Large Annotated Datasets Research Articles

Comparative analysis of vision transformers and fine-tuned transfer learning models for brain tumor classification

Brain tumors are a primary factor causing cancer-related deaths globally, and their classification remains a significant research challenge due to the variability in tumor intensity, size, and shape, as well as the similar appearances of different tumor types. Accurate differentiation is further complicated by these factors, making diagnosis difficult even with advanced imaging techniques such as magnetic resonance imaging (MRI). Recent techniques in artificial intelligence (AI), in particular deep learning (DL), have improved the speed and accuracy of medical image analysis, but they still face challenges like overfitting and the need for large annotated datasets. This study addresses these challenges by presenting two approaches for brain tumor classification using MRI images. The first approach involves fine-tuning transfer learning cutting-edge models, including SEResNet, ConvNeXtBase, and ResNet101V2, with global average pooling 2D and dropout layers to minimize overfitting and reduce the need for extensive preprocessing. The second approach leverages the Vision Transformer (ViT), optimized with the AdamW optimizer and extensive data augmentation. Experiments on the BT-Large-4C dataset demonstrate that SEResNet achieves the highest accuracy of 97.96%, surpassing ViT’s 95.4%. These results suggest that fine-tuning and transfer learning models are more effective at addressing the challenges of overfitting and dataset limitations, ultimately outperforming the Vision Transformer and existing state-of-the-art techniques in brain tumor classification.

NOVEL DEVICE FOR ENHANCING TUBERCULOSIS DIAGNOSIS FOR FASTER, MORE ACCURATE SCREENING RESULTS

Tuberculosis (TB) continues to be a leading global cause of illness and death, highlighting an urgent need for rapid, precise diagnostic solutions. Conventional methods, such as chest X-rays and sputum smear microscopy, often lack adequate sensitivity and specificity, resulting in treatment delays and increased transmission. This research introduces a novel device leveraging deep learning to enhance TB diagnosis, utilizing a convolutional neural network (CNN) model designed to automatically detect TB-related abnormalities in chest X-ray images. The device, powered by a large annotated dataset, incorporates transfer learning to fine-tune pre-trained models, maximizing diagnostic performance. Key evaluation metrics—accuracy, precision, recall, and F1-score—demonstrate the model’s efficacy compared to traditional diagnostics. Preliminary findings show that the AI-driven approach significantly enhances diagnostic accuracy, reducing false negatives and facilitating faster screenings. These results suggest that embedding AI technology in clinical workflows can enable earlier TB detection, optimize healthcare resources, and improve patient outcomes. Future directions include testing the device across diverse clinical environments and exploring integration with mobile health technologies to broaden access to rapid TB diagnostics.

Role of Radiology in the Diagnosis and Treatment of Breast Cancer in Women: A Comprehensive Review.

Breast cancer remains a leading cause of morbidity and mortality among women worldwide. Early detection and precise diagnosis are critical for effective treatment and improved patient outcomes. This review explores the evolving role of radiology in the diagnosis and treatment of breast cancer, highlighting advancements in imaging technologies and the integration of artificial intelligence (AI). Traditional imaging modalities such as mammography, ultrasound, and magnetic resonance imaging have been the cornerstone of breast cancer diagnostics, with each modality offering unique advantages. The advent of radiomics, which involves extracting quantitative data from medical images, has further augmented the diagnostic capabilities of these modalities. AI, particularly deep learning algorithms, has shown potential in improving diagnostic accuracy and reducing observer variability across imaging modalities. AI-driven tools are increasingly being integrated into clinical workflows to assist in image interpretation, lesion classification, and treatment planning. Additionally, radiology plays a crucial role in guiding treatment decisions, particularly in the context of image-guided radiotherapy and monitoring response to neoadjuvant chemotherapy. The review also discusses the emerging field of theranostics, where diagnostic imaging is combined with therapeutic interventions to provide personalized cancer care. Despite these advancements, challenges such as the need for large annotated datasets and the integration of AI into clinical practice remain. The review concludes that while the role of radiology in breast cancer management is rapidly evolving, further research is required to fully realize the potential of these technologies in improving patient outcomes.

Zero-shot prompt-based video encoder for surgical gesture recognition.

In order to produce a surgical gesture recognition system that can support a wide variety of procedures, either a very large annotated dataset must be acquired, or fitted models must generalize to new labels (so-called zero-shot capability). In this paper we investigate the feasibility of latter option. Leveraging the bridge-prompt framework, we prompt-tune a pre-trained vision-text model (CLIP) for gesture recognition in surgical videos. This can utilize extensive outside video data such as text, but also make use of label meta-data and weakly supervised contrastive losses. Our experiments show that prompt-based video encoder outperforms standard encoders in surgical gesture recognition tasks. Notably, it displays strong performance in zero-shot scenarios, where gestures/tasks that were not provided during the encoder training phase are included in the prediction phase. Additionally, we measure the benefit of inclusion text descriptions in the feature extractor training schema. Bridge-prompt and similar pre-trained+prompt-tuned video encoder models present significant visual representation for surgical robotics, especially in gesture recognition tasks. Given the diverse range of surgical tasks (gestures), the ability of these models to zero-shot transfer without the need for any task (gesture) specific retraining makes them invaluable.

HindiSumm: A Hindi Abstractive Summarization benchmark dataset

Abstractive Text Summarization (ATS) is a task to create a novel summary by generating fresh sentences incorporating new words or rephrasing the article. It is a complex task as the model needs to understand the semantic similarity between the sentences of the text. To fulfill this, there is a need for a large annotated benchmark dataset, which is available for resource-rich languages such as English and non-indic languages. In contrast, for the less-resourced languages, such as Indic languages, the available datasets are limited and involve very short summary sentences. Hence, a language-specific abstractive summarization dataset called HindiSumm was introduced for Hindi, consisting of 570,000 text-summary pairs from Navbharat Times across 21 domains. The HindiSumm dataset’s efficiency is evaluated extrinsically and intrinsically by using various metrics. Furthermore, two recent multilingual-cased pre-trained models are fine-tuned on the HindiSumm dataset individually. In addition, an ensembled approach using weighted averaging is also incorporated to check the efficacy of the proposed dataset. The model is tested with the in-house created dataset, and results are evaluated on ROUGE scores and show significant improvements of around 13.2% for the proposed HindiSumm compared to other benchmark datasets. In the future, the HindiSumm dataset will promote the progress of ATS for the Indian language.

A social context-aware graph-based multimodal attentive learning framework for disaster content classification during emergencies

In times of crisis, the prompt and precise classification of disaster-related information shared on social media platforms is of paramount importance for effective disaster response and public safety. During such critical events, people utilize social media as a medium for communication, sharing multimodal textual and visual content. However, due to the substantial influx of unfiltered and diverse data, humanitarian organizations face challenges in effectively leveraging this information. Numerous methods have been proposed for classifying disaster-related content, but these methods lack modeling users’ credibility, emotional context, and social interaction information, which is crucial for classification. In this context, we propose a method, CrisisSpot, that leverages a Graph-based Neural Network to comprehend intricate relationships between textual and visual modalities and Social Context Features to incorporate user-centric and content-centric information. We also propose Inverted Dual Embedded Attention, which captures both harmonious and contrary patterns present in the data to harness complex interactions and facilitate richer insights in multimodal data. We have developed a multimodal disaster dataset, TSEqD (Turkey-Syria Earthquake Dataset), which is a large annotated dataset for a single disaster event containing 10,352 data samples. Through extensive experimentation, CrisisSpot has demonstrated significant improvements, achieving an average gain of 9.45% and 5.01% in F1-score compared to the state-of-the-art methods on the publicly available CrisisMMD dataset and TSEqD dataset, respectively.

AnnotateXR: An Extended Reality Workflow for Automating Data Annotation to Support Computer Vision Applications

Abstract Computer vision (CV) algorithms require large annotated datasets that are often labor-intensive and expensive to create. We propose AnnotateXR, an extended reality (XR) workflow to collect various high-fidelity data and auto-annotate it in a single demonstration. AnnotateXR allows users to align virtual models over physical objects, tracked with six degrees-of-freedom (6DOF) sensors. AnnotateXR utilizes a hand tracking capable XR head-mounted display coupled with 6DOF information and collision detection to enable algorithmic segmentation of different actions in videos through its digital twin. The virtual–physical mapping provides a tight bounding volume to generate semantic segmentation masks for the captured image data. Alongside supporting object and action segmentation, we also support other dimensions of annotation required by modern CV, such as human–object, object–object, and rich 3D recordings, all with a single demonstration. Our user study shows AnnotateXR produced over 112,000 annotated data points in 67 min.

Optimizing Automated Optical Inspection: An Adaptive Fusion and Semi-Supervised Self-Learning Approach for Elevated Accuracy and Efficiency in Scenarios with Scarce Labeled Data.

In the field of automatic optical inspection (AOI), this study presents innovative strategies to enhance object detection accuracy while minimizing dependence on large annotated datasets. We initially developed a defect detection model using a dataset of 3579 images across 32 categories, created in collaboration with a major Taiwanese panel manufacturer. This model was evaluated using 12,000 ambiguously labeled images, with improvements achieved through data augmentation and annotation refinement. To address the challenges of limited labeled data, we proposed the Adaptive Fused Semi-Supervised Self-Learning (AFSL) method. This approach, designed for anchor-based object detection models, leverages a small set of labeled data alongside a larger pool of unlabeled data to enable continuous model optimization. Key components of AFSL include the Bounding Box Assigner, Adaptive Training Scheduler, and Data Allocator, which together facilitate dynamic threshold adjustments and balanced training, significantly enhancing the model's performance on AOI datasets. The AFSL method improved the mean average precision (mAP) from 43.5% to 57.1% on the COCO dataset and by 2.6% on the AOI dataset, demonstrating its effectiveness in achieving high levels of precision and efficiency in AOI with minimal labeled data.

Application and investigation of knowledge graph in biomedical field

A large amount of drug and disease research knowledge is scattered in unstructured literature data, presenting significant challenges in text mining within the field of biomedicine. These challenges include handling professional knowledge, integrating related knowledge, and disambiguating different meanings of the same words. Therefore, constructing a biomedical knowledge graph can significantly save expert human resources and make efficient use of medical literature resources. This review paper aims to summarize the construction methods used during the development of Biomedical Knowledge Graphs. It also outlines the latest models and frameworks, such as BioBERT and LSTM+CRF, highlighting their contributions and applications. In addition, this paper points out the limitations of current biomedical knowledge graphs, such as scalability issues and the need for large annotated datasets. To address these limitations, it proposes the use of Apache Spark for improved processing capabilities and transfer learning to enhance model performance and adaptability in diverse biomedical contexts.

ORASIS-MAE Harnesses the Potential of Self-Learning from Partially Annotated Clinical Eye Movement Records

Self-supervised learning (SSL) has gained significant attention in the past decade for its capacity to utilize non-annotated datasets to learn meaningful data representations. In the medical domain, the challenge of constructing large annotated datasets presents a significant limitation, rendering SSL an ideal approach to address this constraint. In this study, we introduce a novel pretext task tailored to stimulus-driven eye movement data, along with a denoising task to improve the robustness against simulated eye tracking failures. Our proposed task aims to capture both the characteristics of the pilot (brain) and the motor (eye) by learning to reconstruct the eye movement position signal using up to 12.5% of the unmasked eye movement signal patches, along with the entire REMOBI target signal. Thus, the encoder learns a high-dimensional representation using a multivariate time series of length 8192 points, corresponding to approximately 40 s. We evaluate the learned representation on screening eight distinct groups of pathologies, including dyslexia, reading disorder, and attention deficit disorder, across four datasets of varying complexity and size. Furthermore, we explore various head architecture designs along with different transfer learning methods, demonstrating promising results with improvements of up to approximately 15%, leading to an overall macro F1 score of 61% and 61.5% on the Saccade and the Vergence datasets, respectively. Notably, our method achieves macro F1 scores of 64.7%, 66.1%, and 61.1% for screening dyslexia, reading disorder, and attention deficit disorder, respectively, on clinical data. These findings underscore the potential of self-learning algorithms in pathology screening, particularly in domains involving complex data such as stimulus-driven eye movement analysis.

A veracity dissemination consistency-based few-shot fake news detection framework by synergizing adversarial and contrastive self-supervised learning

With the rapid growth of social media, fake news (rumors) are rampant online, seriously endangering the health of mainstream social consciousness. Fake news detection (FEND), as a machine learning solution for automatically identifying fake news on Internet, is increasingly gaining the attentions of academic community and researchers. Recently, the mainstream FEND approaches relying on deep learning primarily involves fully supervised fine-tuning paradigms based on pre-trained language models (PLMs), relying on large annotated datasets. In many real scenarios, obtaining high-quality annotated corpora are time-consuming, expertise-required, labor-intensive, and expensive, which presents challenges in obtaining a competitive automatic rumor detection system. Therefore, developing and enhancing FEND towards data-scarce scenarios is becoming increasingly essential. In this work, inspired by the superiority of semi-/self- supervised learning, we propose a novel few-shot rumor detection framework based on semi-supervised adversarial learning and self-supervised contrastive learning, named Detection Yet See Few (DetectYSF). DetectYSF synergizes contrastive self-supervised learning and adversarial semi-supervised learning to achieve accurate and efficient FEND capabilities with limited supervised data. DetectYSF uses Transformer-based PLMs (e.g., BERT, RoBERTa) as its backbone and employs a Masked LM-based pseudo prompt learning paradigm for model tuning (prompt-tuning). Specifically, during DetectYSF training, the enhancement measures for DetectYSF are as follows: (1) We design a simple but efficient self-supervised contrastive learning strategy to optimize sentence-level semantic embedding representations obtained from PLMs; (2) We construct a Generation Adversarial Network (GAN), utilizing random noises and negative fake news samples as inputs, and employing Multi-Layer Perceptrons (MLPs) and an extra independent PLM encoder to generate abundant adversarial embeddings. Then, incorporated with the adversarial embeddings, we utilize semi-supervised adversarial learning to further optimize the output embeddings of DetectYSF during its prompt-tuning procedure. From the news veracity dissemination perspective, we found that the authenticity of the news shared by these collectives tends to remain consistent, either mostly genuine or predominantly fake, a theory we refer to as “news veracity dissemination consistency”. By employing an adjacent sub-graph feature aggregation algorithm, we infuse the authenticity characteristics from neighboring news nodes of the constructed veracity dissemination network during DetectYSF inference. It integrates the external supervisory signals from “news veracity dissemination consistency” to further refine the news authenticity detection results of PLM prompt-tuning, thereby enhancing the accuracy of fake news detection. Furthermore, extensive baseline comparisons and ablated experiments on three widely-used benchmarks demonstrate the effectiveness and superiority of DetectYSF for few-shot fake new detection under low-resource scenarios.

Artificial disfluency detection, uh no, disfluency generation for the masses

Existing approaches for disfluency detection typically require the existence of large annotated datasets. However, current datasets for this task are limited, suffer from class imbalance, and lack some types of disfluencies that are encountered in real-world scenarios. At the same time, augmentation techniques for disfluency detection are not able to model complex types of disfluencies. This limits such approaches to only performing pre-training since the generated data are not indicative of disfluencies that occur in real scenarios and, as a result, cannot be directly used for training disfluency detection models, as we experimentally demonstrate. This imposes significant constraints on the usefulness of such approaches in practice since real disfluencies still have to be collected in order to train the models. In this work, we propose Large-scale ARtificial Disfluency Generation (LARD), a method for automatically generating artificial disfluencies, and more specifically repairs, from fluent text. Unlike existing augmentation techniques, LARD can simulate all the different and complex types of disfluencies. In addition, it incorporates contextual embeddings into the disfluency generation to produce realistic, context-aware artificial disfluencies. LARD can be used effectively for training disfluency detection models, bypassing the requirement of annotated disfluent data. Our empirical evaluation shows that LARD outperforms existing rule-based augmentation methods and increases the accuracy of existing disfluency detectors. In addition, experiments demonstrate that the proposed method can be effectively used in a low-resource setup.

Probabilistic Attention Based on Gaussian Processes for Deep Multiple Instance Learning.

Multiple instance learning (MIL) is a weakly supervised learning paradigm that is becoming increasingly popular because it requires less labeling effort than fully supervised methods. This is especially interesting for areas where the creation of large annotated datasets remains challenging, as in medicine. Although recent deep learning MIL approaches have obtained state-of-the-art results, they are fully deterministic and do not provide uncertainty estimations for the predictions. In this work, we introduce the attention Gaussian process (AGP) model, a novel probabilistic attention mechanism based on Gaussian processes (GPs) for deep MIL. AGP provides accurate bag-level predictions as well as instance-level explainability and can be trained end-to-end. Moreover, its probabilistic nature guarantees robustness to overfit on small datasets and uncertainty estimations for the predictions. The latter is especially important in medical applications, where decisions have a direct impact on the patient's health. The proposed model is validated experimentally as follows. First, its behavior is illustrated in two synthetic MIL experiments based on the well-known MNIST and CIFAR-10 datasets, respectively. Then, it is evaluated in three different real-world cancer detection experiments. AGP outperforms state-of-the-art MIL approaches, including deterministic deep learning ones. It shows a strong performance even on a small dataset with less than 100 labels and generalizes better than competing methods on an external test set. Moreover, we experimentally show that predictive uncertainty correlates with the risk of wrong predictions, and therefore it is a good indicator of reliability in practice. Our code is publicly available.

Building RadiologyNET: an unsupervised approach to annotating a large-scale multimodal medical database

BackgroundThe use of machine learning in medical diagnosis and treatment has grown significantly in recent years with the development of computer-aided diagnosis systems, often based on annotated medical radiology images. However, the lack of large annotated image datasets remains a major obstacle, as the annotation process is time-consuming and costly. This study aims to overcome this challenge by proposing an automated method for annotating a large database of medical radiology images based on their semantic similarity.ResultsAn automated, unsupervised approach is used to create a large annotated dataset of medical radiology images originating from the Clinical Hospital Centre Rijeka, Croatia. The pipeline is built by data-mining three different types of medical data: images, DICOM metadata and narrative diagnoses. The optimal feature extractors are then integrated into a multimodal representation, which is then clustered to create an automated pipeline for labelling a precursor dataset of 1,337,926 medical images into 50 clusters of visually similar images. The quality of the clusters is assessed by examining their homogeneity and mutual information, taking into account the anatomical region and modality representation.ConclusionsThe results indicate that fusing the embeddings of all three data sources together provides the best results for the task of unsupervised clustering of large-scale medical data and leads to the most concise clusters. Hence, this work marks the initial step towards building a much larger and more fine-grained annotated dataset of medical radiology images.

Lip Reading using Deep Learning

Lip reading, the process of interpreting speech by visually observing the movements of the lips, has emerged as a critical area of research with applications spanning communication aids for the hearing impaired, silent speech interfaces, and enhanced human-computer interaction. This paper reviews recent advancements in lip reading technologies, focusing on the integration of machine learning and computer vision techniques. We explore state-of-the-art methods including convolutional neural networks (CNNs), recurrent neural networks (RNNs), and transformer-based models that have significantly improved the accuracy and robustness of lip reading systems. The study highlights the importance of large annotated datasets, such as LipNet and LRW, which have facilitated the training of deep learning models. Additionally, we examine multimodal approaches that combine visual information with audio signals to enhance performance, especially in noisy environments. Despite substantial progress, challenges remain in addressing speaker variability, low resolution, and real-time processing. Future research directions are discussed, emphasizing the need for more diverse datasets, improved model generalization, and real-world application testing. This comprehensive review underscores the potential of advanced lip reading technologies to revolutionize communication accessibility and human-computer interaction. This paper presents the method for Vision based Lip Reading system that uses convolutional neural network (CNN) with attention-based Long Short-Term Memory (LSTM). The dataset includes video clips pronouncing words sentence. The pretrained CNN is used for extracting features from pre- processed video frames which then are processed for learning temporal characteristics by LSTM. The SoftMax layer of architecture provides the result of lip reading. In the present work experiments are performed with two pre-trained models. The system provides 80% accuracy using Tensorflow and ensemble learning. Keywords— CNN; RNN; LSTM; Tensorflow; lip reading; deep learning

A comparative study of large language model-based zero-shot inference and task-specific supervised classification of breast cancer pathology reports.

Although supervised machine learning is popular for information extraction from clinical notes, creating large annotated datasets requires extensive domain expertise and is time-consuming. Meanwhile, large language models (LLMs) have demonstrated promising transfer learning capability. In this study, we explored whether recent LLMs could reduce the need for large-scale data annotations. We curated a dataset of 769 breast cancer pathology reports, manually labeled with 12 categories, to compare zero-shot classification capability of the following LLMs: GPT-4, GPT-3.5, Starling, and ClinicalCamel, with task-specific supervised classification performance of 3 models: random forests, long short-term memory networks with attention (LSTM-Att), and the UCSF-BERT model. Across all 12 tasks, the GPT-4 model performed either significantly better than or as well as the best supervised model, LSTM-Att (average macro F1-score of 0.86 vs 0.75), with advantage on tasks with high label imbalance. Other LLMs demonstrated poor performance. Frequent GPT-4 error categories included incorrect inferences from multiple samples and from history, and complex task design, and several LSTM-Att errors were related to poor generalization to the test set. On tasks where large annotated datasets cannot be easily collected, LLMs can reduce the burden of data labeling. However, if the use of LLMs is prohibitive, the use of simpler models with large annotated datasets can provide comparable results. GPT-4 demonstrated the potential to speed up the execution of clinical NLP studies by reducing the need for large annotated datasets. This may increase the utilization of NLP-based variables and outcomes in clinical studies.

Review on Decision Tree Algorithm in Healthcare Applications

Decision tree algorithms have emerged as a pivotal tool in healthcare, offering substantial benefits in diagnostics, prognosis, and health monitoring. This paper provides a comprehensive review of decision tree applications in medical settings, highlighting their ability to simplify complex decision-making processes and improve accuracy in disease diagnosis and outcome prediction. By dissecting various research studies and clinical implementations, we demonstrate the versatility of decision trees in handling diverse datasets—from genetic markers to electronic health records and real-time patient data. This review also explores the integration of decision trees with machine learning techniques to enhance diagnostic procedures and prognostic evaluations, underscoring the significant role of these algorithms in advancing personalized medicine and public health strategies. Challenges such as data sensitivity, privacy concerns, and the need for large annotated datasets are discussed to provide a balanced perspective on the capabilities and limitations of decision tree algorithms in healthcare. Through this analysis, we aim to illuminate the transformative potential of decision trees in improving patient care and streamlining healthcare operations.

From classification to segmentation with explainable AI: A study on crack detection and growth monitoring

Monitoring surface cracks in infrastructure is crucial for structural health monitoring. Automatic visual inspection offers an effective solution, especially in hard-to-reach areas. Machine learning approaches have proven their effectiveness but typically require large annotated datasets for supervised training. Once a crack is detected, monitoring its severity often demands precise segmentation of the damage. However, pixel-level annotation of images for segmentation is labor-intensive. To mitigate this cost, one can leverage explainable artificial intelligence (XAI) to derive segmentations from the explanations of a classifier, requiring only weak image-level supervision. This paper proposes applying this methodology to segment and monitor surface cracks. We evaluate the performance of various XAI methods and examine how this approach facilitates severity quantification and growth monitoring. Results reveal that while the resulting segmentation masks may exhibit lower quality than those produced by supervised methods, they remain meaningful and enable severity monitoring, thus reducing substantial labeling costs. Code and data available at https://github.com/EPFL-IMOS/crack-explanations.

Unified multimodal fusion transformer for few shot object detection for remote sensing images

Object detection is a fundamental computer vision task with wide applications in remote sensing, but traditional methods strongly rely on large annotated datasets which are difficult to obtain, especially for novel object classes. Few-shot object detection (FSOD) aims to address this by using detectors to learn from very limited labeled data. Recent work fuse multi-modalities like image–text pairs to tackle data scarcity but require external region proposal network (RPN) to align cross-modal pairs which leads to a bias towards base classes and insufficient cross-modal contextual learning. To address these problems, we propose a unified multi-modal fusion transformer (UMFT), which extracts visual features from ViT and textual encodings from BERT to align multi-modal representations in an end-to-end manner. Specifically, affinity-guided fusion (AFM) captures semantically related image–text pairs by modeling their affinity relationships to selectively combine most informative pairs. In addition, cross-modal correlation module (CCM) captures discriminative inter-modal patterns between image and text representations and filters out unrelated features to enhance cross-modal alignment. By leveraging AFM to focus on semantic relationships and CCM to refine inter-modal features, the model better aligns multimodal data without RPN. These representations are passed to detection decoder that predicts bounding boxes, probabilities of class and cross-modal attributes. Evaluation of UMFT on benchmark datasets NWPU VHR-10 and DIOR demonstrated its ability to leverage limited image–text training data via dynamic fusion, achieving new state-of-the-art mean average precision (mAP) for few-shot object detection. Our code will be publicly available at https://github.com/abdullah-azeem/umft.

FetalBrainAwareNet: Bridging GANs with anatomical insight for fetal ultrasound brain plane synthesis

Over the past decade, deep-learning (DL) algorithms have become a promising tool to aid clinicians in identifying fetal head standard planes (FHSPs) during ultrasound (US) examination. However, the adoption of these algorithms in clinical settings is still hindered by the lack of large annotated datasets. To overcome this barrier, we introduce FetalBrainAwareNet, an innovative framework designed to synthesize anatomically accurate images of FHSPs. FetalBrainAwareNet introduces a cutting-edge approach that utilizes class activation maps as a prior in its conditional adversarial training process. This approach fosters the presence of the specific anatomical landmarks in the synthesized images. Additionally, we investigate specialized regularization terms within the adversarial training loss function to control the morphology of the fetal skull and foster the differentiation between the standard planes, ensuring that the synthetic images faithfully represent real US scans in both structure and overall appearance. The versatility of our FetalBrainAwareNet framework is highlighted by its ability to generate high-quality images of three predominant FHSPs using a singular, integrated framework. Quantitative (Fréchet inception distance of 88.52) and qualitative (t-SNE) results suggest that our framework generates US images with greater variability compared to state-of-the-art methods. By using the synthetic images generated with our framework, we increase the accuracy of FHSP classifiers by 3.2% compared to training the same classifiers solely with real acquisitions. These achievements suggest that using our synthetic images to increase the training set could provide benefits to enhance the performance of DL algorithms for FHSPs classification that could be integrated in real clinical scenarios.