A new artificial intelligence tool could help dermatologists determine whether suspicious skin spots are cancerous, though researchers say it should not be used on its own.

HealthDay News reports the tool, called Dermoscopy Explainable Intelligence (DEXI) correctly identified nearly 94% of skin cancers and accurately ruled out about 75% of noncancerous skin lesions.

The findings were published Oct. 22 in a report by Mayo Clinic Proceedings.

DEXI is part of an imaging system that helps dermatologists track skin lesions over time. Trained on more than 66,000 skin images, the tool assigns each lesion a score from zero to 10 based on its likelihood of being cancerous. It also identifies characteristics such as asymmetry, border and color that contribute to its assessment.

Researchers tested DEXI on 261 skin lesions from 178 patients who underwent full-body photography. They compared the results with assessments from dermatologists.

The average score was 2.5 for noncancerous lesions, 7.8 for skin cancers and 6.0 for melanomas. Researchers identified a score of 3.0 as a potential threshold that could help doctors decide whether to monitor a lesion or perform a biopsy.

Study senior researcher Dr. Jenny Link, a dermatologist at Mayo Clinic in Rochester, Minnesota, said the tool could provide valuable information when evaluating suspicious skin spots.

However, Link cautioned that DEXI requires further testing and does not account for individual factors such as a patient’s age, melanoma history or medications.

Researchers emphasized that the tool should supplement, rather than replace, a doctor’s clinical judgment.