On 21 September, the US Food and Drug Administration (FDA) authorised Paige Prostate, the first artificial intelligence (AI) based software designed to detect areas at risk of being cancerous in prostate biopsy images. This software allows pathologists to make a higher number of accurate diagnoses of cancerous tissues, which could help address the growing number of cancer cases worldwide.
Prostate cancer is a cancer that develops in the prostate, a male gland responsible for producing a fluid that nourishes and transports semen. According to the US Centres for Disease Control and Prevention (CDC), prostate cancer is the most common cancer in men in the US, aside from skin cancers. It is also one of the leading causes of cancer death in men, along with lung cancer. In the US, 13 out of 100 men will get prostate cancer during their lifetime, and two to three of these men will die as a result. Prostate cancer that is detected earlier has a higher likelihood of being treated successfully.
According to the International Agency for Research on Cancer (IARC), there were 17 million new cancer cases worldwide and 9.5 million cancer deaths in 2018. By 2040, the number of new cases is projected to increase by 60% and the number of cancer deaths is projected to increase by 70%. With this increase in diagnostic demand, Paige Prostate will allow pathologists to work more efficiently and confidently and will allow them to focus more on the crucial elements of the diagnostic process.
In the clinical study submitted to the FDA, 16 pathologists were asked to examine 527 prostate biopsy images. Of these 527 samples, 171 were cancerous and the rest were benign. The results of this study showed that when pathologists used Paige Prostate when examining the images, their detection of cancer improved by 7.3% compared to when they did not use the software.
The potential risks of introducing a new test to the diagnosis process include false-negative and false-positive results. Pathologists using Paige Prostate had a 70% reduction in producing false-negative results and a 24% reduction in producing false-positive results. Although these results seem good, it is important that any false-negative or false-positive results be mitigated by limiting the role of the software in the diagnoses, because a false-negative result may result in delayed detection and ultimately more complicated and difficult treatment, while a false-positive can put undue psychological stress on the patient.
AI and machine learning have the potential to revolutionise healthcare. Medical device manufacturers have recently begun to incorporate AI technology into their products to advance healthcare through the ability to learn from real-world data and constantly improve performance, which could be greatly beneficial to healthcare providers and patient care. To manage potential risks caused by these changes, however, the FDA has imposed tight regulations on medical devices. Despite being heavily regulated, GlobalData forecasts that the market for AI-based healthcare will grow from $1.5bn in 2019 to $4.3bn in 2024. As such, AI has already attracted the attention of investors and will continue to do so despite the risks that come with the FDA’s tight regulations.