Publications by authors named "L Kostner"
Since December 2019, the COVID-19 pandemic has profoundly affected healthcare. The real effects of the COVID-19 pandemic on skin cancer are still unclear, more than 3 years later. This study aims to summarise the pandemic's impact on skin cancer diagnosis and outcome.
View Article and Find Full Text PDFBackground: Artificial intelligence (AI) shows promising potential to enhance human decision-making as synergistic augmented intelligence (AuI), but requires critical evaluation for skin cancer screening in a real-world setting.
Objectives: To investigate the perspectives of patients and dermatologists after skin cancer screening by human, artificial and augmented intelligence.
Methods: A prospective comparative cohort study conducted at the University Hospital Basel included 205 patients (at high-risk of developing melanoma, with resected or advanced disease) and 8 dermatologists.
Article Synopsis
- The study explores how well two commercial convolutional neural networks (CNNs) assess melanoma risk in real-world dermoscopic images compared to experienced dermatologists.* -
- Conducted at the University Hospital Basel, the research involved analyzing 117 image sets of skin lesions to compare and evaluate the assessment reliability between the two CNNs using variation and correlation metrics.* -
- Results showed that CNN-1 was more consistent in identifying clinically benign lesions with cancerous risk, while CNN-2 excelled with benign-scored lesions; both struggled with lesions that had conflicting risk assessments.*
Article Synopsis
- This study compared the effectiveness of 2D and 3D convolutional neural networks (CNNs) and dermatologists in detecting melanoma in real-life scenarios, involving 1,690 melanocytic lesions in high-risk patients.
- The results showed that 3D-CNN outperformed both 2D-CNN and dermatologists in terms of sensitivity (90%) and had a high ROC-AUC score (0.92), although dermatologists and augmented intelligence matched the sensitivity of 3D-CNN while having superior specificity.
- The 2D-CNN performed poorly with a sensitivity of 70% and specificity of only 40%, indicating that the 3D-CNN is more reliable in early melanoma detection.