Explainable artificial intelligence in skin cancer recognition: A systematic review

Hauser, Katja; Kurz, Alexander; Haggenmüller, Sarah; Maron, Roman C.; Kalle, Christof von; Utikal, Jochen; Meier, Friedegund; Hobelsberger, Sarah; Gellrich, Frank Friedrich; Sergon, Mildred; Hauschild, Axel; French, Lars E.; Heinzerling, Lucie; Schlager, Justin Gabriel; Ghoreschi, Kamran; Schlaak, Max; Hilke, Franz J.; Poch, Gabriela; Kutzner, Heinz; Berking, Carola; Heppt, Markus V.; Erdmann, Michael; Haferkamp, Sebastian; Schadendorf, Dirk; Sondermann, Wiebke; Goebeler, Matthias; Schilling, Bastian; Kather, Jakob Nikolas; Fröhling, Stefan; Lipka, Daniel B.; Hekler, Achim; Krieghoff-Henning, Eva; Brinker, Titus J.

doi:10.1016/j.ejca.2022.02.025

Cited by 63 publications

(31 citation statements)

References 61 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Besides the aforementioned infrastructural features, digitalization of tissue sections allows the application of deep learning algorithms with the potential to support the objective, consistent clinical and diagnostic decision making. 25,26 Deep learning algorithms and CNN have been previously used to classify benign and malignant diseases on conventionally stained, scanned histopathological slides in various organs such as skin, 12,27 lung, 10,13 breast, 28,29 prostate, 30,31 or intestinal tissue. 11,32 In addition, it was used for the assignment of the tumour origin in unknown primary cancers using a metastasized tumours from different anatomical sites (e.g., lymph node, liver), including different types of adenocarcinoma, squamous cell carcinoma, renal, or urothelial carcinoma; although without discrimination of adjacent non-neoplastic or benign similarly appearing structures, 33 as shown here.…”

Section: Discussionmentioning

confidence: 99%

Implementation of deep learning in liver pathology optimizes diagnosis of benign lesions and adenocarcinoma metastasis

Kriegsmann

Steinbuß

et al. 2023

Clinical & Translational Med

View full text Add to dashboard Cite

Introduction Differentiation of histologically similar structures in the liver, including anatomical structures, benign bile duct lesions, or common types of liver metastases, can be challenging with conventional histological tissue sections alone. Accurate histopathological classification is paramount for the diagnosis and adequate treatment of the disease. Deep learning algorithms have been proposed for objective and consistent assessment of digital histopathological images. Materials and methods In the present study, we trained and evaluated deep learning algorithms based on the EfficientNetV2 and ResNetRS architectures to discriminate between different histopathological classes. For the required dataset, specialized surgical pathologists annotated seven different histological classes, including different non‐neoplastic anatomical structures, benign bile duct lesions, and liver metastases from colorectal and pancreatic adenocarcinoma in a large patient cohort. Annotation resulted in a total of 204.159 image patches, followed by discrimination analysis using our deep learning models. Model performance was evaluated on validation and test data using confusion matrices. Results Evaluation of the test set based on tiles and cases revealed overall highly satisfactory prediction capability of our algorithm for the different histological classes, resulting in a tile accuracy of 89% (38 413/43 059) and case accuracy of 94% (198/211). Importantly, the separation of metastasis versus benign lesions was certainly confident on case level, confirming the classification model performed with high diagnostic accuracy. Moreover, the whole curated raw data set is made publically available. Conclusions Deep learning is a promising approach in surgical liver pathology supporting decision making in personalized medicine.

show abstract

Section: Discussionmentioning

confidence: 99%

Implementation of deep learning in liver pathology optimizes diagnosis of benign lesions and adenocarcinoma metastasis

Kriegsmann

Steinbuß

et al. 2023

Clinical & Translational Med

View full text Add to dashboard Cite

show abstract

“…16 Another recent review of XAI techniques in the skin cancer domain analyzed 37 studies and reported that only three of these studies conducted user experiments to evaluate human performance with XAI systems. 17 This scarcity speaks to both the lack of awareness for evaluation as well as the high cost of evaluation. Much of our work in this article is focused on developing algorithmic metrics of interpretability that could help reduce the cost of user studies.…”

Section: Related Workmentioning

confidence: 99%

Evaluating perceptual and semantic interpretability of saliency methods: A case study of melanoma

Bokadia

Yang

et al. 2022

Applied AI Letters

View full text Add to dashboard Cite

In order to be useful, XAI explanations have to be faithful to the AI system they seek to elucidate and also interpretable to the people that engage with them. There exist multiple algorithmic methods for assessing faithfulness, but this is not so for interpretability, which is typically only assessed through expensive user studies. Here we propose two complementary metrics to algorithmically evaluate the interpretability of saliency map explanations. One metric assesses perceptual interpretability by quantifying the visual coherence of the saliency map. The second metric assesses semantic interpretability by capturing the degree of overlap between the saliency map and textbook features-features human experts use to make a classification. We use a melanoma dataset and a deep-neural network classifier as a case-study to explore how our two interpretability metrics relate to each other and a faithfulness metric. Across six commonly used saliency methods, we find that none achieves high scores across all three metrics for all test images, but that different methods perform well in different regions of the data distribution. This variation between methods can be leveraged to consistently achieve high interpretability and faithfulness by using our metrics to inform saliency mask selection on a case-by-case basis. Our interpretability metrics provide a new way to evaluate saliency-based explanations and allow for the adaptive combination of saliency-based explanation methods.

show abstract

“…less computational power and better interpretability. While there are studies which investigate more interpretable ensembles [38,39], the majority of interpretability studies focus on single models, which is already a challenging task [21].…”

Section: Soup Performance: Generalisation Robustness and Calibrationmentioning

confidence: 99%

“…However, with ever improving hardware, the severity of this limitation will continue to decrease in the future. A tougher problem is the necessity for transparency [20], which is complicated through the black box nature of DL-algorithms [21]. Adding more 'black boxes' to a classifier is unlikely to simplify this problem.…”

Section: Introductionmentioning

confidence: 99%

Model soups improve performance of dermoscopic skin cancer classifiers

Maron

Hekler

Haggenmüller

et al. 2022

European Journal of Cancer

Self Cite

View full text Add to dashboard Cite

Explainable artificial intelligence in skin cancer recognition: A systematic review

Cited by 63 publications

References 61 publications

Implementation of deep learning in liver pathology optimizes diagnosis of benign lesions and adenocarcinoma metastasis

Implementation of deep learning in liver pathology optimizes diagnosis of benign lesions and adenocarcinoma metastasis

Evaluating perceptual and semantic interpretability of saliency methods: A case study of melanoma

Model soups improve performance of dermoscopic skin cancer classifiers

Contact Info

Product

Resources

About