Convolutional Neural Networks for the evaluation of cancer in Barrett's esophagus: Explainable AI to lighten up the black-box

• Published in: Computers in biology and medicine Vol. 135; p. 104578
• Main Authors: de Souza, Luis A., Mendel, Robert, Strasser, Sophia, Ebigbo, Alanna, Probst, Andreas, Messmann, Helmut, Papa, João P., Palm, Christoph
• Format: Journal Article
• Language: English
• Published: Oxford Elsevier Ltd 01.08.2021; Elsevier Limited
• Subjects: Adenocarcinoma; Algorithms; Annotations; Artificial intelligence; Artificial neural networks; Back propagation; Back propagation networks; Barrett's esophagus; Breast cancer; Classification; Computer applications; Computer-aided diagnosis; Decision making; Deep learning; Endoscopy; Esophageal cancer; Esophagus; Evaluation; Explainable artificial intelligence; Image segmentation; Internal Medicine; Learning algorithms; Machine learning; Mammography; Medical diagnosis; Medical imaging; Neural networks; Other; Reliability analysis; Remission (Medicine); Salience; Sensitivity; Adenocarcinoma; Explainable artificial intelligence; Computer-aided diagnosis; Barrett's esophagus; Machine learning
• ISSN: 0010-4825; 1879-0534; 1879-0534
• DOI: 10.1016/j.compbiomed.2021.104578

Even though artificial intelligence and machine learning have demonstrated remarkable performances in medical image computing, their level of accountability and transparency must be provided in such evaluations. The reliability related to machine learning predictions must be explained and interpreted, especially if diagnosis support is addressed. For this task, the black-box nature of deep learning techniques must be lightened up to transfer its promising results into clinical practice. Hence, we aim to investigate the use of explainable artificial intelligence techniques to quantitatively highlight discriminative regions during the classification of early-cancerous tissues in Barrett's esophagus-diagnosed patients. Four Convolutional Neural Network models (AlexNet, SqueezeNet, ResNet50, and VGG16) were analyzed using five different interpretation techniques (saliency, guided backpropagation, integrated gradients, input × gradients, and DeepLIFT) to compare their agreement with experts' previous annotations of cancerous tissue. We could show that saliency attributes match best with the manual experts' delineations. Moreover, there is moderate to high correlation between the sensitivity of a model and the human-and-computer agreement. The results also lightened that the higher the model's sensitivity, the stronger the correlation of human and computational segmentation agreement. We observed a relevant relation between computational learning and experts' insights, demonstrating how human knowledge may influence the correct computational learning. •To introduce the use of Explainable AI at classifying early cancer in Barrett’s esophagus samples.•To propose a quantitative analysis of CNN’s learning based on Explainable AI techniques.•To assess whether there is an agreement in the visual interpretation of cancerous tissues between Explainable AI and experts annotations.•The manual and automatic segmentations presented from moderate to strong correlation for correct classification of early cancer.