Development of a Deep Learning System to Detect Esophageal Cancer by Barium Esophagram

Evaluation of Spectrum-Aided Visual Enhancer (SAVE) in Esophageal Cancer Detection Using YOLO Frameworks

The early detection of esophageal cancer presents a substantial difficulty, which contributes to its status as a primary cause of cancer-related fatalities. This study used You Only Look Once (YOLO) frameworks, specifically YOLOv5 and YOLOv8, to predict and detect early-stage EC by using a dataset sourced from the Division of Gastroenterology and Hepatology, Ditmanson Medical Foundation, Chia-Yi Christian Hospital. The dataset comprised 2741 white-light images (WLI) and 2741 hyperspectral narrowband images (HSI-NBI). They were divided into 60% training, 20% validation, and 20% test sets to facilitate robust detection. The images were produced using a conversion method called the spectrum-aided vision enhancer (SAVE). This algorithm can transform a WLI into an NBI without requiring a spectrometer or spectral head. The main goal was to identify dysplasia and squamous cell carcinoma (SCC). The model's performance was evaluated using five essential metrics: precision, recall, F1-score, mAP, and the confusion matrix. The experimental results demonstrated that the HSI model exhibited improved learning capabilities for SCC characteristics compared with the original RGB images. Within the YOLO framework, YOLOv5 outperformed YOLOv8, indicating that YOLOv5's design possessed superior feature-learning skills. The YOLOv5 model, when used in conjunction with HSI-NBI, demonstrated the best performance. It achieved a precision rate of 85.1% (CI95: 83.2-87.0%, p < 0.01) in diagnosing SCC and an F1-score of 52.5% (CI95: 50.1-54.9%, p < 0.01) in detecting dysplasia. The results of these figures were much better than those of YOLOv8. YOLOv8 achieved a precision rate of 81.7% (CI95: 79.6-83.8%, p < 0.01) and an F1-score of 49.4% (CI95: 47.0-51.8%, p < 0.05). The YOLOv5 model with HSI demonstrated greater performance than other models in multiple scenarios. This difference was statistically significant, suggesting that the YOLOv5 model with HSI significantly improved detection capabilities.

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.