Artificial intelligence, or AI, is a fast-growing area of research in many fields, including medicine. In cancer research, teams are studying its potential to learn about and better diagnose and develop treatments for patients. The early results are promising and show that this technology could provide people with earlier diagnoses and more effective, less damaging treatments.
One of the best hopes a cancer patient has of remission or extending life expectancy is with an early diagnosis. The earlier treatment begins, the better the chances are that a cancer can be cured, sent into remission, or at least significantly slowed in growth.
Cancer diagnosis is fraught with human error. There is rarely a single test that indicates a patient has cancer. Most diagnoses rely on radiologists reading imaging scans to find suspect areas of tissue, followed by a biopsy of cells to be examined by a pathologist. It is easy to both miss tumors and to make false positive diagnoses.
A number of studies have recently shown that AI can do the job of diagnosing cancer better than human oncologists and pathologists. One study that brought together Google and researchers from Northwestern University used a Google AI program trained with over 40,000 CT scans from thousands of patients.
The program used those scans to learn how to diagnose lung cancer. The researchers tested it against several expert diagnostic radiologists. The AI program boosted the diagnosis rate by 5% but also reduced by 11% the number of false positives. A similar study used another algorithm and trained AI to diagnose breast cancer. That system had a 97% success rate.
Getting a better, earlier diagnosis for patients is important, but AI may also be able to improve the effectiveness of treatments. One big area for utilizing AI is in personalizing treatments. Researchers at the Cleveland Clinic have used AI systems to look at patients’ electronic medical records and CT scans to determine a personalized radiation dosage for treatment.
Normally radiation is given as a uniform dose, but there are unique differences between patients and their tumors. The AI strategy of determining individual doses has proven successful in reducing the probability of treatment failure to just 5%.
Researchers at MIT have done similar research but have included chemotherapy. A study that involved patients with glioblastoma, a cancer of the brain and spinal cord, has found success in using AI to reduce the toxicity of treatments. The researchers used a machine-learning technique to adjust doses so that patients get the maximum effect with the lowest potency of drugs.
While any patient may benefit from AI diagnosis and treatment, those with cancers traditionally difficult to diagnose and manage may benefit the most. Mesothelioma is one example. This cancer caused by asbestos exposure is notoriously difficult to diagnose and is often not found until it is in later stages, giving patients few treatment options.
Early diagnosis for this kind of cancer could significantly extend survival time, and that is where AI efforts have focused. Researchers in Scotland along with the Cancer Innovation Challenge have begun using AI in screening CT scans to try to specifically diagnose pleural mesothelioma. Because it has been successful in better diagnosing lung and other types of cancers, there is hope it can find early signs of mesothelioma in people especially at risk, including those exposed to asbestos.
Research in all areas of cancer care is expected to continue using AI. The potential for making earlier diagnoses and for developing better and more effective treatment is vast and is giving patients even with the most difficult cancers hope.
In July 2018 I wrote about the part that point of care ultrasound (POCUS) would play in the future of the imaging market. I’m not sure where two years have gone, but since then the companies, personnel and their technology have evolved considerably.
Aesthetics is a hot bed of innovation and pioneering start-ups are thriving at the heart of this market.