Evaluation of digital dermoscopy in a pigmented lesion clinic: clinician versus computer assessment of malignancy risk

• HIT Description: decision support system for dermatology More info...
  
• Purpose of Study: determine the level of agreement in malignancy risk assessment between a pigmented lesion specialist and a digital demoscopy system based on an articificial neural network-based algorithm of the international, multi-center Diagnostic and Neural analysis of Skin Cancer trial
  
• Years of study: 2002-2005
  
• Study Design: other predictive analysis
  

• Settings: patients with pigmented lesions attending the Stanford Pigmented Lesion and Cutaneous Melanoma Clinic
• Intervention: pigmented lesions were evaluated by the specialists as well as photographed using the microDERM camera with subsequent comparison between the malignancy risk score assigned by the clinician viewing the photograph and the score calculated suing the artificial neural network
• Evaluation Method: assessment of interrater reliability and comparison of ratings to biopsy results
• Description: microDERM artificial neural network (ANN) mage analysis and feature assessment to calculate a risk assessment score and was developed using 2218 dermoscopic images and 968 histopathologic correlations to train the ANN to recognize and differentiate benign lesions from suspected melanoma
• Clinical Outcomes: inter-observer agreement was poor between the clinician scoring and the ANN scoring and the ANN malignancy risk scores did not necessarily correlate with the degree of histologic atypia