Diagnose Skin Cancer With Artificial Intelligence And Machine Learning

(CTN News) – Almost everyone diagnosed with skin cancer first goes to primary care. It’s hard for primary care providers to tell benign skin lesions from rarer skin cancers.

In recent weeks, artificial intelligence and machine learning (AI/ML) algorithms have been talked about a lot for diagnosing skin cancer. Owain Jones is a clinical research fellow at Cambridge’s Department of Public Health and Primary Care.

AI/ML algorithms could make a big difference in primary care.

Early diagnosis of skin cancer could improve outcomes for patients and boost survival rates.”

Having low cancer probability lesions reassure patients and reduce the burden on dermatology services.

Jones and a team of Cambridge researchers conducted a study assessing the current evidence for the efficacy and safety of AI/ML algorithms based on this thinking.

Jones said “we knew AI/ML algorithms and technologies to help diagnose skin cancer were emerging, but there wasn’t much evidence about their safety and effectiveness.”

The study investigators carefully considered what kinds of studies to include.

Jones said we couldn’t find any research studies that used AI/ML algorithms to diagnose skin cancer in primary care settings.

We included all studies that came up with AI/ML algorithms that could be used in primary care. The systematic review included a lot of studies, but it also provided a nice overview.”

The review included 272 studies, but no primary care studies, and only 2 used data from populations similar to primary care populations.

It turns out AI/ML algorithms are pretty good at diagnosing skin cancer. Jones noted that there aren’t any real-life clinical studies on AI/ML algorithms and their accuracy.

The main focus of our review was primary care settings, and we found that there is no evidence in settings with low skin cancer rates, so widespread adoption isn’t recommended,” he said.

Also, we questioned whether the datasets used to develop many AI/ML algorithms were sufficiently representative of the general population to prevent biases against minorities.

In Jones’s opinion, the researchers didn’t anticipate that patients from Black and ethnic minority backgrounds would not be included in the datasets used to develop the AI/ML algorithms at the beginning of the review.

“Another surprising outcome was that there wasn’t much implementation research and study in real life clinical settings, despite all the research that has been done in this field in recent years,” he said.

According to this study, AI/ML technologies aimed at diagnosis of skin cancer are probably at a much earlier stage than we expected.

Based on their findings, AI/ML algorithms could help clinicians detect skin cancers in primary care settings more accurately.

Researchers are still working on this area, so there’s a concern about whether the diagnostic performance demonstrated in the included studies would hold true in populations with a lower skin cancer prevalence or in settings with low-quality images.

“These algorithms need to be evaluated carefully to make sure they’re accurate, efficient, cost-effective, and safe enough for clinical use, and that more access to skin lesion assessment won’t put more burden on specialist care providers or lead to overdiagnosis,” Jones said.

A checklist was created for future studies that highlights things AI/ML developers should consider.

“We hope the results of our systematic review, combined with this checklist, will improve the quality of research in this area and help develop implementable technologies that will benefit patients and clinicians,” Jones said.

As a result of this study, we’re working on a qualitative study to see what patients, the public, healthcare providers, and data scientists think about using AI/ML to diagnose skin cancer in primary care.

SEE ALSO:

With Endometriosis, The Likelihood Of Having a First Live Birth Is Lower