Trial results for an observational study evaluating an AI-based medical device for early Melanoma detection were posted on ClinicalTrials.gov on 2026-03-16, demonstrating an Area Under the ROC Curve (AUC) of 0.8482 for melanoma detection in patients with suspected cutaneous malignancy.
Background
Melanoma is a serious form of skin cancer where early detection significantly improves treatment success. However, distinguishing melanoma from benign lesions can be challenging, making accurate and timely diagnosis critical for patient outcomes. Artificial intelligence (AI) tools are being developed to assist healthcare professionals in this process, aiming to enhance diagnostic accuracy and efficiency.
Trial design
This completed observational study enrolled 105 adult participants with skin lesions suspected of being cancerous. The study's objective was to assess if a computer-aided diagnosis (CAD) system could accurately identify cutaneous melanoma in skin images. The research focused on patients with conditions including Melanoma and Melanoma, Skin. The AI tool's performance was compared against final diagnoses made by doctors or through skin biopsy.
Key results
The study reported several key measurements for the AI tool's performance in detecting melanoma and other malignant conditions in patients with suspected cutaneous malignancy:
- The Area Under the ROC Curve (AUC) for Melanoma Detection was 0.8482 (95% Confidence Interval).
- Accuracy for Melanoma Detection was 0.81 (95% Confidence Interval).
- Sensitivity for Melanoma Detection was 0.93 (95% Confidence Interval).
- Specificity for Melanoma Detection was 0.80 (95% Confidence Interval).
- Top-1 Accuracy for Multiple ICD Categories was 0.55 (95% Confidence Interval).
- Top-3 Accuracy for Multiple ICD Categories was 0.7569 (95% Confidence Interval).
- Top-5 Accuracy for Multiple ICD Categories was 0.8422 (95% Confidence Interval).
- The Area Under the ROC Curve (AUC) for Malignancy Detection was 0.90 (95% Confidence Interval).
- Sensitivity for Multiple Malignant Conditions Detection was 0.81 (95% Confidence Interval).
- Specificity for Multiple Malignant Conditions Detection was 0.86 (95% Confidence Interval).
- Predictive Values (PPV and NPV) for Malignancy were 0.9247 (95% Confidence Interval) and 0.6789 (95% Confidence Interval).
What this means
The results suggest that the AI-based medical device shows strong performance in aiding the detection of melanoma and other cutaneous malignancies. A high sensitivity of 0.93 for melanoma detection indicates the tool's effectiveness at identifying true positive cases, crucial for early cancer diagnosis. The AUC values of 0.8482 for melanoma and 0.90 for general malignancy detection demonstrate good overall discriminative ability. These findings imply that such AI tools could be valuable aids for clinicians, potentially improving diagnostic accuracy and supporting timely intervention for patients with suspected skin cancer.
Source
The information regarding these trial results was obtained from ClinicalTrials.gov, a public database of clinical studies. The results for the study NCT06221397, titled "AI-based Medical Device Validation for Early Melanoma Detection", were posted on 2026-03-16 on clinicaltrials.gov.