Improved AlexNet–Bi-LSTM Model-Based Gastrointestinal Cancer Classification with Shape and Texture Feature Set
Year: 2026
Type: article
Abstract: Background: Gastrointestinal (GI) cancer classification is the process of identifying and categorizing cancers that occur in the GI tract based on their morphological, histopathological, and molecular characteristics. Existing GI cancer classification systems often struggle with low interpretability and limited generalizability across diverse datasets. This study aims to improve the accuracy of classifying cancer. Methods: The Improved AlexNet–Bi-LSTM model-based Gastrointestinal Cancer Classification (IAB-GCC) with a shape andtexture feature set is proposed to improve diagnostic accuracy. The input images are from a dataset consisting of 411,890 unique image patches generated from formalin-fixed paraffin-embedded (FFPE) diagnostic slides. Then, the proposed IAB-GCC process begins with input images, which are preprocessed using an Adaptive Weighted Gaussian Filtering (AWGF) technique to reduce noise and enhance image quality. Key features including Weighted Shannon Entropy (WSE)-based features, texture features using Gray Level Co-occurrence Matrix (GLCM), shape, and morphological features are then extracted and fed into the Alex–-Bi-LSTM network, where the extracted features are passed into an Improved AlexNet (IANet) model to train the features, and then the learned feature maps of IANet are subjected to a Bi-directional Long Short-Term Memory (Bi-LSTM) model for classification. Results: The combination of IANet and Bi-LSTM enables the system to classify cancer cases as stable or unstable, which provides precise and reliable diagnostic results. The proposed IAB-GCC model achieves a higher precision 97% and an accuracy score of 96% in the classification process of GI cancers. Conclusion: The proposed hybrid IAB-GCC provides accurate, reliable, and interpretable GI cancer classification, demonstrating potential for clinical diagnostic applications.
Language: English
Topic: AI in cancer detection
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W7225742920
FWCI:
Cites:
Cited by:
Related to: