Research Article
Automated Diagnosis of Gastric Cancer Through VGG16 Convolutional Neural Networks
@INPROCEEDINGS{10.4108/eai.28-4-2025.2358101, author={Sahil Kumar Gupta and Raushan Kumar Gupta and Samip Aanand Shah and Adrija Adhikary and Aryan Kumar Sah and Gayathri Ramasamy}, title={Automated Diagnosis of Gastric Cancer Through VGG16 Convolutional Neural Networks}, proceedings={Proceedings of the 4th International Conference on Information Technology, Civil Innovation, Science, and Management, ICITSM 2025, 28-29 April 2025, Tiruchengode, Tamil Nadu, India, Part II}, publisher={EAI}, proceedings_a={ICITSM PART II}, year={2025}, month={10}, keywords={gastric cancer pathology survival prediction machine learning convolutional neural networks (cnn) resnet-50 tissue classification}, doi={10.4108/eai.28-4-2025.2358101} }- Sahil Kumar Gupta
Raushan Kumar Gupta
Samip Aanand Shah
Adrija Adhikary
Aryan Kumar Sah
Gayathri Ramasamy
Year: 2025
Automated Diagnosis of Gastric Cancer Through VGG16 Convolutional Neural Networks
ICITSM PART II
EAI
DOI: 10.4108/eai.28-4-2025.2358101
Abstract
Stomach cancer remains among the leading causes of mortality, this is why its early detection is another need for a new cancer diagnostic test. This paper illustrates design a CNN-based model for predicting gastric cancer. Specifically, the GASHissDB which is an open access image database with normal and abnormal gastric tissue images was used for this work. Some changes made in the input images included normalization and converting the images to tensors with reference to the enhancement of the model. Fine-tuned five CNN architectures: For feature extractors, used VGG16, ResNet30 and ResNet50 were used to extract features from the images for binary classification tasks. MobileNet-v2 used a different classifier head where the last dropout layer was removed and replaced by fully connected layers to produce probabilities by using the sigmoid function of Keras. The selected performance measures used to benchmark the models were accuracy, precision, recall or sensitivity, and F1-score. Real-time screen interaction was implemented in Streamlit for image classification in addition to probability scoring. The suggested approach possed high accuracy and depended on the chosen indexes which can be a potential applicative tool for clinicians in diagnosis of Gastric cancer. This research also backs the trend where pre-trained CNNs can be utilized in the classification of medical images and throws light into the integration of AI with the clinical related setting.
