Related Experiment Video
Updated: Sep 27, 2025
09:37
Combining Reflectance Confocal Microscopy with Optical Coherence Tomography for Noninvasive Diagnosis of Skin Cancers via Image Acquisition
Published on: August 18, 2022
2.5K
Explainable artificial intelligence in skin cancer recognition: A systematic review.
Katja Hauser1, Alexander Kurz1, Sarah Haggenmüller1
1Digital Biomarkers for Oncology Group, National Center for Tumor Diseases (NCT), German Cancer Research Center (DKFZ), Heidelberg, Germany.
European Journal of Cancer (Oxford, England : 1990)
|April 7, 2022
Summary
Explainable artificial intelligence (XAI) is frequently used in developing deep neural networks (DNNs) for skin cancer detection. However, rigorous evaluations of XAI
Area of Science:
- Dermatology
- Artificial Intelligence
- Medical Imaging
Background:
- Deep neural networks (DNNs) are increasingly used in medicine for complex problem-solving.
- The
- black-box
- nature of DNNs hinders physician trust in their diagnostic reliability.
- Explainable artificial intelligence (XAI) is proposed to address this opacity in medical AI.
Purpose of the Study:
- To investigate the application of XAI in the development of DNNs for skin cancer detection.
- To identify common visualization techniques used in XAI for dermatology.
- To assess the extent of systematic evaluations of XAI with dermatologists and dermatopathologists.
Main Methods:
- A systematic literature search was conducted across major scientific databases (Google Scholar, PubMed, IEEE Explore, Science Direct, Scopus).
- Peer-reviewed studies published between January 2017 and October 2021 focusing on XAI in skin cancer detection using various dermatological image types were included.
- Search terms included histopathological, whole-slide, clinical, and dermoscopic images, alongside dermatology, explainable, interpretable, and XAI.
Main Results:
- Out of 37 included publications, most (19/37) applied existing XAI methods to interpret DNN decisions.
- A smaller number (4/37) proposed novel or improved XAI techniques.
- While 14/37 studies explored bias detection and human-computer interaction, only three rigorously evaluated the impact of XAI on human performance and confidence in computer-aided diagnosis (CAD) systems.
Conclusions:
- XAI is commonly integrated into the development pipeline of DNNs for skin cancer detection.
- There is a significant lack of systematic and rigorous evaluation regarding the practical utility and effectiveness of XAI in this clinical domain.
