Authors:
G. Rajesh, A. B. Gurulakshmi, B. Manjunatha, Thirumalraj Karthikeyan
Addresses:
Department of Electronics and Communication Engineering, New Horizon College of Engineering, Bengaluru, Karnataka, India. Department of Mechanical Engineering, New Horizon College of Engineering, Bengaluru, Karnataka, India. Department of Artificial Intelligence, Trichy Research Labs, Quest Technologies, Tiruchirappalli, Tamil Nadu, India.
Abstract:
Deep learning is a new method that is frequently utilised in the medical industry for illness diagnosis. For the analysis of (EGC), a new research avenue has been devised. Because of their efficacy, computer-aided diagnostic (CAD) systems minimise mortality rates. We proposed a new approach for feature extraction, utilising a Deep Q Network (DQN), for this study. To identify stomach cancer from breath samples, an upgraded Swin Transformer network was coupled with the suggested feature extraction approach. An Improved Attention Swin with feature pyramid network (IAS-FPN) architecture aims to identify different phases of GC proficiently by adding context information to the fundamental Swin Transformer. Furthermore, to prevent small-cell information loss, an enhanced weighted bidirectional network is provided by incorporating skip connections with enhanced Dwarf Mongoose Optimisation (IDMO). To be more specific, we found fifty to differentiate EGC, Advanced Gastric Cancer (AGC), and healthy people. By learning the characteristics and preserving the construction of the input tasters, this CAD scheme lowers the distance between the input and output. The traits were derived from unlabeled breath-sample data. The suggested network design achieves good results for advanced gastric cancer classification using breath data, with an overall accuracy of 94.07%. Furthermore, the created model achieves high accuracy and F1 score, making it ideal for scientific use.
Keywords: Deep Learning; Breath Analysis; Computer-Aided Diagnostic (CAD); Deep Q Network (DQN); Advanced Gastric Cancer (AGC); F-Score Value; Fundamental Swin Transformer.
Received: 10/09/2024, Revised: 28/12/2024, Accepted: 23/02/2025, Published: 05/06/2025
DOI: 10.64091/ATIHL.2025.000169
AVE Trends in Intelligent Health Letters, 2025 Vol. 2 No. 2 , Pages: 94-107