In vivo Study to Evaluate an Intelligent Algorithm for Time Efficient Detection of Malignant Melanoma Using...

Area of Science:

• Dermatology
• Medical Imaging
• Artificial Intelligence

Background:

• Dermatofluoroscopy is an optical, noninvasive method for differentiating melanoma from nevi.
• Current clinical trials show 89% sensitivity and 45% specificity, but long measurement times limit its clinical application.
• An intelligent algorithm was developed to decrease measurement duration without compromising diagnostic accuracy.

Purpose of the Study:

• To evaluate the diagnostic accuracy and time efficiency of a newly developed intelligent algorithm for dermatofluoroscopy.
• To compare the intelligent algorithm's performance against conventional dermatofluoroscopy in differentiating skin lesions.

Main Methods:

• A clinical study included 27 patients with 29 lesions suggestive of cutaneous melanoma.
• Lesions were measured using both conventional dermatofluoroscopy and the intelligent algorithm.
• Results were compared against histopathology findings from two independent pathologists.

Main Results:

• The intelligent algorithm reduced measurement points by a median of 40% (from 265 to 158).
• The algorithm demonstrated higher diagnostic accuracy (AUC 72%) compared to conventional dermatofluoroscopy (AUC 63%).

Conclusions:

• The intelligent algorithm is non-inferior to the conventional method and saves 40% of measurement time.
• Despite improvements, measurement times remain lengthy compared to other noninvasive diagnostic methods.
• Further research is necessary to determine the clinical suitability of this intelligent algorithm.