Building Confidence in Cancer Diagnosis
Early diagnosis is the first step in successful cancer treatment and medical imaging has become intrinsic to the diagnostic process.
Professor Gustavo Carneiro’s research team has been using AI to boost diagnostic screening for potentially deadly bowel and rectal cancer.
Polyps are abnormal growths that look like small bumps in the colon; some are benign but others may grow into cancer and need to be biopsied or removed.
Some are already cancerous and will need surgery, chemotherapy or radiotherapy to stop the spread.
It’s very difficult to determine during an examination which kind are present and only a few experts are able to do that.
Gustavo’s research team used more than 1000 images of polyps found during colonoscopies to teach an AI program to identify differences and report level of certainty around the identification process.
Identifying and treating bowel cancer is quicker and more effective when doctors can diagnose it during a colonoscopy, rather than having to wait for further tests.
Real-time diagnosis assists the endoscopist to make decisions: leave it alone, biopsy or remove.
This will fast track treatment, decrease patient risk and reduce costs thanks to faster treatment and a reduction in colonoscopy related complications.
“Our target is that this system will be as accurate as an experienced endoscopist in the detection and classification of polyps,” Gustavo says.
The research has been very positive, offering the hope of moving the computer-assisted diagnostic system to clinical trials in the next few years.
Translating research findings into use by medical experts will be a big step forward for machine learning in health.
Story provided by Professor Gustavo Carneiro.