Though AI systems have shown promise for detecting skin cancer, more work is needed before they can be utilized in “real world” applications, according to researchers at the 2019 American Academy of Dermatology annual meeting in Washington, D.C.
“AI systems for skin cancer detection are still in their very early stages,” said Roger S. Ho, MD, MPH, dermatologist and assistant professor at NYU Langone Health in New York, in a prepared statement. “Nothing is 100 percent clear-cut yet.”
Opinions regarding the efficacy of AI and skin cancer detection are clearly varied. A recent study published in the Annals of Oncology found an AI-based network outperformed several dermatologists of different levels of experiences. Researchers at the University of Waterloo found a machine learning software was capable of detecting melanoma skin cancer in its earliest and most treatable stages and noted their method could “be a very powerful tool for skin cancer clinical decision support.
When it comes to AI-based skin cancer detection and smartphone apps, dermatologists noted the need for improvement in the following areas: scoring systems and racial diversity.
Al algorithms assign skin cancer scores to suspicious spots, Ho said, although dermatologists may not know how to interpret such scores. And what’s more, the images used to train skin cancer spotting AI algorithms are usually photographs captured in “optimal conditions.” This may be a problem when using a smartphone-based method to detect skin cancer.
“Just because the computer can read these validated data sets with near 100 percent accuracy doesn’t mean they can read any image,” Ho said. “Everyone has a different phone, lighting, background.”
Though people with lighter skin color are more susceptible, individuals of color can also develop skin cancer. But AI-based systems have mostly been trained using light-skinned patients. Still, AI systems have not been trained to detect skin cancer on the palms or the soles of feet—which is where people with darker skin are disproportionately impacted.
“The algorithm is only as good as what you’ve taught it to do,” Adewole Adamson, MD, MPP, professor at UT Austin Dell Medical School in Austin, Texas, said in the same statement. “If you’ve not taught it to diagnose melanoma in skin of color, then you’re at risk of not being able to do it when the algorithm is complete.”
As AI quickly becomes a popular method for diagnoses in other areas of medicine, Ho added dermatologists should learn to embrace AI. However, he disclaimed the idea that AI is 100 percent accurate. Furthermore, he added AI-based skin cancer detection methods and smartphones should not replace a visit with a dermatologist.
“I don’t think the ‘man versus machine’ framing of AI and machine learning is correct,” Adamson concluded. “It’s going to be more like AI is going to support the dermatologist and make the dermatologist even better.”
As a senior news writer for TriMed, Subrata covers cardiology, clinical innovation and healthcare business. She has a master’s degree in communication management and 12 years of experience in journalism and public relations.