Simplicity bias is a critical challenge in neural net-works since it often leads to favoring simpler solutions and learning unintended decision rules captured by spuri-ous correlations, causing models to be biased and dimin-ishing their generalizability. While existing solutions rely on human supervision, obtaining annotations of the dif-ferent bias attributes is often impractical. To tackle this, we present Debiasify, a novel self-distillation approach that works without any prior information about the nature of biases. Our method leverages a new distillation loss to distill knowledge within a network; from a deep layer where complex, highly-predictive features reside, to a shallow layer where simpler yet attribute-conditioned features are found in an unsupervised manner. In this way, Debiasify learns robust, debiased representations that generalize well across various biases and datasets, enhancing worst-group performance and overall accuracy. Extensive experiments on computer vision and medical imaging benchmarks show the efficacy of our method, significantly outperforming the previous unsupervised debiasing methods (e.g., a 10.13% improvement in worst-group accuracy on Wavy Hair classi-fication in CelebA) while achieving comparable or superior performance to supervised methods. Our code is publicly available at the following link:Debiasify.

Continual Learning (CL) enables neural networks to learn new tasks while retaining previous knowledge. However, most CL methods fail to address bias transfer, where spurious correlations propagate to future tasks or influence past knowledge. This bidirectional bias transfer negatively impacts model performance and fairness, especially in medical imaging, where it can lead to misdiagnoses and unequal treatment. In this work, we show that conventional CL methods amplify these biases, posing risks for diverse patient cohorts. To address this, we propose BiasPruner, a framework that mitigates bias propagation through debiased subnetworks, while preserving sequential learning and avoiding catastrophic forgetting. BiasPruner computes a bias attribution score to identify and prune network units responsible for spurious correlations, creating task-specific subnetworks that learn unbiased representations. As new tasks are learned, the framework integrates non-biased units from previous subnetworks to preserve transferable knowledge and prevent bias transfer. During inference, a task-agnostic gating mechanism selects the optimal subnetwork for robust predictions. We evaluate BiasPruner on medical imaging benchmarks, including skin lesion and chest X-ray classification tasks, where biased data (e.g., spurious skin tone correlations) can exacerbate disparities. Our experiments show that BiasPruner outperforms state-of-the-art CL methods in both accuracy and fairness. Code is available at: BiasPruner.

Deep learning models have achieved remarkable success in medical image classification. These models are typically trained once on the available annotated images and thus lack the ability of continually learning new tasks (i.e., new classes or data distributions) due to the problem of catastrophic forgetting. Recently, there has been more interest in designing continual learning methods to learn different tasks presented sequentially over time while preserving previously acquired knowledge. However, these methods focus mainly on preventing catastrophic forgetting and are tested under a closed-world assumption; i.e., assuming the test data is drawn from the same distribution as the training data. In this work, we advance the state-of-the-art in continual learning by proposing GC2 for medical image classification, which learns a sequence of tasks while simultaneously enhancing its out-of-distribution robustness. To alleviate forgetting, GC2 employs a gradual culpability-based network pruning to identify an optimal subnetwork for each task. To improve generalization, GC2 incorporates adversarial image augmentation and knowledge distillation approaches for learning generalized and robust representations for each subnetwork. Our extensive experiments on multiple benchmarks in a task-agnostic inference demonstrate that GC2 significantly outperforms baselines and other continual learning methods in reducing forgetting and enhancing generalization. Our code is publicly available at the following link: https://github.com/nourhanb/TMI2024-GC2.

Deep learning (DL) models trained to minimize empirical risk on a single domain often fail to generalize when applied to other domains. Model failures due to poor generalizability are quite common in practice and may prove quite perilous in mission-critical applications, e.g., diagnostic imaging where real-world data often exhibits pronounced variability. Such limitations have led to increased interest in domain generalization (DG) approaches that improve the ability of models learned from a single or multiple source domains to generalize to out-of-distribution (OOD) test domains. In this work, we propose BoosterNet, a lean add-on network that can be simply appended to any arbitrary core network to improve its generalization capability without requiring any changes in its architecture or training procedure. Specifically, using a novel measure of feature culpability, BoosterNet is trained episodically on the most and least culpable data features extracted from critical units in the core network based on their contribution towards class-specific prediction errors, which have shown to improve generalization. At inference time, corresponding test image features are extracted from the closest class-specific units, determined by smart gating via a Siamese network, and fed to BoosterNet for improved generalization. We evaluate the performance of BoosterNet within two very different classification problems, digits and skin lesions, and demonstrate a marked improvement in model generalization to OOD test domains compared to SOTA.

This paper presents the design of a fully integrated electrocardiogram (ECG) signal processor (ESP) for the prediction of ventricular arrhythmia using a unique set of ECG features and a naive Bayes classifier. Real-time and adaptive techniques for the detection and the delineation of the P-QRS-T waves were investigated to extract the fiducial points. Those techniques are robust to any variations in the ECG signal with high sensitivity and precision. Two databases of the heart signal recordings from the MIT PhysioNet and the American Heart Association were used as a validation set to evaluate the performance of the processor. Based on application-specified integrated circuit (ASIC) simulation results, the overall classification accuracy was found to be 86% on the out-of-sample validation data with 3-s window size. The architecture of the proposed ESP was implemented using 65-nm CMOS process. It occupied 0.112- ${\rm mm}^{2}$ area and consumed 2.78- $\mu \text{W}$ power at an operating frequency of 10 kHz and from an operating voltage of 1 V. It is worth mentioning that the proposed ESP is the first ASIC implementation of an ECG-based processor that is used for the prediction of ventricular arrhythmia up to 3 h before the onset.

This paper presents a real-time adaptive ECG detection and delineation algorithm alongside an architecture based on time-domain signal processing of the ECG signal. The algorithm is enhanced to detect large number of different P-QRS-T waveform morphologies using adaptive search windows and adaptive threshold levels. The proposed architecture has been implemented in the state-of-the-art 65-nm CMOS technology. It occupied 0.03416 mm2 area and consumed 0.614 mW power. Furthermore, the non-complex nature of the architecture resulted with a realization using smaller number of computation and higher performance. The design of the QRS detector was tested on ECG records obtained from the Physionet QT database and achieved a sensitivity of Se =99.83% and a positive predictivity of P+= 98.65%. Similarly, the mean error values of the T peak, T offset, P peak and P offset were found to be -1.367, 6.36, 5.5 and -2.59 milliseconds, respectively, using the same database. The small area, low power, and high performance of our architecture makes it suitable for inclusion in System On Chips (SOCs) targeting wearable mobile medical devices.

ECG intervals such as QRS, QT and PR provide significant information and are widely used as clinical parameters for diagnosing cardiac diseases. This paper presents a novel QRS detection technique based on Curve Length Transform (CLT) and a refined delineation of P-wave and T-wave using Discrete Wavelet Transform (DWT). The proposed technique was verified using the PhysioNet database. The QRS detection achieved a sensitivity of 98.59% and a positive predictivity of 97.86%. The QRS duration, QT interval and PR interval had a mean error of -1.56± 28.8ms, -5.39± 42.4ms and 0.86± 40.3ms respectively. The proposed algorithm is computationally efficient and is simpler to implement in hardware, hence, will lead to a faster execution time, smaller design area and consequently low power consumption.