Early gastric cancer detection and lesion segmentation based on deep learning and gastroscopic images

Kezhi Zhang,Haibao Wang,Yaru Cheng,Hongyan Liu,Qi Gong,Qian Zeng,Tao Zhang,Guoqiang Wei,Zhi Wei,Dong Chen

DOI: https://doi.org/10.1038/s41598-024-58361-8

IF: 4.6

2024-04-04

Scientific Reports

Abstract:Gastric cancer is a highly prevalent disease that poses a serious threat to public health. In clinical practice, gastroscopy is frequently used by medical practitioners to screen for gastric cancer. However, the symptoms of gastric cancer at different stages of advancement vary significantly, particularly in the case of early gastric cancer (EGC). The manifestations of EGC are often indistinct, leading to a detection rate of less than 10%. In recent years, researchers have focused on leveraging deep learning algorithms to assist medical professionals in detecting EGC and thereby improve detection rates. To enhance the ability of deep learning to detect EGC and segment lesions in gastroscopic images, an Improved Mask R-CNN (IMR-CNN) model was proposed. This model incorporates a "Bi-directional feature extraction and fusion module" and a "Purification module for feature channel and space" based on the Mask R-CNN (MR-CNN). Our study includes a dataset of 1120 images of EGC for training and validation of the models. The experimental results indicate that the IMR-CNN model outperforms the original MR-CNN model, with Precision, Recall, Accuracy, Specificity and F1-Score values of 92.9%, 95.3%, 93.9%, 92.5% and 94.1%, respectively. Therefore, our proposed IMR-CNN model has superior detection and lesion segmentation capabilities and can effectively aid doctors in diagnosing EGC from gastroscopic images.

multidisciplinary sciences

What problem does this paper attempt to address?

The paper primarily addresses the issues of Early Gastric Cancer (EGC) detection and lesion segmentation. Specifically, the research team developed an improved version of the Mask R-CNN model (IMR-CNN) aimed at enhancing the accuracy of detecting and segmenting early gastric cancer through endoscopic images.

Currently, in clinical practice, although white light endoscopy is the standard method for screening gastric cancer, its accuracy highly depends on the professional skills and experience of the endoscopist, resulting in a low detection rate of early gastric cancer, generally not exceeding 10%. Additionally, the workload of manually analyzing a large number of medical images can also affect diagnostic results. To improve this situation, researchers utilized deep learning technology to assist doctors in enhancing the efficiency and accuracy of early gastric cancer detection.

The IMR-CNN model proposed in the paper includes two key modules: the "Bidirectional Feature Extraction and Fusion Module" and the "Feature Channel and Spatial Purification Module." These improvements enable the model to not only accurately detect early gastric cancer but also perform high-precision lesion segmentation. Experimental results show that compared to the original Mask R-CNN model, the IMR-CNN model has significant improvements in major evaluation metrics such as precision, recall, accuracy, specificity, and F1 score, reaching 92.9%, 95.3%, 93.9%, 92.5%, and 94.1%, respectively.

Therefore, this study aims to improve the detection rate of early gastric cancer, reduce the risk of misdiagnosis and missed diagnosis, and ultimately help doctors better diagnose the disease by proposing the IMR-CNN model.