ExAID: A multimodal explanation framework for computer-aided diagnosis of skin lesions.

BACKGROUND AND OBJECTIVES: One principal impediment in the successful deployment of Artificial Intelligence (AI) based Computer-Aided Diagnosis (CAD) systems in everyday clinical workflows is their lack of transparent decision-making. Although commonly used eXplainable AI (XAI) methods provide insights into these largely opaque algorithms, such explanations are usually convoluted and not readily comprehensible. The explanation of decisions regarding the malignancy of skin lesions from dermoscopic images demands particular clarity, as the underlying medical problem definition is ambiguous in itself. This work presents ExAID (Explainable AI for Dermatology), a novel XAI framework for biomedical image analysis that provides multi-modal concept-based explanations, consisting of easy-to-understand textual explanations and visual maps, to justify the predictions. METHODS: Our framework relies on Concept Activation Vectors to map human-understandable concepts to those learned by an arbitrary Deep Learning (DL) based algorithm, and Concept Localisation Maps to highlight those concepts in the input space. This identification of relevant concepts is then used to construct fine-grained textual explanations supplemented by concept-wise location information to provide comprehensive and coherent multi-modal explanations. All decision-related information is presented in a diagnostic interface for use in clinical routines. Moreover, the framework includes an educational mode providing dataset-level explanation statistics as well as tools for data and model exploration to aid medical research and education processes. RESULTS: Through rigorous quantitative and qualitative evaluation of our framework on a range of publicly available dermoscopic image datasets, we show the utility of multi-modal explanations for CAD-assisted scenarios even in case of wrong disease predictions. We demonstrate that concept detectors for the explanation of pre-trained networks reach accuracies of up to 81.46%, which is comparable to supervised networks trained end-to-end. CONCLUSIONS: We present a new end-to-end framework for the multi-modal explanation of DL-based biomedical image analysis in Melanoma classification and evaluate its utility on an array of datasets. Since perspicuous explanation is one of the cornerstones of any CAD system, we believe that ExAID will accelerate the transition from AI research to practice by providing dermatologists and researchers with an effective tool that they can both understand and trust. ExAID can also serve as the basis for similar applications in other biomedical fields.

Correction to: Two-stage framework for optic disc localization and glaucoma classification in retinal fundus images using deep learning.

Following publication of the original article.

Two-stage framework for optic disc localization and glaucoma classification in retinal fundus images using deep learning.

BACKGROUND: With the advancement of powerful image processing and machine learning techniques, Computer Aided Diagnosis has become ever more prevalent in all fields of medicine including ophthalmology. These methods continue to provide reliable and standardized large scale screening of various image modalities to assist clinicians in identifying diseases. Since optic disc is the most important part of retinal fundus image for glaucoma detection, this paper proposes a two-stage framework that first detects and localizes optic disc and then classifies it into healthy or glaucomatous. METHODS: The first stage is based on Regions with Convolutional Neural Network (RCNN) and is responsible for localizing and extracting optic disc from a retinal fundus image while the second stage uses Deep Convolutional Neural Network to classify the extracted disc into healthy or glaucomatous. Unfortunately, none of the publicly available retinal fundus image datasets provides any bounding box ground truth required for disc localization. Therefore, in addition to the proposed solution, we also developed a rule-based semi-automatic ground truth generation method that provides necessary annotations for training RCNN based model for automated disc localization. RESULTS: The proposed method is evaluated on seven publicly available datasets for disc localization and on ORIGA dataset, which is the largest publicly available dataset with healthy and glaucoma labels, for glaucoma classification. The results of automatic localization mark new state-of-the-art on six datasets with accuracy reaching 100% on four of them. For glaucoma classification we achieved Area Under the Receiver Operating Characteristic Curve equal to 0.874 which is 2.7% relative improvement over the state-of-the-art results previously obtained for classification on ORIGA dataset. CONCLUSION: Once trained on carefully annotated data, Deep Learning based methods for optic disc detection and localization are not only robust, accurate and fully automated but also eliminates the need for dataset-dependent heuristic algorithms. Our empirical evaluation of glaucoma classification on ORIGA reveals that reporting only Area Under the Curve, for datasets with class imbalance and without pre-defined train and test splits, does not portray true picture of the classifier's performance and calls for additional performance metrics to substantiate the results.