Evidence-based XAI: An empirical approach to design more effective and explainable decision support systems

This paper proposes a user study aimed at evaluating the impact of Class Activation Maps (CAMs) as an eXplainable AI (XAI) method in a radiological diagnostic task, the detection of thoracolumbar (TL) fractures from vertebral X-rays. In particular, we focus on two oft-neglected features of CAMs, that is granularity and coloring, in terms of what features, lower-level vs higher-level, should the maps highlight and adopting which coloring scheme, to bring better impact to the decision-making process, both in terms of diagnostic accuracy (that is effectiveness) and of user-centered dimensions, such as perceived confidence and utility (that is satisfaction), depending on case complexity, AI accuracy, and user expertise. Our findings show that lower-level features CAMs, which highlight more focused anatomical landmarks, are associated with higher diagnostic accuracy than higher-level features CAMs, particularly among experienced physicians. Moreover, despite the intuitive appeal of semantic CAMs, traditionally colored CAMs consistently yielded higher diagnostic accuracy across all groups. Our results challenge some prevalent assumptions in the XAI field and emphasize the importance of adopting an evidence-based and human-centered approach to design and evaluate AI- and XAI-assisted diagnostic tools. To this aim, the paper also proposes a hierarchy of evidence framework to help designers and practitioners choose the XAI solutions that optimize performance and satisfaction on the basis of the strongest evidence available or to focus on the gaps in the literature that need to be filled to move from opinionated and eminence-based research to one more based on empirical evidence and end-user work and preferences.

Deep learning based identification of bone scintigraphies containing metastatic bone disease foci

PURPOSE

Metastatic bone disease (MBD) is the most common form of metastases, most frequently deriving from prostate cancer. MBD is screened with bone scintigraphy (BS), which have high sensitivity but low specificity for the diagnosis of MBD, often requiring further investigations. Deep learning (DL) - a machine learning technique designed to mimic human neuronal interactions- has shown promise in the field of medical imaging analysis for different purposes, including segmentation and classification of lesions. In this study, we aim to develop a DL algorithm that can classify areas of increased uptake on bone scintigraphy scans.

METHODS

We collected 2365 BS from three European medical centres. The model was trained and validated on 1203 and 164 BS scans respectively. Furthermore we evaluated its performance on an external testing set composed of 998 BS scans. We further aimed to enhance the explainability of our developed algorithm, using activation maps. We compared the performance of our algorithm to that of 6 nuclear medicine physicians.

RESULTS

The developed DL based algorithm is able to detect MBD on BSs, with high specificity and sensitivity (0.80 and 0.82 respectively on the external test set), in a shorter time compared to the nuclear medicine physicians (2.5 min for AI and 30 min for nuclear medicine physicians to classify 134 BSs). Further prospective validation is required before the algorithm can be used in the clinic.

HRel: Filter pruning based on High Relevance between activation maps and class labels

This paper proposes an Information Bottleneck theory based filter pruning method that uses a statistical measure called Mutual Information (MI). The MI between filters and class labels, also called Relevance, is computed using the filter's activation maps and the annotations. The filters having High Relevance (HRel) are considered to be more important. Consequently, the least important filters, which have lower Mutual Information with the class labels, are pruned. Unlike the existing MI based pruning methods, the proposed method determines the significance of the filters purely based on their corresponding activation map's relationship with the class labels. Architectures such as LeNet-5, VGG-16, ResNet-56, ResNet-110 and ResNet-50 are utilized to demonstrate the efficacy of the proposed pruning method over MNIST, CIFAR-10 and ImageNet datasets. The proposed method shows the state-of-the-art pruning results for LeNet-5, VGG-16, ResNet-56, ResNet-110 and ResNet-50 architectures. In the experiments, we prune 97.98%, 84.85%, 76.89%, 76.95%, and 63.99% of Floating Point Operation (FLOP)s from LeNet-5, VGG-16, ResNet-56, ResNet-110, and ResNet-50 respectively. The proposed HRel pruning method outperforms recent state-of-the-art filter pruning methods. Even after pruning the filters from convolutional layers of LeNet-5 drastically (i.e., from 20, 50 to 2, 3, respectively), only a small accuracy drop of 0.52% is observed. Notably, for VGG-16, 94.98% parameters are reduced, only with a drop of 0.36% in top-1 accuracy. ResNet-50 has shown a 1.17% drop in the top-5 accuracy after pruning 66.42% of the FLOPs. In addition to pruning, the Information Plane dynamics of Information Bottleneck theory is analyzed for various Convolutional Neural Network architectures with the effect of pruning. The code is available at https://github.com/sarvanichinthapalli/HRel.

COVID-19 detection and disease progression visualization: Deep learning on chest X-rays for classification and coarse localization

Chest X-rays are playing an important role in the testing and diagnosis of COVID-19 disease in the recent pandemic. However, due to the limited amount of labelled medical images, automated classification of these images for positive and negative cases remains the biggest challenge in their reliable use in diagnosis and disease progression. We implemented a transfer learning pipeline for classifying COVID-19 chest X-ray images from two publicly available chest X-ray datasets1,2. The classifier effectively distinguishes inflammation in lungs due to COVID-19 and Pneumonia from the ones with no infection (normal). We have used multiple pre-trained convolutional backbones as the feature extractor and achieved an overall detection accuracy of 90%, 94.3%, and 96.8% for the VGG16, ResNet50, and EfficientNetB0 backbones respectively. Additionally, we trained a generative adversarial framework (a CycleGAN) to generate and augment the minority COVID-19 class in our approach. For visual explanations and interpretation purposes, we implemented a gradient class activation mapping technique to highlight the regions of the input image that are important for predictions. Additionally, these visualizations can be used to monitor the affected lung regions during disease progression and severity stages.