“The goal is to detect cancer and diseases in their earliest stages, similar to pinpointing an X on a map, and comprehend the decision-making process,” explained Sourya Sengupta, a graduate student at the Beckman Institute for Advanced Science and Technology in the US.
Researchers have created a novel artificial intelligence (AI) model that precisely detects tumors and diseases in medical images while providing a visual map to explain each diagnosis.
Describing the model’s distinctive transparency in the journal IEEE Transactions on Medical Imaging, he highlighted that it enables doctors to readily track its reasoning, verify accuracy, and communicate results effectively to patients.
“Our model aims to simplify this process, benefiting both doctors and patients,” remarked Sengupta, the lead author of the study.
The approach to interpreting medical images varies across different regions of the world.
“In numerous developing nations, there’s a shortage of doctors and a high volume of patients. AI can be advantageous in such situations,” noted Sengupta.
Alternatively, an AI model can pre-scan medical images and highlight any anomalies, such as a tumour or an early indication of disease, referred to as a biomarker, for a doctor’s assessment. This approach saves time and can enhance the efficiency of the individual responsible for interpreting the scan.
While these models are effective, they may fall short in addressing inquiries from patients, such as explaining why an AI system flagged an image as either containing or not containing a tumour.
The latest AI model provides self-interpretation with each decision, offering detailed explanations rather than simply categorizing as “tumour versus non-tumour,” according to Sengupta.
The researchers conducted training for their model across three distinct disease diagnosis tasks, utilizing over 20,000 images.
Read More: Click Here
