Labeled Data Set Research Articles

DanceCaps: Pseudo-Captioning for Dance Videos Using Large Language Models

In recent years, the dance field has been able to create diverse content by leveraging technical advancements such as deep learning models, generating content beyond the unique artistic creations that only humans can create. However, in terms of dance data, there are still a lack of video and label datasets or datasets that contain multiple tags for videos. To address this gap, this paper explores the feasibility of generating dance captions from tags using a pseudo-captioning approach, inspired by the significant improvements large language models (LLMs) have shown in other domains. Various tags are generated from features extracted from videos and audio, and LLMs are then instructed to produce dance captions based on these tags. Captions were generated using both the open dance dataset and Internet dance videos, followed by user evaluations of randomly sampled captions. Participants found the captions effective in describing dance movements, of expert quality, and consistent with video content. Additionally, positive feedback was received on the evaluation of the gap in image extraction and the inclusion of tag data. This paper introduces and validates a novel pseudo-captioning method for generating dance captions using predefined tags, contributing to the expansion of data available for dance research and offering a practical solution to the current lack of datasets in this field.

Inversion of biophysical parameters of potato based on an active learning pool-based sampling strategy

ABSTRACT Timely estimations of leaf chlorophyll content (LCC) and leaf area index (LAI) can provide critical information for potato field management. We employed a hybrid method that integrated machine learning with a radiative transfer model to estimate potato growth parameters. A look-up table was generated using the PROSAIL model, which was used as an unlabelled sample set. Measurements were taken from a potato field, and the data were labelled according to growth period and variety. Then, training samples for potato LCC, LAI, and canopy chlorophyll content (CCC) were selected from the simulated unlabelled sample set using the Euclidean distance-based diversity algorithm based on different labelled data sets. The training sample size required to accurately estimate the parameters varied considerably depending on the parameter, variety, and growth period, despite using the same labelled data set. Moreover, our results indicate that the growth period has a substantial impact on model accuracy and needs to be considered when constructing the labelled data set. The study results indicate that the hybrid method combined with the radiative transfer model and active learning can effectively select informative training samples from a data pool and improve the accuracy of potato parameter estimation, which provides a valid tool for accurately monitoring crop growth and growth health.

GraphPI: Efficient Protein Inference with Graph Neural Networks.

The integration of deep learning approaches in biomedical research has been transformative, enabling breakthroughs in various applications. Despite these strides, its application in protein inference is impeded by the scarcity of extensively labeled data sets, a challenge compounded by the high costs and complexities of accurate protein annotation. In this study, we introduce GraphPI, a novel framework that treats protein inference as a node classification problem. We treat proteins as interconnected nodes within a protein-peptide-PSM graph, utilizing a graph neural network-based architecture to elucidate their interrelations. To address label scarcity, we train the model on a set of unlabeled public protein data sets with pseudolabels derived from an existing protein inference algorithm, enhanced by self-training to iteratively refine labels based on confidence scores. Contrary to prevalent methodologies necessitating data set-specific training, our research illustrates that GraphPI, due to the well-normalized nature of Percolator features, exhibits universal applicability without data set-specific fine-tuning, a feature that not only mitigates the risk of overfitting but also enhances computational efficiency. Our empirical experiments reveal notable performance on various test data sets and deliver significantly reduced computation times compared to common protein inference algorithms.

From Labquakes to Megathrusts: Scaling Deep Learning Based Pickers Over 15 Orders of Magnitude

AbstractThe application of machine learning techniques in seismology has greatly advanced seismological analysis, especially for earthquake detection and seismic phase picking. However, machine learning approaches still face challenges in generalizing to data sets that differ from their original training setting. Previous studies focused on retraining or transfer‐learning models for these scenarios, but require high‐quality labeled data sets. This paper demonstrates a new approach for augmenting already trained models without the need for additional training data. We propose four strategies—rescaling, model aggregation, shifting, and filtering—to enhance the performance of pre‐trained models on out‐of‐distribution data sets. We further devise various methodologies to ensemble the individual predictions from these strategies to obtain a final unified prediction result featuring prediction robustness and detection sensitivity. We develop an open‐source Python module quakephase that implements these methods and can flexibly process input continuous seismic data of any sampling rate. With quakephase and pre‐trained ML models from SeisBench, we perform systematic benchmark tests on data recorded by different types of instruments, ranging from acoustic emission sensors to distributed acoustic sensing, and collected at different scales, spanning from laboratory acoustic emission events to major tectonic earthquakes. Our tests highlight that rescaling is essential for dealing with small‐magnitude seismic events recorded at high sampling rates as well as larger magnitude events having long coda and remote events with long wave trains. Our results demonstrate that the proposed methods are effective in augmenting pre‐trained models for out‐of‐distribution data sets, especially in scenarios with limited labeled data for transfer learning.

Establishing the Foundations of Emotional Intelligence in Care Companion Robots to Mitigate Agitation Among High-Risk Patients With Dementia: Protocol for an Empathetic Patient-Robot Interaction Study.

An estimated 6.7 million persons are living with dementia in the United States, a number expected to double by 2060. Persons experiencing moderate to severe dementia are 4 to 5 times more likely to fall than those without dementia, due to agitation and unsteady gait. Socially assistive robots fail to address the changing emotional states associated with agitation, and it is unclear how emotional states change, how they impact agitation and gait over time, and how social robots can best respond by showing empathy. This study aims to design and validate a foundational model of emotional intelligence for empathetic patient-robot interaction that mitigates agitation among those at the highest risk: persons experiencing moderate to severe dementia. A design science approach will be adopted to (1) collect and store granular, personal, and chronological data using Personicle (an open-source software platform developed to automatically collect data from phones and other devices), incorporating real-time visual, audio, and physiological sensing technologies in a simulation laboratory and at board and care facilities; (2) develop statistical models to understand and forecast the emotional state, agitation level, and gait pattern of persons experiencing moderate to severe dementia in real time using machine learning and artificial intelligence and Personicle; (3) design and test an empathy-focused conversation model, focused on storytelling; and (4) test and evaluate this model for a care companion robot (CCR) in the community. The study was funded in October 2023. For aim 1, architecture development for Personicle data collection began with a search for existing open-source data in January 2024. A community advisory board was formed and met in December 2023 to provide feedback on the use of CCRs and provide personal stories. Full institutional review board approval was received in March 2024 to place cameras and CCRs at the sites. In March 2024, atomic marker development was begun. For aim 2, after a review of open-source data on patients with dementia, the development of an emotional classifier was begun. Data labeling was started in April 2024 and completed in June 2024 with ongoing validation. Moreover, the team established a baseline multimodal model trained and validated on healthy-person data sets, using transformer architecture in a semisupervised manner, and later retrained on the labeled data set of patients experiencing moderate to severe dementia. In April 2024, empathy alignment of large language models was initiated using prompt engineering and reinforcement learning. This innovative caregiving approach is designed to recognize the signs of agitation and, upon recognition, intervene with empathetic verbal communication. This proposal has the potential to have a significant impact on an emerging field of computational dementia science by reducing unnecessary agitation and falls of persons experiencing moderate to severe dementia, while reducing caregiver burden. PRR1-10.2196/55761.

Clustering of Global Magnetospheric Observations

AbstractThe use of supervised methods in space science have demonstrated powerful capability in classification tasks, but purely unsupervised methods have been less utilized for the classification of spacecraft observations. We use a combination of unsupervised methods, being principal component analysis, Self‐Organizing Maps, and hierarchical agglomerative clustering, to classify THEMIS and MMS observations as having occurred in the magnetosphere, magnetosheath, or the solar wind. The resulting classification are validated visually by analyzing the distribution of classifications and studying individual time series as well as by comparison to the labeled data set of a previous model, against which ours has an accuracy of 99.4. The model has a variety of applications beyond region classification such as deeper hierarchical analysis, magnetopause and bow shock crossing identification, and identification of bursty bulk flows, hot flow anomalies, and foreshock bubbles.

Asymmetric Network Combining CNN and Transformer for Building Extraction from Remote Sensing Images.

The accurate extraction of buildings from remote sensing images is crucial in fields such as 3D urban planning, disaster detection, and military reconnaissance. In recent years, models based on Transformer have performed well in global information processing and contextual relationship modeling, but suffer from high computational costs and insufficient ability to capture local information. In contrast, convolutional neural networks (CNNs) are very effective in extracting local features, but have a limited ability to process global information. In this paper, an asymmetric network (CTANet), which combines the advantages of CNN and Transformer, is proposed to achieve efficient extraction of buildings. Specifically, CTANet employs ConvNeXt as an encoder to extract features and combines it with an efficient bilateral hybrid attention transformer (BHAFormer) which is designed as a decoder. The BHAFormer establishes global dependencies from both texture edge features and background information perspectives to extract buildings more accurately while maintaining a low computational cost. Additionally, the multiscale mixed attention mechanism module (MSM-AMM) is introduced to learn the multiscale semantic information and channel representations of the encoder features to reduce noise interference and compensate for the loss of information in the downsampling process. Experimental results show that the proposed model achieves the best F1-score (86.7%, 95.74%, and 90.52%) and IoU (76.52%, 91.84%, and 82.68%) compared to other state-of-the-art methods on the Massachusetts building dataset, the WHU building dataset, and the Inria aerial image labeling dataset.

Multi-label remote sensing classification with self-supervised gated multi-modal transformers.

With the great success of Transformers in the field of machine learning, it is also gradually attracting widespread interest in the field of remote sensing (RS). However, the research in the field of remote sensing has been hampered by the lack of large labeled data sets and the inconsistency of data modes caused by the diversity of RS platforms. With the rise of self-supervised learning (SSL) algorithms in recent years, RS researchers began to pay attention to the application of "pre-training and fine-tuning" paradigm in RS. However, there are few researches on multi-modal data fusion in remote sensing field. Most of them choose to use only one of the modal data or simply splice multiple modal data roughly. In order to study a more efficient multi-modal data fusion scheme, we propose a multi-modal fusion mechanism based on gated unit control (MGSViT). In this paper, we pretrain the ViT model based on BigEarthNet dataset by combining two commonly used SSL algorithms, and propose an intra-modal and inter-modal gated fusion unit for feature learning by combining multispectral (MS) and synthetic aperture radar (SAR). Our method can effectively combine different modal data to extract key feature information. After fine-tuning and comparison experiments, we outperform the most advanced algorithms in all downstream classification tasks. The validity of our proposed method is verified.

Method for Augmenting Side-Scan Sonar Seafloor Sediment Image Dataset Based on BCEL1-CBAM-INGAN.

In this paper, a method for augmenting samples of side-scan sonar seafloor sediment images based on CBAM-BCEL1-INGAN is proposed, aiming to address the difficulties in acquiring and labeling datasets, as well as the insufficient diversity and quantity of data samples. Firstly, a Convolutional Block Attention Module (CBAM) is integrated into the residual blocks of the INGAN generator to enhance the learning of specific attributes and improve the quality of the generated images. Secondly, a BCEL1 loss function (combining binary cross-entropy and L1 loss functions) is introduced into the discriminator, enabling it to focus on both global image consistency and finer distinctions for better generation results. Finally, augmented samples are input into an AlexNet classifier to verify their authenticity. Experimental results demonstrate the excellent performance of the method in generating images of coarse sand, gravel, and bedrock, as evidenced by significant improvements in the Frechet Inception Distance (FID) and Inception Score (IS). The introduction of the CBAM and BCEL1 loss function notably enhances the quality and details of the generated images. Moreover, classification experiments using the AlexNet classifier show an increase in the recognition rate from 90.5% using only INGAN-generated images of bedrock to 97.3% using images augmented using our method, marking a 6.8% improvement. Additionally, the classification accuracy of bedrock-type matrices is improved by 5.2% when images enhanced using the method presented in this paper are added to the training set, which is 2.7% higher than that of the simple method amplification. This validates the effectiveness of our method in the task of generating seafloor sediment images, partially alleviating the scarcity of side-scan sonar seafloor sediment image data.

Identifying Mergers in the Legacy Surveys with Few-shot Learning

Galaxy mergers exert a pivotal influence on the evolutionary trajectory of galaxies and the expansive development of cosmic structures. The primary challenge encountered in machine learning–based identification of merging galaxies arises from the scarcity of meticulously labeled data sets specifically dedicated to merging galaxies. In this paper, we propose a novel framework utilizing few-shot learning techniques to identify galaxy mergers in the Legacy Surveys. Few-shot learning enables effective classification of merging galaxies even when confronted with limited labeled training samples. We employ a deep convolutional neural network architecture trained on data sets sampled from Galaxy Zoo Decals to learn essential features and generalize to new instances. Our experimental results demonstrate the efficacy of our approach, achieving high accuracy and precision in identifying galaxy mergers with few labeled training samples. Furthermore, we investigate the impact of various factors, such as the number of training samples and network architectures, on the performance of the few-shot learning model. The proposed methodology offers a promising avenue for automating the identification of galaxy mergers in large-scale surveys, facilitating the comprehensive study of galaxy evolution and structure formation. In pursuit of identifying galaxy mergers, our methodology is applied to analyze the Data Release 9 of the Dark Energy Spectroscopic Instrument Legacy Imaging Surveys. As a result, we have unveiled an extensive catalog encompassing 648,183 galaxy merger candidates. We publicly release the catalog alongside this paper.

Modified deep inductive transfer learning diagnostic systems for diabetic retinopathy severity levels classification

Diabetic Retinopathy (DR), a retinal illness that degenerates the retina and causes blindness, can be effectively treated with early detection and examination. Although expensive and unpleasant, manual retinography is the gold standard for DR diagnosis. Many Deep Learning (DL)-based algorithms have shown promise as deep learning (DR) diagnostic tools, performing similarly to human picture evaluation. These strategies work best with fine-tuned hyperparameters and huge databases. To develop Deep Inductive Transfer Learning-based Diagnostic Systems (DITL-DS) for Multiclass DR Severity Classification. The benchmark Indian Diabetic Retinopathy Image Dataset (IDRiD) is chosen first. Next, the dataset is pre-processed by proportionally scaling the image to 128*128 and using Convolutional Local Area Histogram Equalization (CLAHE). Third, the uneven class label dataset is balanced for efficiency. Fourth, a global average pooling layer and bottlenecking are added to five DITL models—Inception V3, ResNet34, EfficientNet B0, VGG16, and Xception—to compensate for information loss. Finally, we compare our improved approaches to five basic DITL classifiers. Our modified Xception model separates severity stages best, with an accuracy of 0.9840, precision of 0.9845, recall of 0.9650, AUC of 0.9979, and AUC-ROC of 0.9902. Precision, recall, and F1 scores are 0.97, 0.99, and 0.98 for stage0, 0.99, 1, and 1 for stage1, 0.98, 0.94, and 0.96 for stage2, 0.99, 0.99, and 0.99 for stage3, and 1, 1, and 1 for stage4. This may help clinicians objectively diagnose DR early. To ensure robustness, the model’s generalizability was validated using APTOS (Asia Pacific Tele-Ophthalmology Society).

Matching Occupational Injury and Illness Data Using Augmented Twin Neural Networks

When record linkage efforts involve complex characteristics there is ample potential for general purpose machine learning (ML) techniques to succeed where traditional probabilistic approaches might fall short. However, there can still be pre-processing (e.g. geocoding) and hand-picked comparators that can further improve linkage outcomes using standard ML models. In this project we present a fusion of these sides we are calling an Augmented Twin Neural Network. This approach leverages the inherent flexibility of Twin Neural Networks in a record linkage context while adding additional layers to allow for hand curated comparators that may be difficult for ML optimizers to implicitly identify without sufficiently large, labeled data sets. The framework is used to match establishments from the BLS Survey of Occupational Injuries and Illnesses to establishments in the OSHA Injury Tracking Application data. The difficulties inherent in matching company names and addresses and the existence of multi-establishment firms make this a prime application for testing. Linkage outcome metrics of this augmented algorithm are compared both with results from probabilistic methods (e.g. Fellegi-Sunter) and standard machine learning methods to illustrate the added benefits.

DeepKINET: a deep generative model for estimating single-cell RNA splicing and degradation rates

Messenger RNA splicing and degradation are critical for gene expression regulation, the abnormality of which leads to diseases. Previous methods for estimating kinetic rates have limitations, assuming uniform rates across cells. DeepKINET is a deep generative model that estimates splicing and degradation rates at single-cell resolution from scRNA-seq data. DeepKINET outperforms existing methods on simulated and metabolic labeling datasets. Applied to forebrain and breast cancer data, it identifies RNA-binding proteins responsible for kinetic rate diversity. DeepKINET also analyzes the effects of splicing factor mutations on target genes in erythroid lineage cells. DeepKINET effectively reveals cellular heterogeneity in post-transcriptional regulation.

Machine Learning-Based Identification of Risk Factors of Keratoconus Progression Using Raw Corneal Tomography Data.

The purpose of this study was to identify early indicators of keratoconus progression in Pentacam data using machine learning (ML) techniques. A retrospective Pentacam tabular data set was created by retrieving 11,760 tomography tests performed in patients with keratoconus. Data for eyes labeled unstable based on their referral for cross-linking were differentiated from data for eyes labeled stable and not referred for follow-up procedures. A boosted decision tree was trained on the final data set using a cross-validation method. The final labeled data set included 1218 tomography tests. Training a ML model on a single test for each eye did not accurately predict disease progression, as indicated by the mean receiver-operating characteristic area under the curve of 0.59 ± 0.1, with precision of 0.27, recall of 0.3, and F1 score of 0.28. Training on serial tests for each eye included 819 tomography scans and yielded good prognostic abilities: a receiver-operating characteristic area under the curve of 0.75 ± 0.07, precision of 0.32, recall of 0.67, and F1 score of 0.43. In addition, 4 of the 55 Pentacam raw data parameters predominantly used the algorithm decision: age, central keratoconus index, Rs B, and D10 mm pachy. This study revealed specific dominant parameters attributing to the classification of stability, which are not routinely assessed in determining progression in common practice. Using ML techniques, keratoconus deterioration was evaluated algorithmically with training on multiple tests, yet was not predicted by a single tomography test. Hence, our study highlights novel factors to the current consideration of cross-linking referral and may serve as a supportive tool for clinicians.

SSP: self-supervised pertaining technique for classification of shoulder implants in x-ray medical images: a broad experimental study

Multiple pathologic conditions can lead to a diseased and symptomatic glenohumeral joint for which total shoulder arthroplasty (TSA) replacement may be indicated. The long-term survival of implants is limited. With the increasing incidence of joint replacement surgery, it can be anticipated that joint replacement revision surgery will become more common. It can be challenging at times to retrieve the manufacturer of the in situ implant. Therefore, certain systems facilitated by AI techniques such as deep learning (DL) can help correctly identify the implanted prosthesis. Correct identification of implants in revision surgery can help reduce perioperative complications and complications. DL was used in this study to categorise different implants based on X-ray images into four classes (as a first case study of the small dataset): Cofield, Depuy, Tornier, and Zimmer. Imbalanced and small public datasets for shoulder implants can lead to poor performance of DL model training. Most of the methods in the literature have adopted the idea of transfer learning (TL) from ImageNet models. This type of TL has been proven ineffective due to some concerns regarding the contrast between features learnt from natural images (ImageNet: colour images) and shoulder implants in X-ray images (greyscale images). To address that, a new TL approach (self-supervised pertaining (SSP)) is proposed to resolve the issue of small datasets. The SSP approach is based on training the DL models (ImageNet models) on a large number of unlabelled greyscale medical images in the domain to update the features. The models are then trained on a small labelled data set of X-ray images of shoulder implants. The SSP shows excellent results in five ImageNet models, including MobilNetV2, DarkNet19, Xception, InceptionResNetV2, and EfficientNet with precision of 96.69%, 95.45%, 98.76%, 98.35%, and 96.6%, respectively. Furthermore, it has been shown that different domains of TL (such as ImageNet) do not significantly affect the performance of shoulder implants in X-ray images. A lightweight model trained from scratch achieves 96.6% accuracy, which is similar to using standard ImageNet models. The features extracted by the DL models are used to train several ML classifiers that show outstanding performance by obtaining an accuracy of 99.20% with Xception+SVM. Finally, extended experimentation has been carried out to elucidate our approach’s real effectiveness in dealing with different medical imaging scenarios. Specifically, five different datasets are trained and tested with and without the proposed SSP, including the shoulder X-ray with an accuracy of 99.47% and CT brain stroke with an accuracy of 98.60%.

Downstream lingering attention transformer network (DsLATNet) for land use land cover classification: A bicolor deep learning framework

Land Use and Land Cover (LULC) classification is the process of locating and classifying regions of the Earth's surface (land cover) based on their physical attributes and human utilization (land use purpose). It is essential for mapping and monitoring changes in ecosystems, facilitating effective resource management and informed decision-making in land-use planning. Convolutional Neural Network (CNN’s) is incapable of capturing long-range dependencies. The effectiveness of transformers relies on extensive training datasets, yet many satellite datasets have comparatively limited sample sizes. Moreover, due to the numerous factors influencing visual prominence, it is challenging to obtain abundant features from a single-color space. To overcome these limitations, we introduce an innovative bicolor architecture Downstream Lingering Attention Transformer Network (DsLATNet). DsLATNet characterizes remote sensing images on two color spaces viz RGB and HSV by processing the information using Residual Network −50(ResNet-50) backbone network.Extracted features of backbone network are further refined to acquire detailed aware features through Downstream Lingering Feature Pyramid Network (DsLFPN).This is further integrated with a DuoTransformer that utilizes attention mechanisms to obtain long-range dependencies. Eventually, the features are aggregated using residual connections that are dilated in-depth multiple times to produce class labels. The efficacy of the algorithm is evaluated through standard classification accuracy metrics. Experimental results on both the Wuhan Dense Labeling Dataset (WHDLD) and the Gaofen Image Dataset (GID) exhibit that the suggested technique exceeds cutting-edge classification methods. WHDLD achieved an Overall Accuracy (OA) of 86.34 %, Average Accuracy (AA) of 75.62 %, and Kappa coefficient (K) of 80.75, while the GID exhibited superior performance with an OA of 86.43 %, AA of 88.13 %, and K of 81.25.

Chaotic time series prediction based on physics-informed neural operator

This paper investigates the prediction of chaotic time series using physics-informed neural operator (PINO) with different driven methods, such as data-driven method, physics-driven method, and hybrid data-physics-driven method. Here, the chaotic time series are generated from two classical time delayed chaotic systems, including Mackey–Glass equation (MG) and Optoelectronic Oscillator (OEO). The simulation results from these two chaotic systems or experimental results from Optoelectronic Oscillator demonstrate that when there is enough data with good quality, it is possible to train the model solely using data-driven method. Conversely, when possessing complete mastery of the physical prior knowledge, it is also available to train the model solely using physics-driven method, indicating there is no need to prepare label datasets as in data-driven method. However, when dataset quality is poor, for example, contaminated by noise, and precise physical prior knowledge is not fully acquired, the hybrid method will have a better performance.

Implication of Different Data Split Ratios on the Performance of Models in Price Prediction of Used Vehicles Using Regression Analysis

Introduction; Artificial intelligence (AI) and Machine Learning have become buzzwords lately due to technological changes and data quality testing, especially in shape and finish analysis. Lots of research has been conducted for linear regression algorithms to predict the price in different sectors for share stock, rental properties, prices of used cars etc. This study provides suitable data split ratio for optimum cost estimation based on linear regression model. In present days there is an increasing demand for having own car for every middle class family therefore this have given opportunity to motor vehicle business to offer wide range of used vehicle for re-sale especially companies like Maruti Suzuki, Tata motors & Mahendra motors in Indian motor vehicle industries. Therefore, it is important to know the current value of your car before spending your hard-earned money on any item. Objective; The objective of this paper is finding appropriate value of cars in Metropolitans or even in state capitals. Features like model, mileage, AC, seating capacities, fuel type automatic will be taken into account when doing this. This estimate is designed to help customers find the right options to suit their needs. Methods; We have used a linear regression model to estimate the value of the respective car. Results; For doing this price prediction in this paper using liner regression we have tried to find the optimum accuracy of model by varying data split ratio for training and test data set and concluded with the result that 80/20 ratio is the best ratio with optimum model accuracy for business domain analysis with labelled data set. Conclusion; The findings underscore the importance of careful consideration when selecting a data split ratio for price prediction models in the used vehicle market. The insights gleaned from this study can inform future research and contribute to the development of more accurate and reliable regression models in similar domains.

Better performance of deep learning pulmonary nodule detection using chest radiography with pixel level labels in reference to computed tomography: data quality matters

Labeling errors can significantly impact the performance of deep learning models used for screening chest radiographs. The deep learning model for detecting pulmonary nodules is particularly vulnerable to such errors, mainly because normal chest radiographs and those with nodules obscured by ribs appear similar. Thus, high-quality datasets referred to chest computed tomography (CT) are required to prevent the misclassification of nodular chest radiographs as normal. From this perspective, a deep learning strategy employing chest radiography data with pixel-level annotations referencing chest CT scans may improve nodule detection and localization compared to image-level labels. We trained models using a National Institute of Health chest radiograph-based labeling dataset and an AI-HUB CT-based labeling dataset, employing DenseNet architecture with squeeze-and-excitation blocks. We developed four models to assess whether CT versus chest radiography and pixel-level versus image-level labeling would improve the deep learning model’s performance to detect nodules. The models' performance was evaluated using two external validation datasets. The AI-HUB dataset with image-level labeling outperformed the NIH dataset (AUC 0.88 vs 0.71 and 0.78 vs.0.73 in two external datasets, respectively; both p < 0.001). However, the AI-HUB data annotated at the pixel level produced the best model (AUC 0.91 and 0.86 in external datasets), and in terms of nodule localization, it significantly outperformed models trained with image-level annotation data, with a Dice coefficient ranging from 0.36 to 0.58. Our findings underscore the importance of accurately labeled data in developing reliable deep learning algorithms for nodule detection in chest radiography.

Automatic text classification of drug-induced liver injury using document-term matrix and XGBoost.

Regulatory agencies generate a vast amount of textual data in the review process. For example, drug labeling serves as a valuable resource for regulatory agencies, such as U.S. Food and Drug Administration (FDA) and Europe Medical Agency (EMA), to communicate drug safety and effectiveness information to healthcare professionals and patients. Drug labeling also serves as a resource for pharmacovigilance and drug safety research. Automated text classification would significantly improve the analysis of drug labeling documents and conserve reviewer resources. We utilized artificial intelligence in this study to classify drug-induced liver injury (DILI)-related content from drug labeling documents based on FDA's DILIrank dataset. We employed text mining and XGBoost models and utilized the Preferred Terms of Medical queries for adverse event standards to simplify the elimination of common words and phrases while retaining medical standard terms for FDA and EMA drug label datasets. Then, we constructed a document term matrix using weights computed by Term Frequency-Inverse Document Frequency (TF-IDF) for each included word/term/token. The automatic text classification model exhibited robust performance in predicting DILI, achieving cross-validation AUC scores exceeding 0.90 for both drug labels from FDA and EMA and literature abstracts from the Critical Assessment of Massive Data Analysis (CAMDA). Moreover, the text mining and XGBoost functions demonstrated in this study can be applied to other text processing and classification tasks.