Large Labeled Training Datasets Research Articles

DPI-MoCo: Deep Prior Image Constrained Motion Compensation Reconstruction for 4D CBCT.

4D cone-beam computed tomography (CBCT) plays a critical role in adaptive radiation therapy for lung cancer. However, extremely sparse sampling projection data will cause severe streak artifacts in 4D CBCT images. Existing deep learning (DL) methods heavily rely on large labeled training datasets which are difficult to obtain in practical scenarios. Restricted by this dilemma, DL models often struggle with simultaneously retaining dynamic motions, removing streak degradations, and recovering fine details. To address the above challenging problem, we introduce a Deep Prior Image Constrained Motion Compensation framework (DPI-MoCo) that decouples the 4D CBCT reconstruction into two sub-tasks including coarse image restoration and structural detail fine-tuning. In the first stage, the proposed DPI-MoCo combines the prior image guidance, generative adversarial network, and contrastive learning to globally suppress the artifacts while maintaining the respiratory movements. After that, to further enhance the local anatomical structures, the motion estimation and compensation technique is adopted. Notably, our framework is performed without the need for paired datasets, ensuring practicality in clinical cases. In the Monte Carlo simulation dataset, the DPI-MoCo achieves competitive quantitative performance compared to the state-of-the-art (SOTA) methods. Furthermore, we test DPI-MoCo in clinical lung cancer datasets, and experiments validate that DPI-MoCo not only restores small anatomical structures and lesions but also preserves motion information.

Geo-knowledge-guided GPT models improve the extraction of location descriptions from disaster-related social media messages

Social media messages posted by people during natural disasters often contain important location descriptions, such as the locations of victims. Recent research has shown that many of these location descriptions go beyond simple place names, such as city names and street names, and are difficult to extract using typical named entity recognition (NER) tools. While advanced machine learning models could be trained, they require large labeled training datasets that can be time-consuming and labor-intensive to create. In this work, we propose a method that fuses geo-knowledge of location descriptions and a Generative Pre-trained Transformer (GPT) model, such as ChatGPT and GPT-4. The result is a geo-knowledge-guided GPT model that can accurately extract location descriptions from disaster-related social media messages. Also, only 22 training examples encoding geo-knowledge are used in our method. We conduct experiments to compare this method with nine alternative approaches on a dataset of tweets from Hurricane Harvey. Our method demonstrates an over 40% improvement over typically used NER approaches. The experiment results also show that geo-knowledge is indispensable for guiding the behavior of GPT models. The extracted location descriptions can help disaster responders reach victims more quickly and may even save lives.

Using self-supervised feature learning to improve the use of pulse oximeter signals to predict paediatric hospitalization.

Background: The success of many machine learning applications depends on knowledge about the relationship between the input data and the task of interest (output), hindering the application of machine learning to novel tasks. End-to-end deep learning, which does not require intermediate feature engineering, has been recommended to overcome this challenge but end-to-end deep learning models require large labelled training data sets often unavailable in many medical applications. In this study, we trained self-supervised learning (SSL) models for automatic feature extraction from raw photoplethysmography (PPG) obtained using a pulse oximeter, with the aim of predicting paediatric hospitalization. Methods: We compared logistic regression models fitted using features extracted using SSL with models trained using both clinical and SSL features. In addition, we compared end-to-end deep learning models initialized randomly or using weights from the SSL models. We also compared the performance of SSL models trained on labelled data alone (n=1,031) with SSL trained using both labelled and unlabelled signals (n=7,578). Results: Logistic regression models were more predictive of hospitalization when trained on features extracted using labelled PPG signals only compared to SSL models trained on both labelled and unlabelled signals (AUC 0.83 vs 0.80). However, features extracted using SSL model trained on both labelled and unlabelled PPG signals were more predictive of hospitalization when concatenated with clinical features (AUC 0.89 vs 0.87). The end-to-end deep learning model had an AUC of 0.80 when initialized using the SSL model trained on all PPG signals, 0.77 when initialized using SSL trained on labelled data only, and 0.73 when initialized randomly. Conclusions: This study shows that SSL can extract features from PPG signals that are predictive of hospitalization or initialize end-to-end deep learning models. Furthermore, SSL can leverage larger unlabelled data sets to improve performance of models fitted using small labelled data sets.

Witan

Effective supervised training of modern machine learning models often requires large labelled training datasets, which could be prohibitively costly to acquire for many practical applications. Research addressing this problem has sought ways to leverage weak supervision sources, such as the user-defined heuristic labelling functions used in the data programming paradigm, which are cheaper and easier to acquire. Automatic generation of these functions can make data programming even more efficient and effective. However, existing approaches rely on initial supervision in the form of small labelled datasets or interactive user feedback. In this paper, we propose Witan, an algorithm for generating labelling functions without any initial supervision. This flexibility affords many interaction modes, including unsupervised dataset exploration before the user even defines a set of classes. Experiments in binary and multi-class classification demonstrate the efficiency and classification accuracy of Witan compared to alternative labelling approaches.

Cross-Domain Activity Recognition Using Shared Representation in Sensor Data

Existing models based on sensor data for human activity recognition are reporting state-of-the-art performances. Most of these models are conducted based on single-domain learning in which for each domain a model is required to be trained. However, the generation of adequate labelled data and a learning model for each domain separately is often time-consuming and computationally expensive. Moreover, the deployment of multiple domain-wise models is not scalable as it obscures domain distinctions, introduces extra computational costs, and limits the usefulness of training data. To mitigate this, we propose a multi-domain learning network to transfer knowledge across different but related domains and alleviate isolated learning paradigms using a shared representation. The proposed network consists of two identical causal convolutional sub-networks that are projected to a shared representation followed by a linear attention mechanism. The proposed network can be trained using the full training dataset of the source domain and a dataset of restricted size of the target training domain to reduce the need of large labelled training datasets. The network processes the source and target domains jointly to learn powerful and mutually complementary features to boost the performance in both domains. The proposed multi-domain learning network on six real-world sensor activity datasets outperforms the existing methods by applying only 50% of the labelled data. This confirms the efficacy of the proposed approach as a generic model to learn human activities from different but related domains in a joint effort, to reduce the number of required models and thus improve system efficiency.

Comprehensive study of semi-supervised learning for DNA methylation-based supervised classification of central nervous system tumors.

BackgroundPrecision medicine for cancer treatment relies on an accurate pathological diagnosis. The number of known tumor classes has increased rapidly, and reliance on traditional methods of histopathologic classification alone has become unfeasible. To help reduce variability, validation costs, and standardize the histopathological diagnostic process, supervised machine learning models using DNA-methylation data have been developed for tumor classification. These methods require large labeled training data sets to obtain clinically acceptable classification accuracy. While there is abundant unlabeled epigenetic data across multiple databases, labeling pathology data for machine learning models is time-consuming and resource-intensive, especially for rare tumor types. Semi-supervised learning (SSL) approaches have been used to maximize the utility of labeled and unlabeled data for classification tasks and are effectively applied in genomics. SSL methods have not yet been explored with epigenetic data nor demonstrated beneficial to central nervous system (CNS) tumor classification.ResultsThis paper explores the application of semi-supervised machine learning on methylation data to improve the accuracy of supervised learning models in classifying CNS tumors. We comprehensively evaluated 11 SSL methods and developed a novel combination approach that included a self-training with editing using support vector machine (SETRED-SVM) model and an L2-penalized, multinomial logistic regression model to obtain high confidence labels from a few labeled instances. Results across eight random forest and neural net models show that the pseudo-labels derived from our SSL method can significantly increase prediction accuracy for 82 CNS tumors and 9 normal controls.ConclusionsThe proposed combination of semi-supervised technique and multinomial logistic regression holds the potential to leverage the abundant publicly available unlabeled methylation data effectively. Such an approach is highly beneficial in providing additional training examples, especially for scarce tumor types, to boost the prediction accuracy of supervised models.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12859-022-04764-1.

GFSNet: Generalization-friendly siamese network for thermal infrared object tracking

The lack of large labeled training datasets hinders the usage of deep neural network for Thermal Infrared (TIR) tracking. Regular practice is to train a tracking network with large-scale RGB datasets and then retrain it to the TIR domain with limited TIR data. However, we observe that existing Siamese-based trackers can hardly generalize to TIR images though they achieve outstanding performance on RGB tracking. Therefore, the main challenge is the generalization problem: How to design a generalization-friendly Siamese tracking network and what affects the network generalization. To tackle this problem, we introduce the self-adaption structure into Siamese network and propose an effective TIR tracking model, GFSNet. GFSNet is successfully generalized to different TIR tracking tasks, including ground target, aircraft and high-diversity object tracking tasks. To estimate generalization ability, we present a notion of Growth Rate, the improvement of overall performance after retraining. Experimental results show that the Growth Rates of GFSNet exceed state-of-the-art SiamRPN++ by more than 7 times, which indicates the great power of GFSNet in generalization. In addition to experimental validations, we provide the theoretical analysis of network generalization from a novel perspective, model sensitivity. By performing some tests to analyze the sensitivity, we conclude that the self-adaption structure helps GFSNet converge to a more sensitive minimum with better generalization to new tasks. Furthermore, when compared with popular tracking methods, GFSNet maintains comparable accuracy while achieving real-time tracking with the speed of 112 FPS, 5 times faster than other TIR trackers.

Real Time Landmark Detection for Within- and Cross Subject Tracking With Minimal Human Supervision

Landmark detection plays an important role for a variety of image processing and analysis tasks. Current methods rely on either supervised or semi-supervised learning which often requires large labeled training datasets. Also, retrospective addition of further target landmarks after completion of training is difficult in current methods. In this paper we propose a framework that addresses these limitations and allows for landmark detection based on only few examples and for definition of target landmarks after completed training without retraining. Our proposed approach relies on self-supervised training on a within-image template matching task with regularization by data augmentation. The trained network generalizes to cross-image matching and can thus be extended to example-based landmark detection and tracking. We extensively evaluate the proposed framework on chest X-ray images and abdominal MRI scans and demonstrate high accuracy with only few or even only one labeled example. Additionally we apply it to the task of liver and liver lesion tracking in CINE MRI scans.

Using self-supervised feature learning to improve the use of pulse oximeter signals to predict paediatric hospitalization

Background: The success of many machine learning applications depends on knowledge about the relationship between the input data and the task of interest (output), hindering the application of machine learning to novel tasks. End-to-end deep learning, which does not require intermediate feature engineering, has been recommended to overcome this challenge but end-to-end deep learning models require large labelled training data sets often unavailable in many medical applications. In this study, we trained machine learning models to predict paediatric hospitalization given raw photoplethysmography (PPG) signals obtained from a pulse oximeter. We trained self-supervised learning (SSL) for automatic feature extraction from PPG signals and assessed the utility of SSL in initializing end-to-end deep learning models trained on a small labelled data set with the aim of predicting paediatric hospitalization.Methods: We compared logistic regression models fitted using features extracted using SSL with end-to-end deep learning models initialized either randomly or using weights from the SSL model. We also compared the performance of SSL models trained on labelled data alone (n=1,031) with SSL trained using both labelled and unlabelled signals (n=7,578). Results: The SSL model trained on both labelled and unlabelled PPG signals produced features that were more predictive of hospitalization compared to the SSL model trained on labelled PPG only (AUC of logistic regression model: 0.78 vs 0.74). The end-to-end deep learning model had an AUC of 0.80 when initialized using the SSL model trained on all PPG signals, 0.77 when initialized using SSL trained on labelled data only, and 0.73 when initialized randomly. Conclusions: This study shows that SSL can improve the classification of PPG signals by either extracting features required by logistic regression models or initializing end-to-end deep learning models. Furthermore, SSL can leverage larger unlabelled data sets to improve performance of models fitted using small labelled data sets.

Semi-supervised classification of radiology images with NoTeacher: A teacher that is not mean

Deep learning models achieve strong performance for radiology image classification, but their practical application is bottlenecked by the need for large labeled training datasets. Semi-supervised learning (SSL) approaches leverage small labeled datasets alongside larger unlabeled datasets and offer potential for reducing labeling cost. In this work, we introduce NoTeacher, a novel consistency-based SSL framework which incorporates probabilistic graphical models. Unlike Mean Teacher which maintains a teacher network updated via a temporal ensemble, NoTeacher employs two independent networks, thereby eliminating the need for a teacher network. We demonstrate how NoTeacher can be customized to handle a range of challenges in radiology image classification. Specifically, we describe adaptations for scenarios with 2D and 3D inputs, with uni and multi-label classification, and with class distribution mismatch between labeled and unlabeled portions of the training data. In realistic empirical evaluations on three public benchmark datasets spanning the workhorse modalities of radiology (X-Ray, CT, MRI), we show that NoTeacher achieves over 90–95% of the fully supervised AUROC with less than 5–15% labeling budget. Further, NoTeacher outperforms established SSL methods with minimal hyperparameter tuning, and has implications as a principled and practical option for semi-supervised learning in radiology applications.

Intelligent Condition-Based Monitoring of Rotary Machines With Few Samples

Recently, intelligent condition based monitoring systems build on deep learning methods have gained popularity. The success of these methods relies upon the large labeled training datasets, which are crucial to collect in industries. Therefore, building an effective fault diagnosis system becomes challenging. In this paper, a novel fault diagnosis framework is proposed to tackle the issue of limited samples in the training dataset. In the proposed framework, firstly, new training samples termed as synthetic samples are generated to increase the size of the dataset. After that, both original and synthetic samples are stacked, and the classifier model is trained. This study proposes a modified Conditional Variational Autoencoder (CVAE) to generate synthetic samples. In the proposed CVAE, centroid loss is added to the standard CVAE objective function. This loss directs generated samples to remain close with the centroid of their respective class, which helps in generating synthetic samples quite similar to the original samples. This paper also investigates the performance of proposed model in the presence of noise and effect of transformed data and original data. To verify the effectiveness of the proposed approach, the Air compressor and Case Western Reserve University datasets have been investigated. For the CWRU dataset with only 80 samples, accuracy of 96.39%, 99.58%, 98.33% was obtained using multilayer neural network, support vector machine, and RF classifiers respectively. Classification accuracy increased to 63.33% when modified CVAE is used instead of standard CVAE. Finally, a comparative analysis between proposed methods and other state-of-the-art methods has been presented.

Improved automatic detection of herpesvirus secondary envelopment stages in electron microscopy by augmenting training data with synthetic labelled images generated by a generative adversarial network.

Detailed analysis of secondary envelopment of the herpesvirus human cytomegalovirus (HCMV) by transmission electron microscopy (TEM) is crucial for understanding the formation of infectious virions. Here, we present a convolutional neural network (CNN) that automatically recognises cytoplasmic capsids and distinguishes between three HCMV capsid envelopment stages in TEM images. 315 TEM images containing 2,610 expert-labelled capsids of the three classes were available for CNN training. To overcome the limitation of small training datasets and thus poor CNN performance, we used a deep learning method, the generative adversarial network (GAN), to automatically increase our labelled training dataset with 500 synthetic images and thus to 9,192 labelled capsids. The synthetic TEM images were added to the ground truth dataset to train the Faster R-CNN deep learning-based object detector. Training with 315 ground truth images yielded an average precision (AP) of 53.81% for detection, whereas the addition of 500 synthetic training images increased the AP to 76.48%. This shows that generation and additional use of synthetic labelled images for detector training is an inexpensive way to improve detector performance. This work combines the gold standard of secondary envelopment research with state-of-the-art deep learning technology to speed up automatic image analysis even when large labelled training datasets are not available.

Weak supervision as an efficient approach for automated seizure detection in electroencephalography

Automated seizure detection from electroencephalography (EEG) would improve the quality of patient care while reducing medical costs, but achieving reliably high performance across patients has proven difficult. Convolutional Neural Networks (CNNs) show promise in addressing this problem, but they are limited by a lack of large labeled training datasets. We propose using imperfect but plentiful archived annotations to train CNNs for automated, real-time EEG seizure detection across patients. While these weak annotations indicate possible seizures with precision scores as low as 0.37, they are commonly produced in large volumes within existing clinical workflows by a mixed group of technicians, fellows, students, and board-certified epileptologists. We find that CNNs trained using such weak annotations achieve Area Under the Receiver Operating Characteristic curve (AUROC) values of 0.93 and 0.94 for pediatric and adult seizure onset detection, respectively. Compared to currently deployed clinical software, our model provides a 31% increase (18 points) in F1-score for pediatric patients and a 17% increase (11 points) for adult patients. These results demonstrate that weak annotations, which are sustainably collected via existing clinical workflows, can be leveraged to produce clinically useful seizure detection models.

AppendiXNet: Deep Learning for Diagnosis of Appendicitis from A Small Dataset of CT Exams Using Video Pretraining

The development of deep learning algorithms for complex tasks in digital medicine has relied on the availability of large labeled training datasets, usually containing hundreds of thousands of examples. The purpose of this study was to develop a 3D deep learning model, AppendiXNet, to detect appendicitis, one of the most common life-threatening abdominal emergencies, using a small training dataset of less than 500 training CT exams. We explored whether pretraining the model on a large collection of natural videos would improve the performance of the model over training the model from scratch. AppendiXNet was pretrained on a large collection of YouTube videos called Kinetics, consisting of approximately 500,000 video clips and annotated for one of 600 human action classes, and then fine-tuned on a small dataset of 438 CT scans annotated for appendicitis. We found that pretraining the 3D model on natural videos significantly improved the performance of the model from an AUC of 0.724 (95% CI 0.625, 0.823) to 0.810 (95% CI 0.725, 0.895). The application of deep learning to detect abnormalities on CT examinations using video pretraining could generalize effectively to other challenging cross-sectional medical imaging tasks when training data is limited.

A semi-supervised machine learning framework for microRNA classification

BackgroundMicroRNAs (miRNAs) are a family of short, non-coding RNAs that have been linked to critical cellular activities, most notably regulation of gene expression. The identification of miRNA is a cross-disciplinary approach that requires both computational identification methods and wet-lab validation experiments, making it a resource-intensive procedure. While numerous machine learning methods have been developed to increase classification accuracy and thus reduce validation costs, most methods use supervised learning and thus require large labeled training data sets, often not feasible for less-sequenced species. On the other hand, there is now an abundance of unlabeled RNA sequence data due to the emergence of high-throughput wet-lab experimental procedures, such as next-generation sequencing.ResultsThis paper explores the application of semi-supervised machine learning for miRNA classification in order to maximize the utility of both labeled and unlabeled data. We here present the novel combination of two semi-supervised approaches: active learning and multi-view co-training. Results across six diverse species show that this multi-stage semi-supervised approach is able to improve classification performance using very small numbers of labeled instances, effectively leveraging the available unlabeled data.ConclusionsThe proposed semi-supervised miRNA classification pipeline holds the potential to identify novel miRNA with high recall and precision while requiring very small numbers of previously known miRNA. Such a method could be highly beneficial when studying miRNA in newly sequenced genomes of niche species with few known examples of miRNA.

Multi-view Co-training for microRNA Prediction

MicroRNA (miRNA) are short, non-coding RNAs involved in cell regulation at post-transcriptional and translational levels. Numerous computational predictors of miRNA been developed that generally classify miRNA based on either sequence- or expression-based features. While these methods are highly effective, they require large labelled training data sets, which are often not available for many species. Simultaneously, emerging high-throughput wet-lab experimental procedures are producing large unlabelled data sets of genomic sequence and RNA expression profiles. Existing methods use supervised machine learning and are therefore unable to leverage these unlabelled data. In this paper, we design and develop a multi-view co-training approach for the classification of miRNA to maximize the utility of unlabelled training data by taking advantage of multiple views of the problem. Starting with only 10 labelled training data, co-training is shown to significantly (p < 0.01) increase classification accuracy of both sequence- and expression-based classifiers, without requiring any new labelled training data. After 11 iterations of co-training, the expression-based view of miRNA classification experiences an average increase in AUPRC of 15.81% over six species, compared to 11.90% for self-training and 4.84% for passive learning. Similar results are observed for sequence-based classifiers with increases of 46.47%, 39.53% and 29.43%, for co-training, self-training, and passive learning, respectively. The final co-trained sequence and expression-based classifiers are integrated into a final confidence-based classifier which shows improved performance compared to both the expression (1.5%, p = 0.021) and sequence (3.7%, p = 0.006) views. This study represents the first application of multi-view co-training to miRNA prediction and shows great promise, particularly for understudied species with few available training data.

Remaining useful life predictions for turbofan engine degradation using semi-supervised deep architecture

In recent years, research has proposed several deep learning (DL) approaches to providing reliable remaining useful life (RUL) predictions in Prognostics and Health Management (PHM) applications. Although supervised DL techniques, such as Convolutional Neural Network and Long-Short Term Memory, have outperformed traditional prognosis algorithms, they are still dependent on large labeled training datasets. With respect to real-life PHM applications, high-quality labeled training data might be both challenging and time-consuming to acquire. Alternatively, unsupervised DL techniques introduce an initial pre-training stage to extract degradation related features from raw unlabeled training data automatically. Thus, the combination of unsupervised and supervised (semi-supervised) learning has the potential to provide high RUL prediction accuracy even with reduced amounts of labeled training data. This paper investigates the effect of unsupervised pre-training in RUL predictions utilizing a semi-supervised setup. Additionally, a Genetic Algorithm (GA) approach is applied in order to tune the diverse amount of hyper-parameters in the training procedure. The advantages of the proposed semi-supervised setup have been verified on the popular C-MAPSS dataset. The experimental study, compares this approach to purely supervised training, both when the training data is completely labeled and when the labeled training data is reduced, and to the most robust results in the literature. The results suggest that unsupervised pre-training is a promising feature in RUL predictions subjected to multiple operating conditions and fault modes.

A Bayesian Scene-Prior-Based Deep Network Model for Face Verification.

Face recognition/verification has received great attention in both theory and application for the past two decades. Deep learning has been considered as a very powerful tool for improving the performance of face recognition/verification recently. With large labeled training datasets, the features obtained from deep learning networks can achieve higher accuracy in comparison with shallow networks. However, many reported face recognition/verification approaches rely heavily on the large size and complete representative of the training set, and most of them tend to suffer serious performance drop or even fail to work if fewer training samples per person are available. Hence, the small number of training samples may cause the deep features to vary greatly. We aim to solve this critical problem in this paper. Inspired by recent research in scene domain transfer, for a given face image, a new series of possible scenarios about this face can be deduced from the scene semantics extracted from other face individuals in a face dataset. We believe that the “scene” or background in an image, that is, samples with more different scenes for a given person, may determine the intrinsic features among the faces of the same individual. In order to validate this belief, we propose a Bayesian scene-prior-based deep learning model in this paper with the aim to extract important features from background scenes. By learning a scene model on the basis of a labeled face dataset via the Bayesian idea, the proposed method transforms a face image into new face images by referring to the given face with the learnt scene dictionary. Because the new derived faces may have similar scenes to the input face, the face-verification performance can be improved without having background variance, while the number of training samples is significantly reduced. Experiments conducted on the Labeled Faces in the Wild (LFW) dataset view #2 subset illustrated that this model can increase the verification accuracy to 99.2% by means of scenes’ transfer learning (99.12% in literature with an unsupervised protocol). Meanwhile, our model can achieve 94.3% accuracy for the YouTube Faces database (DB) (93.2% in literature with an unsupervised protocol).