An application of explainable artificial intelligence on medical data is presented. There is an increasing demand in machine learning literature for such explainable models in health-related applications. This work aims to generate explanations on how a Convolutional Neural Network (CNN) detects tumor tissue in patches extracted from histology whole slide images. This is achieved using the “locally-interpretable model-agnostic explanations” methodology. Two publicly-available convolutional neural networks trained on the Patch Camelyon Benchmark are analyzed. Three common segmentation algorithms are compared for superpixel generation, and a fourth simpler parameter-free segmentation algorithm is proposed. The main characteristics of the explanations are discussed, as well as the key patterns identified in true positive predictions. The results are compared to medical annotations and literature and suggest that the CNN predictions follow at least some aspects of human expert knowledge.
Over the last decade, Giant Magnetoimpedance (GMI) sensors have been studied as a new possibility for measuring ultra-weak magnetic fields. To achieve that goal, the employment and optimization of gradiometric configurations are a key factor to reduce the magnetic interference. The performance of GMI gradiometers is strongly related to the homogeneity of the impedance characteristics of the GMI samples. This work presents a method and corresponding electronic implementation to homogenize the phase characteristics of heterogeneous GMI samples, aiming at improving the performance of phase-based GMI magnetometers.
Problem: An application of Explainable Artificial Intelligence Methods for COVID CT-Scan classifiers is presented. Motivation: It is possible that classifiers are using spurious artifacts in dataset images to achieve high performances, and such explainable techniques can help identify this issue. Aim: For this purpose, several approaches were used in tandem, in order to create a complete overview of the classificatios. Methodology: The techniques used included GradCAM, LIME, RISE, Squaregrid, and direct Gradient approaches (Vanilla, Smooth, Integrated). Main results: Among the deep neural networks architectures evaluated for this image classification task, VGG16 was shown to be most affected by biases towards spurious artifacts, while DenseNet was notably more robust against them. Further impacts: Results further show that small differences in validation accuracies can cause drastic changes in explanation heatmaps for DenseNet architectures, indicating that small changes in validation accuracy may have large impacts on the biases learned by the networks. Notably, it is important to notice that the strong performance metrics achieved by all these networks (Accuracy, F1 score, AUC all in the 80 to 90% range) could give users the erroneous impression that there is no bias. However, the analysis of the explanation heatmaps highlights the bias.
Being accountable for the signed reports, pathologists may be wary of high-quality deep learning outcomes if the decision-making is not understandable. Applying off-the-shelf methods with default configurations such as Local Interpretable Model-Agnostic Explanations (LIME) is not sufficient to generate stable and understandable explanations. This work improves the application of LIME to histopathology images by leveraging nuclei annotations, creating a reliable way for pathologists to audit black-box tumor classifiers. The obtained visualizations reveal the sharp, neat and high attention of the deep classifier to the neoplastic nuclei in the dataset, an observation in line with clinical decision making. Compared to standard LIME, our explanations show improved understandability for domain-experts, report higher stability and pass the sanity checks of consistency to data or initialization changes and sensitivity to network parameters. This represents a promising step in giving pathologists tools to obtain additional information on image classification models. The code and trained models are available on GitHub.
Giant Magnetoimpedance (GMI) magnetometers are one of the most recent families of magnetic transducers, being characterized by their potential to achieve high sensitivities. The sensitivity of magnetic transducers is directly related to the sensitivity of its sensor elements. Thus, optimizing the sensitivity of these sensor elements is a critical part of the magnetometers development chain. This paper describes an automatic characterization system designed for the measurement of the electrical impedance of Giant Magneto- Impedance samples. The measurement uncertainties of the system were verified and discussed. The high speed of measurement attained with the use of this system, implemented in LabVIEW, allows for the rapid determination of the optimal operational point of GMI magnetometers.
This work presents an experimental study aiming the sensitivity optimization of Giant Magnetoimpedance (GMI) sensors, by means of properly choosing the values of their conditioning parameters. The optimized GMI sensors, together with the development of improved transduction electronic circuits, can lead to the sensitivity enhancement of GMI biomedical, pressure and magnetic transducer prototypes, previously developed by the research group at LaBioMet, PUC-Rio. Those prototypes are, respectively, aimed at the measurement of arterial pulse waves and the localization of magnetic foreign bodies inserted in the human body. Thus, the experimental characterization of GMI ribbon-shaped samples (Co 70 Fe 5 Si 15 B 10 ) was performed, as a function of the external magnetic field, including the experimental evaluation of the asymmetric GMI effect (AGMI). The parameters that affect the behavior of the GMI samples were experimentally analyzed, such as the DC level (0 mA to 100 mA) and frequency (100 kHz to 10 MHz) of the excitation current, as well as the samples length (1 cm, 3 cm, 5 cm and 15 cm). The conditioning parameters, experimentally identified, that optimize the GMI samples sensitivity lead to a maximum specific sensitivity of 0.84(8400 Ω ·T -1 ·cm -1 ). A new electronic circuit was developed for conditioning and reading of the GMI samples, which directly contributed to the performance enhancement of both transducers. The electronic circuits developed were evaluated supposing the operation of the GMI samples in their most sensitive region. Comparing to previously developed prototypes, the optimum sensitivity achieved for the new configuration of the GMI magnetic transducer increased about 9 times (from 0.12 mV/nT to 1.08 mV/nT), and the sensitivity of the modified pressure transducer increased approximately 7 times (from 1 mV/Pa to 7 mV/Pa).
Keywords
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.