Multi-stage domain-specific pretraining for improved detection and localization of Barrett's neoplasia: A comprehensive clinically validated study

Foundation models in gastrointestinal endoscopic AI: Impact of architecture, pre-training approach and data efficiency

Pre-training deep learning models with large data sets of natural images, such as ImageNet, has become the standard for endoscopic image analysis. This approach is generally superior to training from scratch, due to the scarcity of high-quality medical imagery and labels. However, it is still unknown whether the learned features on natural imagery provide an optimal starting point for the downstream medical endoscopic imaging tasks. Intuitively, pre-training with imagery closer to the target domain could lead to better-suited feature representations. This study evaluates whether leveraging in-domain pre-training in gastrointestinal endoscopic image analysis has potential benefits compared to pre-training on natural images. To this end, we present a dataset comprising of 5,014,174 gastrointestinal endoscopic images from eight different medical centers (GastroNet-5M), and exploit self-supervised learning with SimCLRv2, MoCov2 and DINO to learn relevant features for in-domain downstream tasks. The learned features are compared to features learned on natural images derived with multiple methods, and variable amounts of data and/or labels (e.g. Billion-scale semi-weakly supervised learning and supervised learning on ImageNet-21k). The effects of the evaluation is performed on five downstream data sets, particularly designed for a variety of gastrointestinal tasks, for example, GIANA for angiodyplsia detection and Kvasir-SEG for polyp segmentation. The findings indicate that self-supervised domain-specific pre-training, specifically using the DINO framework, results into better performing models compared to any supervised pre-training on natural images. On the ResNet50 and Vision-Transformer-small architectures, utilizing self-supervised in-domain pre-training with DINO leads to an average performance boost of 1.63% and 4.62%, respectively, on the downstream datasets. This improvement is measured against the best performance achieved through pre-training on natural images within any of the evaluated frameworks. Moreover, the in-domain pre-trained models also exhibit increased robustness against distortion perturbations (noise, contrast, blur, etc.), where the in-domain pre-trained ResNet50 and Vision-Transformer-small with DINO achieved on average 1.28% and 3.55% higher on the performance metrics, compared to the best performance found for pre-trained models on natural images. Overall, this study highlights the importance of in-domain pre-training for improving the generic nature, scalability and performance of deep learning for medical image analysis. The GastroNet-5M pre-trained weights are made publicly available in our repository: huggingface.co/tgwboers/GastroNet-5M_Pretrained_Weights.

ReFs: A hybrid pre-training paradigm for 3D medical image segmentation

Self-supervised learning (SSL) has achieved remarkable progress in medical image segmentation. The application of an SSL algorithm often follows a two-stage training process: using unlabeled data to perform label-free representation learning and fine-tuning the pre-trained model on the downstream tasks. One issue of this paradigm is that the SSL step is unaware of the downstream task, which may lead to sub-optimal feature representation for a target task. In this paper, we propose a hybrid pre-training paradigm that is driven by both self-supervised and supervised objectives. To achieve this, a supervised reference task is involved in self-supervised learning, aiming to improve the representation quality. Specifically, we employ the off-the-shelf medical image segmentation task as reference, and encourage learning a representation that (1) incurs low prediction loss on both SSL and reference tasks and (2) leads to a similar gradient when updating the feature extractor from either task. In this way, the reference task pilots SSL in the direction beneficial for the downstream segmentation. To this end, we propose a simple but effective gradient matching method to optimize the model towards a consistent direction, thus improving the compatibility of both SSL and supervised reference tasks. We call this hybrid pre-training paradigm reference-guided self-supervised learning (ReFs), and perform it on a large-scale unlabeled dataset and an additional reference dataset. The experimental results demonstrate its effectiveness on seven downstream medical image segmentation benchmarks.

Deep Learning in Barrett's Esophagus Diagnosis: Current Status and Future Directions

Barrett's esophagus (BE) represents a pre-malignant condition characterized by abnormal cellular proliferation in the distal esophagus. A timely and accurate diagnosis of BE is imperative to prevent its progression to esophageal adenocarcinoma, a malignancy associated with a significantly reduced survival rate. In this digital age, deep learning (DL) has emerged as a powerful tool for medical image analysis and diagnostic applications, showcasing vast potential across various medical disciplines. In this comprehensive review, we meticulously assess 33 primary studies employing varied DL techniques, predominantly featuring convolutional neural networks (CNNs), for the diagnosis and understanding of BE. Our primary focus revolves around evaluating the current applications of DL in BE diagnosis, encompassing tasks such as image segmentation and classification, as well as their potential impact and implications in real-world clinical settings. While the applications of DL in BE diagnosis exhibit promising results, they are not without challenges, such as dataset issues and the "black box" nature of models. We discuss these challenges in the concluding section. Essentially, while DL holds tremendous potential to revolutionize BE diagnosis, addressing these challenges is paramount to harnessing its full capacity and ensuring its widespread application in clinical practice.

Machine learning applications for early detection of esophageal cancer: a systematic review

INTRODUCTION

Esophageal cancer (EC) is a significant global health problem, with an estimated 7th highest incidence and 6th highest mortality rate. Timely diagnosis and treatment are critical for improving patients' outcomes, as over 40% of patients with EC are diagnosed after metastasis. Recent advances in machine learning (ML) techniques, particularly in computer vision, have demonstrated promising applications in medical image processing, assisting clinicians in making more accurate and faster diagnostic decisions. Given the significance of early detection of EC, this systematic review aims to summarize and discuss the current state of research on ML-based methods for the early detection of EC.

METHODS

We conducted a comprehensive systematic search of five databases (PubMed, Scopus, Web of Science, Wiley, and IEEE) using search terms such as "ML", "Deep Learning (DL (", "Neural Networks (NN)", "Esophagus", "EC" and "Early Detection". After applying inclusion and exclusion criteria, 31 articles were retained for full review.

RESULTS

The results of this review highlight the potential of ML-based methods in the early detection of EC. The average accuracy of the reviewed methods in the analysis of endoscopic and computed tomography (CT (images of the esophagus was over 89%, indicating a high impact on early detection of EC. Additionally, the highest percentage of clinical images used in the early detection of EC with the use of ML was related to white light imaging (WLI) images. Among all ML techniques, methods based on convolutional neural networks (CNN) achieved higher accuracy and sensitivity in the early detection of EC compared to other methods.

CONCLUSION

Our findings suggest that ML methods may improve accuracy in the early detection of EC, potentially supporting radiologists, endoscopists, and pathologists in diagnosis and treatment planning. However, the current literature is limited, and more studies are needed to investigate the clinical applications of these methods in early detection of EC. Furthermore, many studies suffer from class imbalance and biases, highlighting the need for validation of detection algorithms across organizations in longitudinal studies.

Optical Biopsy of Dysplasia in Barrett's Oesophagus Assisted by Artificial Intelligence

Optical biopsy in Barrett's oesophagus (BE) using endocytoscopy (EC) could optimize endoscopic screening. However, the identification of dysplasia is challenging due to the complex interpretation of the highly detailed images. Therefore, we assessed whether using artificial intelligence (AI) as second assessor could help gastroenterologists in interpreting endocytoscopic BE images. First, we prospectively videotaped 52 BE patients with EC. Then we trained and tested the AI pm distinct datasets drawn from 83,277 frames, developed an endocytoscopic BE classification system, and designed online training and testing modules. We invited two successive cohorts for these online modules: 10 endoscopists to validate the classification system and 12 gastroenterologists to evaluate AI as second assessor by providing six of them with the option to request AI assistance. Training the endoscopists in the classification system established an improved sensitivity of 90.0% (+32.67%, p < 0.001) and an accuracy of 77.67% (+13.0%, p = 0.020) compared with the baseline. However, these values deteriorated at follow-up (-16.67%, p < 0.001 and -8.0%, p = 0.009). Contrastingly, AI-assisted gastroenterologists maintained high sensitivity and accuracy at follow-up, subsequently outperforming the unassisted gastroenterologists (+20.0%, p = 0.025 and +12.22%, p = 0.05). Thus, best diagnostic scores for the identification of dysplasia emerged through human-machine collaboration between trained gastroenterologists with AI as the second assessor. Therefore, AI could support clinical implementation of optical biopsies through EC.

Deep Learning Approaches for Automatic Localization in Medical Images

Recent revolutionary advances in deep learning (DL) have fueled several breakthrough achievements in various complicated computer vision tasks. The remarkable successes and achievements started in 2012 when deep learning neural networks (DNNs) outperformed the shallow machine learning models on a number of significant benchmarks. Significant advances were made in computer vision by conducting very complex image interpretation tasks with outstanding accuracy. These achievements have shown great promise in a wide variety of fields, especially in medical image analysis by creating opportunities to diagnose and treat diseases earlier. In recent years, the application of the DNN for object localization has gained the attention of researchers due to its success over conventional methods, especially in object localization. As this has become a very broad and rapidly growing field, this study presents a short review of DNN implementation for medical images and validates its efficacy on benchmarks. This study presents the first review that focuses on object localization using the DNN in medical images. The key aim of this study was to summarize the recent studies based on the DNN for medical image localization and to highlight the research gaps that can provide worthwhile ideas to shape future research related to object localization tasks. It starts with an overview on the importance of medical image analysis and existing technology in this space. The discussion then proceeds to the dominant DNN utilized in the current literature. Finally, we conclude by discussing the challenges associated with the application of the DNN for medical image localization which can drive further studies in identifying potential future developments in the relevant field of study.

Convolutional Neural Networks for the evaluation of cancer in Barrett's esophagus: Explainable AI to lighten up the black-box

Even though artificial intelligence and machine learning have demonstrated remarkable performances in medical image computing, their level of accountability and transparency must be provided in such evaluations. The reliability related to machine learning predictions must be explained and interpreted, especially if diagnosis support is addressed. For this task, the black-box nature of deep learning techniques must be lightened up to transfer its promising results into clinical practice. Hence, we aim to investigate the use of explainable artificial intelligence techniques to quantitatively highlight discriminative regions during the classification of early-cancerous tissues in Barrett's esophagus-diagnosed patients. Four Convolutional Neural Network models (AlexNet, SqueezeNet, ResNet50, and VGG16) were analyzed using five different interpretation techniques (saliency, guided backpropagation, integrated gradients, input × gradients, and DeepLIFT) to compare their agreement with experts' previous annotations of cancerous tissue. We could show that saliency attributes match best with the manual experts' delineations. Moreover, there is moderate to high correlation between the sensitivity of a model and the human-and-computer agreement. The results also lightened that the higher the model's sensitivity, the stronger the correlation of human and computational segmentation agreement. We observed a relevant relation between computational learning and experts' insights, demonstrating how human knowledge may influence the correct computational learning.