Label noise and self-learning label correction in cardiac abnormalities classification

Noise correction to improve data and model quality for crowdsourcing

Label noise in medical image classification datasets significantly hampers the training of supervised deep learning methods, undermining their generalizability. The test performance of a model tends to decrease as the label noise rate increases. Over recent years, several methods have been proposed to mitigate the impact of label noise in medical image classification and enhance the robustness of the model. Predominantly, these works have employed CNN-based architectures as the backbone of their classifiers for feature extraction. However, in recent years, Vision Transformer (ViT)-based backbones have replaced CNNs, demonstrating improved performance and a greater ability to learn more generalizable features, especially when the dataset is large. Nevertheless, no prior work has rigorously investigated how transformer-based backbones handle the impact of label noise in medical image classification. In this paper, we investigate the architectural robustness of ViT against label noise and compare it to that of CNNs. We use two medical image classification datasets-COVID-DU-Ex, and NCT-CRC-HE-100K-both corrupted by injecting label noise at various rates. Additionally, we show that pretraining is crucial for ensuring ViT's improved robustness against label noise in supervised training.

Automatic modulation classification (AMC) has a wide range of applications in both civilian and military fields, such as industrial Internet of Things (IIoT) security, communication spectrum management, and military electronic countermeasures. However, label mislabeling often occurs in practical scenarios, significantly impacting the performance and robustness of deep neural networks (DNNs). In this article, we propose a meta-learning guided label noise distillation method to enhance the robustness of AMC models against label noise or errors. Specifically, we propose a teacher-student heterogeneous network (TSHN) to discriminate and distill label noise. Following the notion that labels represent information, a teacher network, utilizing trusted few-shot labeled samples, reevaluates and corrects labels for a considerable number of untrusted labeled samples through meta-learning. By dividing and conquering untrusted labeled samples according to their confidence levels, the student network learns more effectively. Additionally, we propose a multiview signal (MVS) method to further enhance the performance of hard-to-classify categories with few-shot trusted labeled samples. Extensive experiments on the RadioML2016 and HisarMod2019.1 data sets demonstrate that our methods significantly improve accuracy and robustness in signal AMC across diverse label noise scenarios, including symmetric, asymmetric, and mixed label noise. For example, compared to the baseline convolutional neural network with the cross-entropy loss, our proposed TSHN achieves a remarkable 1.26% to 36.84% accuracy improvement under symmetric label noise and 0.12% to 38.59% accuracy improvement under mixed label noise. Moreover, TSHN exhibits greater robustness to varying label noise rates compared to existing methods.

Federated learning (FL) is a privacy-preserving distributed learning paradigm that enables clients to jointly train a global model. In real-world FL implementations, client data could have label noise, and different clients could have vastly different label noise levels. Although there exist methods in centralized learning for tackling label noise, such methods do not perform well on heterogeneous label noise in FL settings, due to the typically smaller sizes of client datasets and data privacy requirements in FL. In this paper, we propose FedCorr, a general multi-stage framework to tackle heterogeneous label noise in FL, without making any assumptions on the noise models of local clients, while still maintaining client data privacy. In particular, (1) FedCorr dynamically identifies noisy clients by exploiting the dimensionalities of the model prediction subspaces independently measured on all clients, and then identifies incorrect labels on noisy clients based on persample losses. To deal with data heterogeneity and to increase training stability, we propose an adaptive local proximal regularization term that is based on estimated local noise levels. (2) We further finetune the global model on identified clean clients and correct the noisy labels for the remaining noisy clients after finetuning. (3) Finally, we apply the usual training on all clients to make full use of all local data. Experiments conducted on CIFAR-10/100 with federated synthetic label noise, and on a real-world noisy dataset, Clothing1M, demonstrate that FedCorr is robust to label noise and substantially outperforms the state-of-the-art methods at multiple noise levels.

Automatic detection and classification of electrocardiogram (ECG) signals is of great importance for diagnosis of cardiac abnormalities. A method is proposed here to classify different cardiac abnormalities like Cardiomyopathy, Myocardial infarction, Dysrhythmia, Myocardial hypertrophy and Valvular heart disease. Support Vector Machine (SVM) has been used to classify the patterns inherent in the features extracted through Continuous Wavelet Transform (CWT) of different ECG signals. CWT allows a time domain signal to be transformed into time-frequency domain such that frequency characteristics and the location of particular features in a time series may be highlighted simultaneously. Thus it allows accurate extraction of feature from non-stationary signals like ECG. SVM transforms the multi-dimensional feature space into a linearly separable feature space with the help of Kernel function. In the present work, SVM in regression mode has been successfully applied for the classification of cardiac abnormalities with good diagnostic accuracy.

Classification algorithms are widely applied to predict failures and detect anomalies in various application areas. It is common to assume that the data and labels are correct when training, but this is challenging to guarantee in the real world. If there are erroneous labels in the training data, a model can easily overfit to these, resulting in poor performance. How to handle label noise has been previously researched, however, few works focus on label noise in anomaly detection. In this work, we propose LDAAD, a novel algorithm framework for label de-noising for anomaly detection that combines unsupervised learning and semi-supervised learning methods. Specifically, we apply anomaly detection to partition the training data into low-risk and high-risk sets. We subsequently build upon ideas from cross-validation and train multiple classification models on segments of the low-risk data. The models are used both to relabel the samples in the high-risk set and to filter the low-risk samples. Finally, we merge the two sets to obtain a final sample set with more confident labels. We evaluate LDAAD on multiple real-world datasets and show that LDAAD achieves robust results that outperform the benchmark methods. Specifically, LDAAD achieves a 5% accuracy improvement over the second-best method for symmetric noise while having a minimal detrimental impact when no label noise is present.

Availability of large annotated datasets in computer vision, speech understanding, or natural language processing is one of the main reasons for deep neural networks popularity. Unfortunately, such data can suffer from label noise, introduced by incorrectly labelled patterns. Since neural networks, as data-driven approaches, are strongly dependent on the quality of training data, the results of building deep neural structures with noisy examples can be unreliable. In this paper, we present preliminary experimental results on how two regularization techniques, namely dropout and batch normalization, influence vulnerability to incorrect labels. On popular MNIST and CIFAR-10 datasets we demonstrate that combination of these two approaches can be considered as a tool to improve network robustness to mislabelled training examples.

Atrial fibrillation (AF) and congestive heart failure (CHF) are the most prevalent types of cardiovascular disorders as the leading cause of death due to delayed diagnosis. Early diagnosis of these cardiac conditions is possible by manually analyzing electrocardiogram (ECG) signals. However, manual diagnosis is complex, owing to the various characteristics of ECG signals. An accurate classification system for AF and CHF has the potential to save patient lives. Therefore, this study proposed an ECG signal classification system for AF and CHF using a one-dimensional convolutional neural network (1-D CNN) to provide a robust classification system performance. This study used ECG signal recording of AF, CHF, and NSR, which can be accessed on the Physionet website. A total of 5600 ECG signal segments were obtained from 56 subjects, divided into train sets from 42 subjects (N = 4200 ECG segments), and test sets from 14 subjects (N = 1400). We applied for leave-one-out cross-validation in training to select the best model. The proposed 1-D CNN algorithm successfully classified raw data of ECG signals into normal sinus rhythm (NSR), AF, and CHF by providing the highest classification accuracy of 99.643%, f1-score, recall, and precision of 0.996, respectively, with an AUC score of 0.999. The results showed that the proposed method extracted the ECG signal information directly without needing several preprocessing steps and feature extraction methods that potentially reduce the information contained in the ECG signals. Furthermore, the proposed method outperformed previous studies in classifying AF, CHF, and NSR. Therefore, this approach can be considered as an adjunct for medical personnel to diagnose AF, CHF, and NSR.

Electrocardiogram (ECG) signals are widely used to diagnose heart health. Experts can detect multiple cardiac abnormalities using the ECG signal. In a clinical setting, 12-lead ECG is mainly used. But using fewer leads can make the ECG more pervasive as it can be integrated with wearable devices. At the same time, we need to build systems that can diagnose cardiac abnormalities automatically. This work develops a channel self-attention-based deep neural network to diagnose cardiac abnormality using a different number of ECG lead combinations. Our approach takes care of the temporal and spatial interdependence of multi-lead ECG signals. Our team participates under the name “cardiochallenger” in the “PhysioNetl-Computing in Cardiology Challenge 2021”. Our method achieves the challenge metric score of 0.55, 0.51, 0.53, 0.51, and 0.53 (ranked 2^nd, 5^th, 4^th, 5^th and 4th) for the 12-lead, 6-lead, 4-lead, 3-lead, and 2-lead cases, respectively, on the test data set.

Electrocardiogram (ECG) signals are vital in diagnosing various cardiac abnormalities, yet they are often corrupted by artifacts such as powerline interference (PLI), muscle noise, and baseline wander (BW), which can obscure important clinical information. The present research explores a novel approach for multiple artifacts removal from ECG signals using cascading of multistage least mean square (LMS) adaptive FIR filter. The proposed method employs a cascade of LMS adaptive FIR filters, each optimized for targeting specific artifacts found in the ECG signal. By iteratively adapting filter coefficients based on the input signal and a reference signal obtained from the preceding stage, the proposed cascaded LMS adaptive FIR filter effectively suppresses different artifacts while preserving the integrity of the ECG signal. The proposed algorithm’s performance has been analyzed by evaluating the mean square error (MSE) value and signal-to-noise ratio (SNR) of the filtered ECG, and evaluation conducted on synthetic and real-world ECG datasets demonstrates the efficacy of the proposed method in reducing various types of artifacts, thereby reliability of ECG signal interpretation and enhancing the quality. This method holds promise for improving the precision of cardiac monitoring and diagnosis in clinical settings. The proposed model is simulated on MATLAB to measure the performance parameter which demonstrates the validation of model.

Animal species classification using deep neural networks with noise labels

An ECG denoising technique based on AHIN block and gradient difference max loss

We present a Locally-Alignment Multi-Kernel SVM (LA-MKL) that adaptively combines linear, RBF, and self-tuned RBF kernels using per-sample weights derived from local kernel–target alignment, while remaining PSD-safe. We evaluate our method on two challenging CKD-like datasets (d=24 features): A (n=600) with a ring-plus-cluster structure and 8% MCAR missingness, and B (n=700) with 5% label noise, 30% domain shift, 12% missingness, and heavy-tailed perturbations. Using 5-fold cross-validation, LA-MKL consistently matches strong linear baselines under moderate heterogeneity and outperforms standard kernels under severe non-stationarity, e.g., on CKD-like B, it attains an Accuracy of 0.911±0.021, F1 score of 0.777±0.046, and ROC-AUC of 0.894±0.034, whereas conventional RBF suffers from recall collapse. These results indicate that local alignment–driven weighting effectively emphasizes the most discriminative kernel in each neighborhood, thereby improving robustness to missingness, imbalance, and distributional shifts in biomedical prediction.

Wearable technologies enable continuous monitoring of various health metrics, such as physical activity, heart rate, sleep, and stress levels. A key challenge with wearable data is obtaining quality labels. Unlike modalities like video where the videos themselves can be effectively used to label objects or events, wearable data do not contain obvious cues about the physical manifestation of the users and usually require rich metadata. As a result, label noise can become an increasingly thorny issue when labeling wearable data. In this paper, we propose a novel solution to address noisy label learning, entitled Few-Shot Human-in-the-Loop Refinement (FHLR). Our method initially learns a seed model using weak labels. Next, it fine-tunes the seed model using a handful of expert corrections. Finally, it achieves better generalizability and robustness by merging the seed and fine-tuned models via weighted parameter averaging. We evaluate our approach on four challenging tasks and datasets, and compare it against eight competitive baselines designed to deal with noisy labels. We show that FHLR achieves significantly better performance when learning from noisy labels and achieves state-of-the-art by a large margin, with up to accuracy improvement under symmetric and asymmetric noise. Notably, we find that FHLR is particularly robust to increased label noise, unlike prior works that suffer from severe performance degradation. Our work not only achieves better generalization in high-stakes health sensing benchmarks but also sheds light on how noise affects commonly-used models.

Point cloud segmentation is a fundamental task in 3D. Despite recent progress on point cloud segmentation with the power of deep networks, current learning methods based on the clean label assumptions may fail with noisy labels. Yet, class labels are often mislabeled at both instance-level and boundary-level in real-world datasets. In this work, we take the lead in solving the instance-level label noise by proposing a Point Noise-Adaptive Learning (PNAL) framework. Compared to noise-robust methods on image tasks, our framework is noise-rate blind, to cope with the spatially variant noise rate specific to point clouds. Specifically, we propose a point-wise confidence selection to obtain reliable labels from the historical predictions of each point. A cluster-wise label correction is proposed with a voting strategy to generate the best possible label by considering the neighbor correlations. To handle boundary-level label noise, we also propose a variant "PNAL-boundary" with a progressive boundary label cleaning strategy. Extensive experiments demonstrate its effectiveness on both synthetic and real-world noisy datasets. Even with 60% symmetric noise and high-level boundary noise, our framework significantly outperforms its baselines, and is comparable to the upper bound trained on completely clean data. Moreover, we cleaned the popular real-world dataset ScanNetV2 for rigorous experiment. Our code and data is available at https://github.com/pleaseconnectwifi/PNAL.