to Sengupta, Automated medical image screening can be used as a helpful tool when time and talent are scarce; it should not be used to replace a doctor's training and experience.
Alternatively, an AI model can pre-scan medical photos and highlight those that have something uncommon for a clinician to analyze, such as a tumor or an early disease symptom known as a biomarker. In addition to saving time, this technique can enhance the performance of the individual assigned to read the scan.
These models perform wonderfully, but they fall short when a patient queries an AI system about why, for example, it identified a tumor in an image or did not. More than 20,000 images were used in three distinct disease diagnosis tasks by the researchers to train their algorithm.
Initially, the model studied mock mammograms and trained to identify early cancer warning signals. Secondly, it examined retinal optical coherence tomography (OCT) pictures, practicing the identification of a buildup known as Drusen, which could be a precursor to macular degeneration.
OCT is a non-invasive imaging procedure that creates cross-sectional images of the retina using light waves. Third, the model learned to identify cardiomegaly, a condition characterized by an enlarged heart that may indicate illness, by examining chest X-rays. Following training, the researchers evaluated the mapmaking model's performance against that of other AI systems—those lacking a self-interpretation option.
Accuracy rates
With accuracy scores of 77.8 percent for mammograms, 99.1 percent for retinal OCT pictures, and 83 percent for chest X-rays, the model outperformed its competitors in all three categories, according to the researchers. They said that the AI's deep neural network, whose non-linear layers imitate the subtleties of human neurons in decision-making, is the source of these remarkable accuracy rates.
ALSO READ: Union Government Conducts Research On Online Gaming, Frames Guidlines To Prevent Impulsive Behaviour
Latest Articles
