Practical X-ray Gastric Cancer Diagnostic Support Using Refined Stochastic Data Augmentation and Hard Boundary Box Training

TL;DR

This paper tackles the challenge of scalable gastric cancer screening by leveraging X-ray imaging interpreted with AI. It introduces two innovations, refined stochastic gastric image augmentation (R-sGAIA) and hard boundary box training (HBBT), applied to a general object detector (EfficientDet-D7) to detect cancer regions while utilizing healthy controls for training. The approach achieves a sensitivity of 90.2%—higher than expert performance at 85.5%—with a processing time of 0.51 seconds per image and a 42.5% cancer-positivity rate among detected boxes, outperforming baselines with a 5.9-point F1 gain. Practically, this system highlights ROI to radiologists, reducing workload while enabling mass screening, and the authors provide public access to the R-sGAIA-HBBT codebase.

Abstract

Endoscopy is widely used to diagnose gastric cancer and has a high diagnostic performance, but it must be performed by a physician, which limits the number of people who can be diagnosed. In contrast, gastric X-rays can be taken by radiographers, thus allowing a much larger number of patients to undergo imaging. However, the diagnosis of X-ray images relies heavily on the expertise and experience of physicians, and few machine learning methods have been developed to assist in this process. We propose a novel and practical gastric cancer diagnostic support system for gastric X-ray images that will enable more people to be screened. The system is based on a general deep learning-based object detection model and incorporates two novel techniques: refined probabilistic stomach image augmentation (R-sGAIA) and hard boundary box training (HBBT). R-sGAIA enhances the probabilistic gastric fold region and provides more learning patterns for cancer detection models. HBBT is an efficient training method that improves model performance by allowing the use of unannotated negative (i.e., healthy control) samples, which are typically unusable in conventional detection models. The proposed system achieved a sensitivity (SE) for gastric cancer of 90.2\%, higher than that of an expert (85.5%). Under these conditions, two out of five candidate boxes identified by the system were cancerous (precision = 42.5%), with an image processing speed of 0.51 seconds per image. The system also outperformed methods using the same object detection model and state-of-the-art data augmentation by showing a 5.9-point improvement in the F1 score. In summary, this system efficiently identifies areas for radiologists to examine within a practical time frame, thus significantly reducing their workload.

Figures (6)

• Figure 1: Overview of the proposed gastric cancer diagnostic support system.
• Figure 2: Overview of refined stochastic gastric image augmentation (R-sGAIA).
• Figure 3: Probability of being detected as a gastric fold edge to be highlighted.
• Figure 4: Positive example and five typical false positives provided by experts (as shown in red box with "malignant" label. In HBBT, these false detection regions are labeled as the "hard boundary box" class, and the model is retrained accordingly.). (a) Positive example. (b) The area overlapping the osteophytes and spinous processes of the vertebral body. (c) Area where folds are gathered due to lack of air in the stomach. (d) Area where the duodenal Kerckring’s folds overlap with the stomach. (e) Areas where the mucosal surface of the stomach is irregular due to chronic gastritis. (f) Areas of advanced cancer that appear as wall changes rather than masses.
• Figure 5: Example of augmented images (leftmost: original, others: enhanced images with R-sGAIA).