AI Tool Achieves 93.4% Accuracy in Identifying Cancerous Skin Lesions, Mayo Clinic Study Shows

AI Tool Transforms Skin Cancer Detection for Dermatologists

Mayo Clinic researchers have validated an AI tool that provides dermatologists with quantifiable data to assess whether skin lesions require biopsy or continued monitoring1

1

2

2

. The Dermoscopy Explainable Intelligence (DEXI) system assigns cancer risk scores from 0.0 to 10.0 based on dermoscopy image analysis, offering dermatologists an additional layer of information alongside their clinical evaluation. Published in Mayo Clinic Proceedings, the study demonstrates how AI-assisted dermatological diagnostics can support clinical judgment in assessing skin lesions for cancer risk.

High Accuracy Rates Using 3.0 Threshold

The retrospective study analyzed 261 lesions from 178 patients who underwent total-body photography, sequential digital dermoscopy imaging and subsequent biopsy3

3

. Using a threshold of 3.0—where scores at or above this level indicate potential skin cancer—DEXI correctly identified 93.4% of cancerous lesions and 75% of noncancerous lesions1

1

. The AI scoring tool revealed stark differences in average scores: noncancerous lesions averaged 2.5, while nonmelanoma skin cancers scored 7.8 and melanomas scored 6.02

2

. These findings establish a clear quantitative benchmark for skin biopsy decision-making.

Deep-Learning Model Provides Explainable Insights

DEXI operates as a deep-learning model trained on more than 66,000 annotated dermoscopy images1

1

. Beyond generating risk scores, the tool uses explainable AI to identify specific image characteristics—including asymmetry, border irregularities and color variations—that contribute to each assessment2

2

. This transparency allows dermatologists to understand the reasoning behind each score rather than relying on a black-box algorithm. "Prior to the study, I would review a patient's DEXI score, but I wouldn't know exactly what to do with the number," says Dr. Jenny Link, the study's lead author and a dermatologist at Mayo Clinic3

3

. "Now I can confidently use the score as another piece of data to inform my clinical judgment on whether I recommend a biopsy or monitoring."

Limitations Point to Future Development Needs

While the AI tool demonstrates strong performance, researchers identified important constraints. DEXI does not account for patient-specific factors such as age, medications or history of melanomas1

1

. The system also showed reduced effectiveness when evaluating severely atypical moles and early evolving melanomas2

2

. Dr. Link emphasizes that "DEXI is a useful and valuable tool to support clinicians as they evaluate cancerous and noncancerous skin lesions, but it should not be used in isolation. The score must be interpreted alongside clinical findings and patient-specific factors, and the tool requires further validation and refinement"3

3

.

Implications for Clinical Practice and Patient Care

The validation of this AI tool addresses a critical gap in dermatological practice where clinicians previously lacked standardized thresholds for cancer risk assessment. By establishing 3.0 as a potential decision threshold, DEXI provides dermatologists with actionable guidance that could reduce unnecessary biopsies while ensuring high-risk lesions receive appropriate intervention. The integration of explainable AI features means clinicians can verify the tool's reasoning against their own observations, fostering trust in AI-assisted diagnostics. As healthcare systems increasingly adopt AI tools for diagnostic support, DEXI's approach of combining quantitative scoring with transparent feature analysis may serve as a model for other specialties. The tool's current limitations with atypical moles and early melanomas suggest that future iterations will likely incorporate patient history and longitudinal data to improve accuracy across all lesion types.