Patients with possible skin cancer trust AI to help

The vast majority of patients with suspected melanoma—94%—would welcome the use of AI as an augmenter of their dermatologist’s diagnostic skills. And more than 40% would trust their diagnosis to a standalone AI system.  

So suggests a survey of around 300 patients, around half of whom had a previous diagnosis of skin cancer.

The study was conducted in Germany and is running online in in Frontiers in Medicine.

Led by Titus Brinker and Tanja Jutzi of the German Cancer Research Center in Heidelberg, researchers sent and promoted a web-based questionnaire to members of melanoma support groups and other individuals likely to have a stake in the science.

The team performed statistical tests on the responses to find associations between sociodemographic data and key items in the questionnaire.  

In addition to the above findings on AI acceptance, the responses showed 88% of patients would make their own health data available—as long as their identities were protected—to support the development of AI-based applications in medicine going forward.

The respondents who’d already survived melanoma (154 in number) were more open to the use of AI for new diagnoses, with 56% saying they might even use it at home via mobile app.

Most in this cohort indicated a preference for interpretations made by their physician and the AI independently of one another.

The researchers further found that, in the broader picture of AI utilization across medicine, all patients tend to be concerned about data protection, impersonal care and technology-based medical errors.

Meanwhile patients expect AI to help physicians deliver diagnostics faster, more accurately, more confidently and with less bias than physicians on their own.

“Importantly, although most participants in our survey thought that AI can contribute to improve diagnostic accuracy, they also believed that it cannot and should not replace the dermatologist,” Brinker et al. write in their discussion. “This ambivalence might reflect a general fear that AI might supplant the expertise and diagnostic skills of physicians and is concordant with the desire for personal contact with the treating physician.”

Noting that impersonal consultations were among the most common concerns mentioned in the survey, they surmise that AI-based applications would probably not be accepted as standalone diagnostic systems.

“Along the same lines,” they write, “the majority of German undergraduate medical students don't think that human radiologists will be replaced by AI.”

The study is posted in full for free.