Annotated Dataset Research Articles

Do we need to label large datasets for river water segmentation? Benchmark and stage estimation with minimum to non-labeled image time series

ABSTRACT Automatic water segmentation often relies on supervised deep learning, which requires large, annotated datasets. We investigate deep learning approaches for water segmentation and, consequently, water stage estimation that need minimal or non-annotated datasets. We evaluated the Space-Time Correspondence Network or STCN, Segment Anything (SAM), and a combination of SAM with Grounding DINO. We used images from three stationary camera gauges and one dynamic river image dataset captured with an unmanned aerial vehicle (UAV). Further, we retrieved the water level using the produced water marks for the static images. For the UAV dataset, our results show that STCN and SAM achieved similar performances for image segmentation. Our findings reveal that the models can produce realistic water masks and water stage measurements. STCN achieved the best results for the camera gauges, making it a suitable option for sub-hour monitoring. Our approaches can be viable tools for ad hoc water stage measurements.

Diagnosis of Sacroiliitis Through Semi-Supervised Segmentation and Radiomics Feature Analysis of MRI Images.

Sacroiliitis is a hallmark of ankylosing spondylitis (AS), and early detection plays an important role in managing the condition effectively. MRI is commonly used for diagnosing sacroiliitis, traditional methods often depend on subjective interpretation or limited automation which can introduce variability in diagnoses. The integration of semi-supervised segmentation and radiomics features may reduce reliance on expert interpretation and the need for large annotated datasets, potentially enhancing diagnostic workflows. To develop a diagnostic model for sacroiliitis and bone marrow edema (BME) using semi-supervised segmentation and radiomics analysis of MRI images. Retrospective cohort study. A total of 257 patients (161 males, 93 females; age 11-74 years), including 155 sacroiliitis and 175 BME patients. A total of 514 sacroiliac joint (SIJ) MRI images are analyzed, with 359 used for training and 155 for testing. 3.0 T, spin echo T1-weighted imaging (T1WI) and short-tau inversion recovery (STIR). SIJ segmentation is automated using the semi-supervised segmentation-based Unimatch framework. Manual delineation of SIJ regions of interest (ROIs) on T1WI images by an experienced radiologist (W.M., 10-year experience) served as the reference standard for segmentation performance evaluation. Radiomics features from T1WI and STIR are used to train machine learning models, including support vector machine (SVM), logistic regression (LR), and light gradient boosting machine (LightGBM), for sacroiliitis and BME detection. Performance is assessed using area under the curve (AUC), sensitivity, specificity, and accuracy. The Dice coefficient is used to assess the performance of the semi-supervised segmentation model on SIJ segmentation. Performance is evaluated using receiver operating characteristic (ROC) curves and decision curve analysis (DCA). The Unimatch model achieves an average Dice coefficient of 0.859 for SIJ segmentation. AUCs for sacroiliitis detection are 0.84 (LR), 0.86 (SVM), and 0.78 (LightGBM), while for BME detection, AUCs are 0.73 (LR), 0.76 (SVM), and 0.70 (LightGBM). This study demonstrates that semi-supervised segmentation combined with radiomics features and machine learning models provides a promising approach for diagnosis of sacroiliitis and BME. This study aimed to improve the diagnosis of sacroiliitis and bone marrow edema in patients with ankylosing spondylitis. The researchers used a method that automatically segments MRI images and analyzes features from those images. By applying machine learning, they created models to help detect sacroiliitis and bone marrow edema more accurately. The results show that this approach can effectively assist in identifying these conditions, with the best accuracy for sacroiliitis and bone marrow edema reaching 81.2% and 74.2%, respectively. This method could help doctors make better decisions, offering a promising tool for improving diagnosis in clinical settings. 3 TECHNICAL EFFICACY: Stage 2.

Boosting GPT Models for Genomics Analysis: Generating Trusted Genetic Variant Annotations and Interpretations through RAG and fine-tuning

Abstract Motivation Large language models (LLMs) have acquired a remarkable level of knowledge through their initial training. However, they lack expertise in particular domains such as genomics. Variant annotation data, an important component of genomics, is crucial for interpreting and prioritizing disease-related variants among millions of variants identified by genetic sequencing. In our project, we aimed to improve LLM performance in genomics by adding variant annotation data to LLMs by retrieval-augmented generation (RAG) and fine-tuning techniques. Results Using RAG, we successfully integrated 190 million highly accurate variant annotations, curated from 5 major annotation datasets and tools, into GPT-4o. This integration empowers users to query specific variants and receive accurate variant annotations and interpretations supported by advanced reasoning and language understanding capabilities of LLMs. Additionally, fine-tuning GPT-4 on variant annotation data also improved model performance in some annotation fields, although the accuracy across more fields remains suboptimal. Our model significantly improved the accessibility and efficiency of the variant interpretation process by leveraging LLM capabilities. Our project also revealed that RAG outperforms fine-tuning in factual knowledge injection in terms of data volume, accuracy, and cost-effectiveness. As a pioneering study for adding genomics knowledge to LLMs, our work paves the way for developing more comprehensive and informative genomics AI systems to support clinical diagnosis and research projects, and it demonstrates the potential of LLMs in specialized domains.

Cooperative learning of Pl@ntNet's Artificial Intelligence algorithm: How does it work and how can we improve it?

Abstract Deep learning models for plant species identification rely on large annotated datasets. The Pl@ntNet system enables global data collection by allowing users to upload and annotate plant observations, leading to noisy labels due to diverse user skills. Achieving consensus is crucial for training, but the vast scale of collected data (number of observations, users and species) makes traditional label aggregation strategies challenging. Existing methods either retain all observations, resulting in noisy training data or selectively keep those with sufficient votes, discarding valuable information. Additionally, as many species are rarely observed, user expertise cannot be evaluated as an inter‐user agreement: otherwise, botanical experts would have a lower weight in the AI training step than the average user. Our proposed label aggregation strategy aims to cooperatively train plant identification AI models. This strategy estimates user expertise as a trust score per user based on their ability to identify plant species from crowdsourced data. The trust score is recursively estimated from correctly identified species given the current estimated labels. This interpretable score exploits botanical experts' knowledge and the heterogeneity of users. Subsequently, our strategy removes unreliable observations but retains those with limited trusted annotations, unlike other approaches. We evaluate Pl@ntNet's strategy on a newly released large subset of the Pl@ntNet database focused on European flora, comprising over 6 M observations and 800 K users. This anonymized dataset of votes and observations is released openly via Lefort, Affouard, et al. (2024). We demonstrate that estimating users' skills based on the diversity of their expertise enhances labelling performance. Our findings emphasize the synergy of human annotation and data filtering in improving AI performance for a refined training dataset. We explore incorporating AI‐based votes alongside human input in the label aggregation. This can further enhance human‐AI interactions to detect unreliable observations (even with few votes).

Open challenges for the automatic synthesis of clinical trials

ObjectiveAn important criterion for selecting clinical trials to be compared in systematic reviews and meta-analyses is that they measure the same outcomes. However, this represents a challenge as there is a wide variety of outcomes, and it is difficult to standardize them for comparing clinical trials containing them. To address this challenge, we utilized our annotated dataset, which includes 211 abstracts of clinical trials related to glaucoma and type 2 diabetes mellitus. We then developed a tool that provides an overview of the annotated clinical trial information and enables users to group them by outcomes.ResultsUsing our visualization tool, we formed groups of outcomes and their respective clinical trials. We were able to determine the most common outcomes in clinical trials for these diseases. As a case study on diabetes, we compared our outcomes with those consented by diabetes stakeholders and found that many of the grouped outcomes are aligned with the consented ones. This demonstrates that tools such as the one presented can help standardize clinical outcomes, which in turn help in the synthesis of clinical trials. Finally, we also offer some recommendations that could help in the automation of clinical trials based on outcome standardization.

Accuracy of a Cascade Network for Semi-Supervised Maxillary Sinus Detection and Sinus Cyst Classification.

Maxillary sinus mucosal cysts represent prevalent oral and maxillofacial diseases, and their precise diagnosis is essential for surgical planning in maxillary sinus floor elevation. This study aimed to develop a deep learning-based pipeline for the classification of maxillary sinus lesions in cone beam computed tomography (CBCT) images to provide auxiliary support for clinical diagnosis. This study utilized 45 136 maxillary sinus images from CBCT scans of 541 patients. A cascade network was designed, comprising a semi-supervised maxillary sinus area object detection module and a maxillary sinus lesions classification module. The object detection module employed a semi-supervised pseudo-labelling training strategy to expand the maxillary sinus annotation dataset. In the classification module, the performance of Convolutional Neural Network and Transformer architectures was compared for maxillary sinus mucosal lesion classification. The object detection and classification modules were evaluated using metrics including Accuracy, Precision, Recall, F1 score, and Average Precision, with the object detection module additionally assessed using Precision-Recall Curve. The fully supervised pseudo-label generation model achieved an average accuracy of 0.9433, while the semi-supervised maxillary sinus detection model attained 0.9403. ResNet-50 outperformed in classification, with accuracies of 0.9836 (sagittal) and 0.9797 (coronal). Grad-CAM visualization confirmed accurate focus on clinically relevant lesion features. The proposed pipeline achieves high-precision detection and classification of maxillary sinus mucosal lesions, reducing manual annotation while maintaining accuracy.

MEVDT: Multi-modal event-based vehicle detection and tracking dataset.

In this data article, we introduce the Multi-Modal Event-based Vehicle Detection and Tracking (MEVDT) dataset. This dataset provides a synchronized stream of event data and grayscale images of traffic scenes, captured using the Dynamic and Active-Pixel Vision Sensor (DAVIS) 240c hybrid event-based camera. MEVDT comprises 63 multi-modal sequences with approximately 13k images, 5M events, 10k object labels, and 85 unique object tracking trajectories. Additionally, MEVDT includes manually annotated ground truth labels - consisting of object classifications, pixel-precise bounding boxes, and unique object IDs - which are provided at a labeling frequency of 24 Hz. Designed to advance the research in the domain of event-based vision, MEVDT aims to address the critical need for high-quality, real-world annotated datasets that enable the development and evaluation of object detection and tracking algorithms in automotive environments.

AI And Deep Learning Models for Early Lung Cancer Detection Using Radiological Data

Lung cancer remains one of the deadliest cancers worldwide, with survival rates significantly dependent on early detection. Traditional diagnostic methods, while effective, often identify lung cancer in advanced stages, limiting treatment options. Recently, artificial intelligence (AI) and deep learning (DL) models have gained attention for their potential to assist in early lung cancer diagnosis through radiological data, particularly computed tomography (CT) and X-ray imaging. This review examines recent advancements in AI and DL models specifically applied to early lung cancer detection, offering a comprehensive look at model architecture, data preprocessing techniques, and performance metrics. AI models, particularly those using convolutional neural networks (CNNs) and recurrent neural networks (RNNs), have shown significant promise in improving diagnostic accuracy. In studies published over the past two years, CNN-based models have achieved sensitivity rates upwards of 90% when applied to large datasets, such as the LIDC-IDRI, a publicly available database containing thousands of annotated lung scans. Furthermore, hybrid approaches combining AI models with radiologist expertise have demonstrated reduced false positives, a frequent challenge in automated image analysis. A critical factor in these advancements has been data quality and volume. Large, annotated datasets with high-resolution images have enabled more robust training and validation, essential for refining these models. However, challenges remain, particularly regarding the standardization of data across institutions, variations in scanner quality, and ethical concerns surrounding patient privacy. The review also highlights studies exploring multi-modal approaches, integrating radiological data with clinical records and genetic markers to create more personalized diagnostic tools. These multi-modal methods have the potential to improve predictive accuracy further, though they currently require more extensive validation. Overall, while AI and DL technologies offer transformative potential for early lung cancer detection, widespread implementation depends on continued research in model accuracy, data standardization, and ethical safeguards. This paper concludes by emphasizing the need for collaboration across disciplines—AI researchers, radiologists, and oncologists—to refine these models into reliable tools that can be integrated seamlessly into clinical workflows for timely and accurate lung cancer diagnosis.

Application of a methodological framework for the development and multicenter validation of reliable artificial intelligence in embryo evaluation

BackgroundArtificial intelligence (AI) models analyzing embryo time-lapse images have been developed to predict the likelihood of pregnancy following in vitro fertilization (IVF). However, limited research exists on methods ensuring AI consistency and reliability in clinical settings during its development and validation process. We present a methodology for developing and validating an AI model across multiple datasets to demonstrate reliable performance in evaluating blastocyst-stage embryos.MethodsThis multicenter analysis utilizes time-lapse images, pregnancy outcomes, and morphologic annotations from embryos collected at 10 IVF clinics across 9 countries between 2018 and 2022. The four-step methodology for developing and evaluating the AI model include: (I) curating annotated datasets that represent the intended clinical use case; (II) developing and optimizing the AI model; (III) evaluating the AI’s performance by assessing its discriminative power and associations with pregnancy probability across variable data; and (IV) ensuring interpretability and explainability by correlating AI scores with relevant morphologic features of embryo quality. Three datasets were used: the training and validation dataset (n = 16,935 embryos), the blind test dataset (n = 1,708 embryos; 3 clinics), and the independent dataset (n = 7,445 embryos; 7 clinics) derived from previously unseen clinic cohorts.ResultsThe AI was designed as a deep learning classifier ranking embryos by score according to their likelihood of clinical pregnancy. Higher AI score brackets were associated with increased fetal heartbeat (FH) likelihood across all evaluated datasets, showing a trend of increasing odds ratios (OR). The highest OR was observed in the top G4 bracket (test dataset G4 score ≥ 7.5: OR 3.84; independent dataset G4 score ≥ 7.5: OR 4.01), while the lowest was in the G1 bracket (test dataset G1 score < 4.0: OR 0.40; independent dataset G1 score < 4.0: OR 0.45). AI score brackets G2, G3, and G4 displayed OR values above 1.0 (P < 0.05), indicating linear associations with FH likelihood. Average AI scores were consistently higher for FH-positive than for FH-negative embryos within each age subgroup. Positive correlations were also observed between AI scores and key morphologic parameters used to predict embryo quality.ConclusionsStrong AI performance across multiple datasets demonstrates the value of our four-step methodology in developing and validating the AI as a reliable adjunct to embryo evaluation.

Automated Icon Extraction from Tourism Maps: A Synergistic Approach Integrating YOLOv8x and SAM

Map symbols play a crucial role in cartographic representation. Among these symbols, icons are particularly valued for their vivid and intuitive designs, making them widely utilized in tourist maps. However, the diversity and complexity of these symbols present significant challenges to cartographic workflows. Icon design often relies on manual drawing, which is not only time-consuming but also heavily dependent on specialized skills. Automating the extraction of symbols from existing maps could greatly enhance the map symbol database, offering a valuable resource to support both symbol design and map production. Nevertheless, the intricate shapes and dense distribution of symbols in tourist maps complicate the accurate and efficient detection and extraction using existing methods. Previous studies have shown that You Only Look Once (YOLO) series models demonstrate strong performance in object detection, offering high accuracy and speed. However, these models are less effective in fine-grained boundary segmentation. To address this limitation, this article proposes integrating YOLO models with the Segment Anything Model (SAM) to tackle the challenges of combining efficient detection with precise segmentation. This article developed a dataset consisting of both paper-based and digital tourist maps, with annotations for five main categories of symbols: human landscapes, natural sceneries, humans, animals, and cultural elements. The performance of various YOLO model variants was systematically evaluated using this dataset. Additionally, a user interaction mechanism was incorporated to review and refine detection results, which were subsequently used as prompts for the SAM to perform precise symbol segmentation. The results indicate that the YOLOv8x model achieved excellent performance on the tourist map dataset, with an average detection accuracy of 94.4% across the five symbol categories, fully meeting the requirements for symbol detection tasks. The inclusion of a user interaction mechanism enhanced the reliability and flexibility of detection outcomes, while the integration of the SAM significantly improved the precision of symbol boundary extraction. In conclusion, the integration of YOLOv8x and SAM provides a robust and effective solution for automating the extraction of map symbols. This approach not only reduces the manual workload involved in dataset annotation, but also offers valuable theoretical and practical insights for enhancing cartographic efficiency.

Generative artificial intelligence enables the generation of bone scintigraphy images and improves generalization of deep learning models in data-constrained environments.

Advancements of deep learning in medical imaging are often constrained by the limited availability of large, annotated datasets, resulting in underperforming models when deployed under real-world conditions. This study investigated a generative artificial intelligence (AI) approach to create synthetic medical images taking the example of bone scintigraphy scans, to increase the data diversity of small-scale datasets for more effective model training and improved generalization. We trained a generative model on 99mTc-bone scintigraphy scans from 9,170 patients in one center to generate high-quality and fully anonymized annotated scans of patients representing two distinct disease patterns: abnormal uptake indicative of (i) bone metastases and (ii) cardiac uptake indicative of cardiac amyloidosis. A blinded reader study was performed to assess the clinical validity and quality of the generated data. We investigated the added value of the generated data by augmenting an independent small single-center dataset with synthetic data and by training a deep learning model to detect abnormal uptake in a downstream classification task. We tested this model on 7,472 scans from 6,448 patients across four external sites in a cross-tracer and cross-scanner setting and associated the resulting model predictions with clinical outcomes. The clinical value and high quality of the synthetic imaging data were confirmed by four readers, who were unable to distinguish synthetic scans from real scans (average accuracy: 0.48% [95% CI 0.46-0.51]), disagreeing in 239 (60%) of 400 cases (Fleiss' kappa: 0.18). Adding synthetic data to the training set improved model performance by a mean (± SD) of 33(± 10)% AUC (p < 0.0001) for detecting abnormal uptake indicative of bone metastases and by 5(± 4)% AUC (p < 0.0001) for detecting uptake indicative of cardiac amyloidosis across both internal and external testing cohorts, compared to models without synthetic training data. Patients with predicted abnormal uptake had adverse clinical outcomes (log-rank: p < 0.0001). Generative AI enables the targeted generation of bone scintigraphy images representing different clinical conditions. Our findings point to the potential of synthetic data to overcome challenges in data sharing and in developing reliable and prognostic deep learning models in data-limited environments.

Aspect category sentiment analysis based on pre-trained BiLSTM and syntax-aware graph attention network

Aspect Category Sentiment Analysis (ACSA) is a fine-grained sentiment analysis task aimed at predicting the sentiment polarity associated with aspect categories within a sentence.Most existing ACSA methods are based on a given aspect category to locate sentiment words related to it. When irrelevant sentiment words have semantic meaning for the given aspect category, it may cause the problem that sentiment words cannot be matched with aspect categories. To address the aforementioned issue, this paper proposes a novel approach for ACSA utilizing pre-trained Bidirectional Long Short-Term Memory (BiLSTM) and syntax-aware graph attention network. To address the issue of insufficient existing annotated datasets, a method of using transfer learning is proposed. Firstly, the BiLSTM model is used to pre-train on the document-level sentiment analysis dataset, and the obtained pre-training parameters are transferred to the aspect-level task model. Then, a syntax-aware graph attention network model is proposed to make full use of the syntactic structure and semantic information in the text, and combine the knowledge learned in pre-training to achieve the ACSA task. The performance evaluation of this method is carried out on five user comment text datasets, and the comprehensive ablation experiments prove that this method performs best compared with baseline models.

IForal: Automated Handwritten Text Transcription for Historical Medieval Manuscripts

The transcription of historical manuscripts aims at making our cultural heritage more accessible to experts and also to the larger public, but it is a challenging and time-intensive task. This paper contributes an automated solution for text layout recognition, segmentation, and recognition to speed up the transcription process of historical manuscripts. The focus is on transcribing Portuguese municipal documents from the Middle Ages in the context of the iForal project, including the contribution of an annotated dataset containing Portuguese medieval documents, notably a corpus of 67 Portuguese royal charter data. The proposed system can accurately identify document layouts, isolate the text, segment, and transcribe it. Results for the layout recognition model achieved 0.98 mAP@0.50 and 0.98 precision, while the text segmentation model achieved 0.91 mAP@0.50, detecting 95% of the lines. The text recognition model achieved 8.1% character error rate (CER) and 25.5% word error rate (WER) on the test set. These results can then be validated by palaeographers with less effort, contributing to achieving high-quality transcriptions faster. Moreover, the automatic models developed can be utilized as a basis for the creation of models that perform well for other historical handwriting styles, notably using transfer learning techniques. The contributed dataset has been made available on the HTR United catalogue, which includes training datasets to be used for automatic transcription or segmentation models. The models developed can be used, for instance, on the eSriptorium platform, which is used by a vast community of experts.

A quantum-optimized approach for breast cancer detection using SqueezeNet-SVM

Breast cancer is one of the most aggressive types of cancer, and its early diagnosis is crucial for reducing mortality rates and ensuring timely treatment. Computer-aided diagnosis systems provide automated mammography image processing, interpretation, and grading. However, since the currently existing methods suffer from such issues as overfitting, lack of adaptability, and dependence on massive annotated datasets, the present work introduces a hybrid approach to enhance breast cancer classification accuracy. The proposed Q-BGWO-SQSVM approach utilizes an improved quantum-inspired binary Grey Wolf Optimizer and combines it with SqueezeNet and Support Vector Machines to exhibit sophisticated performance. SqueezeNet’s fire modules and complex bypass mechanisms extract distinct features from mammography images. Then, these features are optimized by the Q-BGWO for determining the best SVM parameters. Since the current CAD system is more reliable, accurate, and sensitive, its application is advantageous for healthcare. The proposed Q-BGWO-SQSVM was evaluated using diverse databases: MIAS, INbreast, DDSM, and CBIS-DDSM, analyzing its performance regarding accuracy, sensitivity, specificity, precision, F1 score, and MCC. Notably, on the CBIS-DDSM dataset, the Q-BGWO-SQSVM achieved remarkable results at 99% accuracy, 98% sensitivity, and 100% specificity in 15-fold cross-validation. Finally, it can be observed that the performance of the designed Q-BGWO-SQSVM model is excellent, and its potential realization in other datasets and imaging conditions is promising. The novel Q-BGWO-SQSVM model outperforms the state-of-the-art classification methods and offers accurate and reliable early breast cancer detection, which is essential for further healthcare development.

Automated Machine Learning in Dentistry: A Narrative Review of Applications, Challenges, and Future Directions

The adoption of automated machine learning (AutoML) in dentistry is transforming clinical practices by enabling clinicians to harness machine learning (ML) models without requiring extensive technical expertise. This narrative review aims to explore the impact of autoML in dental applications. A comprehensive search of PubMed, Scopus, and Google Scholar was conducted without time and language restrictions. Inclusion criteria focused on studies evaluating autoML applications and performance for dental tasks. Exclusion criteria included non-dental studies, single-case reports, and conference abstracts. This review highlights multiple promising applications of autoML in dentistry. Diagnostic tasks showed high accuracy, such as 95.4% precision in dental implant classification and 92% accuracy in paranasal sinus disease detection. Predictive tasks also demonstrated promise, including 84% accuracy for ICU admissions due to dental infections and 93.9% accuracy in orthodontic extraction predictions. AutoML frameworks like Google Vertex AI and H2O AutoML emerged as key tools for these applications. AutoML shows great promise in transforming dentistry by facilitating data-driven decision-making and improving patient care quality through accessible, automated solutions. Future advancements should focus on enhancing model interpretability, developing large and annotated datasets, and creating pipelines tailored to dental tasks. Educating clinicians on autoML and integrating domain-specific knowledge into automated platforms could further bridge the gap between complex ML technology and practical dental applications.

Symmetry-induced ambiguity in orientation estimation from RGB images

The estimation of object orientation from RGB images is a core component in many modern computer vision pipelines. Traditional techniques mostly predict a single orientation per image, learning a one-to-one mapping between images and rotations. However, when objects exhibit rotational symmetries, they can appear identical from multiple viewpoints. This induces ambiguity in the estimation problem, making images map to rotations in a one-to-many fashion. In this paper, we explore several ways of addressing this problem. In doing so, we specifically consider algorithms that can map an image to a range of multiple rotation estimates, accounting for symmetry-induced ambiguity. Our contributions are threefold. Firstly, we create a data set with annotated symmetry information that covers symmetries induced through self-occlusion. Secondly, we compare and evaluate various learning strategies for multiple-hypothesis prediction models applied to orientation estimation. Finally, we propose to model orientation estimation as a binary classification problem. To this end, based on existing work from the field of shape reconstruction, we design a neural network that can be sampled to reconstruct the full range of ambiguous rotations for a given image. Quantitative evaluation on our annotated data set demonstrates its performance and motivates our design choices.

Mapping Pituitary Neuroendocrine Tumors: An Annotated MRI Dataset Profiling Tumor and Carotid Characteristics

Pituitary neuroendocrine tumors remain one of the most common intracranial tumors. While radiomic research related to pituitary tumors is progressing, public data sets for external validation remain scarce. We introduce an open dataset comprising high-resolution T1 contrast-enhanced MR scans of 136 patients with pituitary tumors, annotated for tumor segmentation and accompanied by clinical, radiological and pathological metadata. This diverse dataset captures variations in tumor size, location, and pathological activity, essential for understanding this complex condition. Expert annotations of both the tumor and adjacent carotid arteries ensure precise delineation, facilitating the development of automated segmentation algorithms. Our initiative addresses the need for standardized data in pituitary oncology, fosters collaboration and innovation, and enables the development and benchmarking of workflows that utilize pituitary radiomics for treatment planning and outcome prediction.

Pneumothorax Detection System in Thoracic Radiography Images Using CNN Method

Purpose: This research aims to develop an automatic pneumothorax detection system using Convolutional Neural Networks (CNN) to classify thoracic radiography images. By leveraging CNN's effectiveness in identifying medical abnormalities, the system seeks to enhance diagnostic accuracy, reduce evaluation time, and minimize subjective interpretation errors. The output will provide a predicted label of "pneumothorax" or "non-pneumothorax," facilitating faster clinical treatment and improving diagnostic services while supporting radiologists in making more accurate and efficient decisions for this critical condition. Methods: This research employs an experimental deep learning approach using Convolutional Neural Networks (CNN) to detect pneumothorax in thoracic radiography images. The CNN model is trained on an annotated dataset with preprocessing steps, including zooming, brightness adjustment, flipping and format adjustment, followed by performance evaluation using accuracy, precision, recall, and F1 score metrics. Result: The results showed that the CNN model detected pneumothorax with 79.59% accuracy, a loss of 1.3056, and 1,092 correct predictions out of 1,372 test data. Precision was 51.12%, recall 78.62%, and F1 score 61.96%, confirming the system's potential, though further optimization is needed. Novelty: The novelty of this research lies in developing an automated pneumothorax detection system using a CNN architecture, improving diagnostic accuracy and efficiency. Despite high accuracy, precision and recall can be improved. Future research can focus on optimizing the model and applying data augmentation techniques.

Facilitating automated fact-checking: a machine learning based weighted ensemble technique for claim detection

The rapid digitization of media, driven by technological advancements, has accelerated the spread of information through online platforms. This has heightened the need for robust fact-checking mechanisms to counter misinformation. The prevalence of misinformation necessitates the development of automated claim detection systems to support efficient automated or semi-automated fact-checking processes. Existing claim detection systems predominantly focus on the English language, with limited resources available for other regional languages like Bangla. This paper proposes a novel ensemble machine learning framework for the effective detection of claims in a low-resource language like Bangla, a critical initial step in the automated fact-checking process. The proposed weighted ensemble technique combines Support Vector Machines, Bernoulli Naive Bayes, and Decision Trees as base classifiers to effectively detect claims. An annotated text dataset comprising 5010 sentences sourced from various online platforms, including several online fact-checking sites, was developed. To determine the optimal model and feature representation for claim detection, various machine learning algorithms were evaluated using BoW, TF-IDF, Word2Vec, and FastText features. The efficacy of ensemble models was examined by investigating both averaging and weighting strategies. Evaluation metrics showcased that the proposed weighted ensemble approach outperformed all baseline models, achieving a maximum F1 score of 0.87. To the best of our knowledge, this study is the first and only approach to claim detection in the Bangla language, with the potential for extension to other resource-constrained languages. Our work aspires to serve as a crucial tool in the fight against misinformation by advancing the accuracy and transparency of information.

Development of avian influenza A(H5) virus datasets for Nextclade enables rapid and accurate clade assignment.

The ongoing panzootic of highly pathogenic avian influenza (HPAI) A(H5) viruses is the largest in history, with unprecedented transmission to multiple mammalian species. Avian influenza A viruses of the H5 subtype circulate globally among birds and are classified into distinct clades based on their hemagglutinin (HA) genetic sequences. Thus, the ability to accurately and rapidly assign clades to newly sequenced isolates is key to surveillance and outbreak response. Co-circulation of endemic, low pathogenic avian influenza (LPAI) A(H5) lineages in North American and European wild birds necessitates the ability to rapidly and accurately distinguish between infections arising from these lineages and epizootic HPAI A(H5) viruses. However, currently available clade assignment tools are limited and often require command line expertise, hindering their utility for public health surveillance labs. To address this gap, we have developed datasets to enable A(H5) clade assignments with Nextclade, a drag-and-drop tool originally developed for SARS-CoV-2 genetic clade classification. Using annotated reference datasets for all historical A(H5) clades, clade 2.3.2.1 descendants, and clade 2.3.4.4 descendants provided by the Food and Agriculture Organization/World Health Organization/World Organisation for Animal Health (FAO/WHO/WOAH) H5 Working Group, we identified clade-defining mutations for every established clade to enable tree-based clade assignment. We then created three Nextclade datasets which can be used to assign clades to A(H5) HA sequences and call mutations relative to reference strains through a drag-and-drop interface. Nextclade assignments were benchmarked with 19,834 unique sequences not in the reference set using a pre-released version of LABEL, a well-validated and widely used command line software. Prospective assignment of new sequences with Nextclade and LABEL produced very well-matched assignments (match rates of 97.8% and 99.1% for the 2.3.2.1 and 2.3.4.4 datasets, respectively). The all-clades dataset also performed well (94.8% match rate) and correctly distinguished between all HPAI and LPAI strains. This tool additionally allows for the identification of polybasic cleavage site sequences and potential N-linked glycosylation sites. These datasets therefore provide an alternative, rapid method to accurately assign clades to new A(H5) HA sequences, with the benefit of an easy-to-use browser interface.