Strong Labels Research Articles

Optimal number of strong labels for curriculum learning with convolutional neural network to classify pulmonary abnormalities in chest radiographs.

It is important to alleviate annotation efforts and costs by efficiently training on medical images. We performed a stress test on several strong labels for curriculum learning with a convolutional neural network to differentiate normal and five types of pulmonary abnormalities in chest radiograph images. The numbers of CXR images of healthy subjects and patients, acquired at Asan Medical Center (AMC), were 6069 and 3465, respectively. The numbers of CXR images of patients with nodules, consolidation, interstitial opacity, pleural effusion, and pneumothorax were 944, 550, 280, 1360, and 331, respectively. The AMC dataset was split into training, tuning, and test, with a ratio of 7:1:2. All lesions were strongly labeled by thoracic expert radiologists, with confirmation of the corresponding CT. For curriculum learning, normal and abnormal patches (N=26658) were randomly extracted around the normal lung and strongly labeled abnormal lesions, respectively. In addition, 1%, 5%, 20%, 50%, and 100% of strong labels were used to determine an optimal number for them. Each patch dataset was trained with the ResNet-50 architecture, and all CXRs with weak labels were used for fine-tuning them in a transfer-learning manner. A dataset acquired from the Seoul National University Bundang Hospital (SNUBH) was used for external validation. The detection accuracies of the 1%, 5%, 20%, 50%, and 100% datasets were 90.51, 92.15, 93.90, 94.54, and 95.39, respectively, in the AMC dataset and 90.01, 90.14, 90.97, 91.92, and 93.00 in the SNUBH dataset. Our results showed that curriculum learning with over 20% sampling rate for strong labels are sufficient to train a model with relatively high performance, which can be easily and efficiently developed in an actual clinical setting.

Combining weakly and strongly supervised learning improves strong supervision in Gleason pattern classification

BackgroundOne challenge to train deep convolutional neural network (CNNs) models with whole slide images (WSIs) is providing the required large number of costly, manually annotated image regions. Strategies to alleviate the scarcity of annotated data include: using transfer learning, data augmentation and training the models with less expensive image-level annotations (weakly-supervised learning). However, it is not clear how to combine the use of transfer learning in a CNN model when different data sources are available for training or how to leverage from the combination of large amounts of weakly annotated images with a set of local region annotations. This paper aims to evaluate CNN training strategies based on transfer learning to leverage the combination of weak and strong annotations in heterogeneous data sources. The trade-off between classification performance and annotation effort is explored by evaluating a CNN that learns from strong labels (region annotations) and is later fine-tuned on a dataset with less expensive weak (image-level) labels.ResultsAs expected, the model performance on strongly annotated data steadily increases as the percentage of strong annotations that are used increases, reaching a performance comparable to pathologists (kappa = 0.691 pm 0.02). Nevertheless, the performance sharply decreases when applied for the WSI classification scenario with kappa = 0.307 pm 0.133. Moreover, it only provides a lower performance regardless of the number of annotations used. The model performance increases when fine-tuning the model for the task of Gleason scoring with the weak WSI labels kappa = 0.528 pm 0.05.ConclusionCombining weak and strong supervision improves strong supervision in classification of Gleason patterns using tissue microarrays (TMA) and WSI regions. Our results contribute very good strategies for training CNN models combining few annotated data and heterogeneous data sources. The performance increases in the controlled TMA scenario with the number of annotations used to train the model. Nevertheless, the performance is hindered when the trained TMA model is applied directly to the more challenging WSI classification problem. This demonstrates that a good pre-trained model for prostate cancer TMA image classification may lead to the best downstream model if fine-tuned on the WSI target dataset. We have made available the source code repository for reproducing the experiments in the paper: https://github.com/ilmaro8/Digital_Pathology_Transfer_Learning

Your rugby mates don't need to know your colleagues: Triadic closure with edge colors

Given an undirected graph G=(V,E) the NP-hard Strong Triadic Closure (STC) problem asks for a labeling of the edges as weak and strong such that at most k edges are weak and for each induced P3 in G at least one edge is weak. We study the following generalizations of STC with c different strong edge colors. In Multi-STC an induced P3 may receive two strong labels as long as they are different. In Edge-List Multi-STC and Vertex-List Multi-STC we may restrict the set of permitted colors for each edge of G. We show that, under the Exponential Time Hypothesis (ETH), Edge-List Multi-STC and Vertex-List Multi-STC cannot be solved in time 2o(|V|2). We proceed with a parameterized complexity analysis in which we extend previous algorithms and kernelizations for STC [11,14] to the three variants or outline the limits of such an extension.

Semi-Supervised NMF-CNN for Sound Event Detection

The lack of strongly labeled data can limit the potential of a Sound Event Detection (SED) system trained using deep learning approaches. To address this issue, this paper proposes a novel method to approximate strong labels for the weakly labeled data using Nonnegative Matrix Factorization (NMF) in a supervised manner. Using a combinative transfer learning and semi-supervised learning framework, two different Convolutional Neural Networks (CNN) are trained using synthetic data, approximated strongly labeled data, and unlabeled data where one model will produce the audio tags. In contrast, the other will produce the frame-level prediction. The proposed methodology is then evaluated on three different subsets of the Detection and Classification of Acoustic Scenes and Events (DCASE) 2020 dataset: validation dataset, challenge evaluation dataset, and public YouTube evaluation dataset. Based on the results, our proposed methodology outperforms the baseline system by a minimum of 7% across these three different data subsets. In addition, our proposed method also outperforms the top 3 submissions from the DCASE 2019 challenge task 4 on the validation and public YouTube evaluation datasets. Our system performance is also competitive against the top submission in DCASE 2020 challenge task 4 on the challenge evaluation data. A post-challenge analysis was also performed using the validation dataset, which revealed the causes of the performance difference between our system and the top submission of the DCASE 2020 challenge task 4. The leading causes that we observed are 1) detection threshold tuning method and 2) augmentation techniques used. We observed that our system could perform better than the first place submission by 1.5% by changing our detection threshold tuning method. In addition, the post-challenge analysis also revealed that our system is more robust than the top submission in DCASE 2020 challenge task 4 on long-duration audio clips, where we outperformed them by 37%.

Semi-Supervised Remote Sensing Image Semantic Segmentation via Consistency Regularization and Average Update of Pseudo-Label

Image segmentation has made great progress in recent years, but the annotation required for image segmentation is usually expensive, especially for remote sensing images. To solve this problem, we explore semi-supervised learning methods and appropriately utilize a large amount of unlabeled data to improve the performance of remote sensing image segmentation. This paper proposes a method for remote sensing image segmentation based on semi-supervised learning. We first design a Consistency Regularization (CR) training method for semi-supervised training, then employ the new learned model for Average Update of Pseudo-label (AUP), and finally combine pseudo labels and strong labels to train semantic segmentation network. We demonstrate the effectiveness of the proposed method on three remote sensing datasets, achieving better performance without more labeled data. Extensive experiments show that our semi-supervised method can learn the latent information from the unlabeled data to improve the segmentation performance.

Accurate segmentation of prostate cancer histomorphometric features using a weakly supervised convolutional neural network.

.Purpose: Prostate cancer primarily arises from the glandular epithelium. Histomophometric techniques have been used to assess the glandular epithelium in automated detection and classification pipelines; however, they are often rigid in their implementation, and their performance suffers on large datasets where variation in staining, imaging, and preparation is difficult to control. The purpose of this study is to quantify performance of a pixelwise segmentation algorithm that was trained using different combinations of weak and strong stroma, epithelium, and lumen labels in a prostate histology dataset.Approach: We have combined weakly labeled datasets generated using simple morphometric techniques and high-quality labeled datasets from human observers in prostate biopsy cores to train a convolutional neural network for use in whole mount prostate labeling pipelines. With trained networks, we characterize pixelwise segmentation of stromal, epithelium, and lumen (SEL) regions on both biopsy core and whole-mount H&E-stained tissue.Results: We provide evidence that by simply training a deep learning algorithm on weakly labeled data generated from rigid morphometric methods, we can improve the robustness of classification over the morphometric methods used to train the classifier.Conclusions: We show that not only does our approach of combining weak and strong labels for training the CNN improve qualitative SEL labeling within tissue but also the deep learning generated labels are superior for cancer classification in a higher-order algorithm over the morphometrically derived labels it was trained on.

Maximizing the strong triadic closure in split graphs and proper interval graphs

In social networks the Strong Triadic Closure is an assignment of the edges with strong or weak labels such that any two vertices that have a common neighbor with a strong edge are adjacent. The problem of maximizing the number of strong edges that satisfy the strong triadic closure was recently shown to be NP-complete for general graphs. Here we initiate the study of graph classes for which the problem is solvable. We show that the problem admits a polynomial-time algorithm for two incomparable classes of graphs: proper interval graphs and trivially-perfect graphs. To complement our result, we show that the problem remains NP-complete on split graphs, and consequently also on chordal graphs. Thus, we contribute to define the first border between graph classes on which the problem is polynomially solvable and on which it remains NP-complete.

3D-BoxSup: Positive-Unlabeled Learning of Brain Tumor Segmentation Networks From 3D Bounding Boxes.

Accurate segmentation is an essential task when working with medical images. Recently, deep convolutional neural networks achieved a state-of-the-art performance for many segmentation benchmarks. Regardless of the network architecture, the deep learning-based segmentation methods view the segmentation problem as a supervised task that requires a relatively large number of annotated images. Acquiring a large number of annotated medical images is time consuming, and high-quality segmented images (i.e., strong labels) crafted by human experts are expensive. In this paper, we have proposed a method that achieves competitive accuracy from a “weakly annotated” image where the weak annotation is obtained via a 3D bounding box denoting an object of interest. Our method, called “3D-BoxSup,” employs a positive-unlabeled learning framework to learn segmentation masks from 3D bounding boxes. Specially, we consider the pixels outside of the bounding box as positively labeled data and the pixels inside the bounding box as unlabeled data. Our method can suppress the negative effects of pixels residing between the true segmentation mask and the 3D bounding box and produce accurate segmentation masks. We applied our method to segment a brain tumor. The experimental results on the BraTS 2017 dataset (Menze et al., 2015; Bakas et al., 2017a,b,c) have demonstrated the effectiveness of our method.

Active learning with point supervision for cost-effective panicle detection in cereal crops

BackgroundPanicle density of cereal crops such as wheat and sorghum is one of the main components for plant breeders and agronomists in understanding the yield of their crops. To phenotype the panicle density effectively, researchers agree there is a significant need for computer vision-based object detection techniques. Especially in recent times, research in deep learning-based object detection shows promising results in various agricultural studies. However, training such systems usually requires a lot of bounding-box labeled data. Since crops vary by both environmental and genetic conditions, acquisition of huge amount of labeled image datasets for each crop is expensive and time-consuming. Thus, to catalyze the widespread usage of automatic object detection for crop phenotyping, a cost-effective method to develop such automated systems is essential.ResultsWe propose a point supervision based active learning approach for panicle detection in cereal crops. In our approach, the model constantly interacts with a human annotator by iteratively querying the labels for only the most informative images, as opposed to all images in a dataset. Our query method is specifically designed for cereal crops which usually tend to have panicles with low variance in appearance. Our method reduces labeling costs by intelligently leveraging low-cost weak labels (object centers) for picking the most informative images for which strong labels (bounding boxes) are required. We show promising results on two publicly available cereal crop datasets—Sorghum and Wheat. On Sorghum, 6 variants of our proposed method outperform the best baseline method with more than 55% savings in labeling time. Similarly, on Wheat, 3 variants of our proposed methods outperform the best baseline method with more than 50% of savings in labeling time.ConclusionWe proposed a cost effective method to train reliable panicle detectors for cereal crops. A low cost panicle detection method for cereal crops is highly beneficial to both breeders and agronomists. Plant breeders can obtain quick crop yield estimates to make important crop management decisions. Similarly, obtaining real time visual crop analysis is valuable for researchers to analyze the crop’s response to various experimental conditions.

PartMitosis: A Partially Supervised Deep Learning Framework for Mitosis Detection in Breast Cancer Histopathology Images

Detection of mitotic tumor cells per tissue area is one of the critical markers of breast cancer prognosis. The aim of this paper is to develop a method for the automatic detection of mitotic figures from breast cancer histological slides using a partially supervised deep learning framework. Unlike the previous literature, which has focused on solving the problem of mitosis detection in the weakly annotated datasets using centroid pixel labels (weak labels) only without taking advantage of the available pixel-level labels (strong labels) of other datasets, in this paper, we design a novel partially supervised framework based on two parallel deep fully convolutional networks. One of them is trained using weak labels and the other is trained using strong labels, together with a weight transfer function. In the detection phase, we fuse the segmentation maps produced by the two networks to obtain the final mitosis detections. Our system exploits the available large sets of mitosis detection samples with mitosis centroid annotation, such as the 2014 ICPR dataset and the AMIDA13 dataset, and only a small set of samples with the annotation of all mitosis pixels, such as the 2012 ICPR dataset, to perform a more accurate mitosis detection on weakly labeled data. This enables us to outperform all previous mitosis detection systems by achieving F-scores of 0.575 and 0.698 on the 2014 ICPR dataset and the AMIDA13 dataset respectively.

Digging Into Pseudo Label: A Low-Budget Approach for Semi-Supervised Semantic Segmentation

The capability to understand visual scenes with limited labeled data has been widely concerned in the field of computer vision. Although semi-supervised learning for image classification has been extensively studied in some cases, semantic segmentation with limited data has only recently gained attention. In this work, we follow the standard semi-supervised segmentation pipeline for image classification and propose a two-branch network that can encode strong and pseudo label spaces respectively, extracting reliable supervision information from pseudo-labels to assist in training network with strong labels. Our method outperforms previous semi-supervised methods with limited annotation cost. On standard benchmark PASCAL VOC 2012 for semi-supervised semantic segmentation, the proposed approach gains fresh state-of-the-art performance.

Deep Convolutional Neural Network with Structured Prediction for Weakly Supervised Audio Event Detection

Audio event detection (AED) is a task of recognizing the types of audio events in an audio stream and estimating their temporal positions. AED is typically based on fully supervised approaches, requiring strong labels including both the presence and temporal position of each audio event. However, fully supervised datasets are not easily available due to the heavy cost of human annotation. Recently, weakly supervised approaches for AED have been proposed, utilizing large scale datasets with weak labels including only the occurrence of events in recordings. In this work, we introduce a deep convolutional neural network (CNN) model called DSNet based on densely connected convolution networks (DenseNets) and squeeze-and-excitation networks (SENets) for weakly supervised training of AED. DSNet alleviates the vanishing-gradient problem and strengthens feature propagation and models interdependencies between channels. We also propose a structured prediction method for weakly supervised AED. We apply a recurrent neural network (RNN) based framework and a prediction smoothness cost function to consider long-term contextual information with reduced error propagation. In post-processing, conditional random fields (CRFs) are applied to take into account the dependency between segments and delineate the borders of audio events precisely. We evaluated our proposed models on the DCASE 2017 task 4 dataset and obtained state-of-the-art results on both audio tagging and event detection tasks.

Ag/CdO NP-Engineered Magnetic Electrochemical Aptasensor for Prostatic Specific Antigen Detection.

A simple magnetic electrochemical aptasensor was established for the detection of prostatic specific antigen (PSA). Ag/CdO nanoparticles (NPs) were fabricated and exhibited strong electroreduction peaks at -1.07 V, attributing to the electron transfer from Cd2+ to Cd0 and the superior electron transportation of Ag. Aptamer-modified Ag/CdO NPs were assembled on the surface of superparamagnetic Fe3O4/graphene oxide nanosheets (GO/Fe3O4 NSs) through the hydrophobic and π-π stacking interaction of aptamers and GO NSs. These assemblies possessed superior electroactive properties, efficient electron transfer, and superparamagnetic response and could serve as sensing units for PSA detection with the aid of a magnetic electrode. With increasing concentrations of PSA, the high affinity of aptamers to PSA enabled the dissociation of Ag/CdO NPs from GO/Fe3O4 NSs, decreasing the intensity of electroreduction peaks. The Ag/CdO NP-engineered magnetic electrochemical aptasensor achieved sensitive and accurate detection of PSA in the range of 50 pg/mL to 50 ng/mL. The limit of detection (LOD) was as low as 28 pg/mL. This developed protocol can be extended to a large set of strong electroactive labels for reliable tumor biomarker detection with high sensitivity and specificity.

Impact of tobacco-related health warning labels across socioeconomic, race and ethnic groups: results from a randomized web-based experiment.

BackgroundThe U.S. Family Smoking Prevention and Tobacco Control Act of 2009 requires updating of the existing text-only health warning labels on tobacco packaging with nine new warning statements accompanied by pictorial images. Survey and experimental research in the U.S. and other countries supports the effectiveness of pictorial health warning labels compared with text-only warnings for informing smokers about the risks of smoking and encouraging cessation. Yet very little research has examined differences in reactions to warning labels by race/ethnicity, education or income despite evidence that population subgroups may differ in their ability to process health information. The purpose of the present study was to evaluate the potential impact of pictorial warning labels compared with text-only labels among U.S. adult smokers from diverse racial/ethnic and socioeconomic subgroups.Methods/FindingsParticipants were adult smokers recruited from two online research panels (n = 3,371) into a web-based experimental study to view either the new pictorial warnings or text-only warnings. Participants viewed the labels and reported their reactions. Adjusted regression models demonstrated significantly stronger reactions for the pictorial condition for each outcome salience (b = 0.62, p<.001); perceived impact (b = 0.44, p<.001); credibility (OR = 1.41, 95% CI = 1.22−1.62), and intention to quit (OR = 1.30, 95% CI = 1.10−1.53). No significant results were found for interactions between condition and race/ethnicity, education, or income. The only exception concerned the intention to quit outcome, where the condition-by-education interaction was nearly significant (p = 0.057).ConclusionsFindings suggest that the greater impact of the pictorial warning label compared to the text-only warning is consistent across diverse racial/ethnic and socioeconomic populations. Given their great reach, pictorial health warning labels may be one of the few tobacco control policies that have the potential to reduce communication inequalities across groups. Policies that establish strong pictorial warning labels on tobacco packaging may be instrumental in reducing the toll of the tobacco epidemic, particularly within vulnerable communities.

Cost-Sensitive Active Visual Category Learning

We present an active learning framework that predicts the tradeoff between the effort and information gain associated with a candidate image annotation, thereby ranking unlabeled and partially labeled images according to their expected "net worth" to an object recognition system. We develop a multi-label multiple-instance approach that accommodates realistic images containing multiple objects and allows the category-learner to strategically choose what annotations it receives from a mixture of strong and weak labels. Since the annotation cost can vary depending on an image's complexity, we show how to improve the active selection by directly predicting the time required to segment an unlabeled image. Our approach accounts for the fact that the optimal use of manual effort may call for a combination of labels at multiple levels of granularity, as well as accurate prediction of manual effort. As a result, it is possible to learn more accurate category models with a lower total expenditure of annotation effort. Given a small initial pool of labeled data, the proposed method actively improves the category models with minimal manual intervention.

Effective Interventions to Reduce Smoking-Induced Heart Disease Around the World

Tobacco is a leading cause of preventable death worldwide. For 2002, the World Health Organization (WHO) estimated that 5 million deaths were attributable to smoking.1 This issue of Circulation contains a detailed analysis of the 1.62 million deaths that smoking contributes to global and regional cause-specific mortality from cardiovascular disease.2 Ezzati et al estimate that >1 in 10 cardiovascular deaths (11%) is caused by smoking. More deaths occurred in men than in women and in industrialized than in developing countries. Nevertheless, smoking is an important contributor to cardiac deaths (17% for men and 5% for women) in both industrialized and developing countries. See p 489 The implications of the results of the study by Ezzati et al are simple and straightforward. If smoking did not exist, then 1.62 million cardiovascular deaths would not have occurred. Even in regions where the number of deaths attributable to smoking was low, thousands of deaths could be prevented. In addition, the finding that regions with a longer history of smoking had the largest percentage of total cardiovascular mortality attributable to smoking is a warning sign for countries in which the tobacco industry is just beginning to spread the smoking epidemic. Cardiovascular health can be improved quickly through sound public health measures that have an impact at a population level.3 Smoke-free workplaces and public places, comprehensive advertising bans, increased taxation on tobacco products, and strong graphic warning labels on cigarette packs quickly reduce tobacco consumption3,4 and, thus, cardiovascular disease. For example, in California, a strong antitobacco program that stressed adherence to laws banning or limiting smoking and exposed the tobacco industry was associated with 59 000 fewer heart disease deaths than would have been expected based on historical trends, ≈13% below the number of deaths expected after 8 years.5 More …

OVERVIEW OF BIOLOGIC FACTORS

Exploring the biology of neurotransmitters and polymorphic genes that influence behavior most certainly will increase understanding of the complexity of personality disorders and the people who suffer with them. Continuing biologic exploration of the personality disorders will also provide new and valuable information that will inform and direct ever more sophisticated and specific psychiatric treatments. Further, it will supply knowledge that might begin to erode the strong biases and negative labels that have for too long been applied to these patients. These labels may have been applied not because these patients are bad people, but because of ignorance in understanding their complexity and limited ability to improve their condition. Nonetheless, clinicians must refrain from concluding that the answers to all problems in psychiatric treatment can be answered by the molecular geneticist or biologic researcher. As Reiss et al have written when speaking about nonshared environmental effects and its role in the development of psychopathology: "Psychiatry has been forced into the chronically uncomfortable position of straddling biomedicine and the social sciences and seems always to hunger for relief.... [Yet] the data simply do not permit a conception of the future centered on a straightforward biomedical answer to the fundamental question of the pathogenesis of major disorders. Indeed, a balanced image of the future contains a growing and equal partnership of the social sciences and molecular biology."