Multiclass Classification of Gastric Cancer using Artificial Neural Network

• Published in: 2023 International Conference on Energy, Power, Environment, Control, and Computing (ICEPECC) pp. 1 - 6
• Main Authors: Aslam, Muhammad Aqeel, Mahnoor, Shahzadi, Bilal, Ahmad, Cheema, Saman, Munir, Muhammad Asif, Hassan, Khawaja Humble
• Format: Conference Proceeding
• Language: English
• Published: IEEE 08.03.2023
• Subjects: Artificial Neural Network; Artificial neural networks; Breath Samples; Design automation; Hardware; Machine Learning; Predictive models; Programmable logic arrays; Receiver Operating Characteristics; Sensitivity; Solid modeling
• DOI: 10.1109/ICEPECC57281.2023.10209471

Gastric Cancer (GC) remains on the top of the list in death troll due to cancer worldwide and has a high mortality due to its vague symptoms. With the lack of effective biomarkers and absence of symptoms, it is an urgent need to develop a noninvasive, rapid, and highly accurate technique for GC screening at early-stage detection, prognostics, and diagnostics. Breath shows a promising multi-constituent oral fluid with a noninvasive source. The proposed solution is based on the unsupervised learning model that is capable to classify the cancer patient in different categories based on the spread of the disease. The proposed model can distinguish the level of disease as healthy person, moderate spread and acute spread. We have three hundred saliva samples of volunteers which comprise 102 EGC, 156 AGC, and 110 Healthy persons. This study has shown pivotal information in distinguishing EGC, AGC from healthy persons, which have a great clinical translational prospect. The developed Artificial Neural Network (ANN) based predictive model has produced an accuracy of 81.4%, 68.6% and 92.5 % for AGC, EGC and Healthy persons.