摘要：—In medical research, non-invasive diagnostic toolshave become an emerging technique for the diagnosis of fataldisease in the last few years. Saliva analysis for the detection ofGastric cancer (GC) also belongs to this powerful new researchfield. According to the WHO, cancer is heterogeneous diseasewith different subtypes. Early prognosis and diagnosis are keyto improve the survival rate. It has become necessary in cancerresearch to facilitate the subsequent clinical management ofpatients. In this study, we have found 10 Amino acid biomarkersin saliva and extracted 19 fingerprint Raman bands producedby these biomarkers, that can be used to distinguish cancerpatients from healthy persons. These Amino acid biomarkersvary according to the health condition of the patient. ComputerAided Diagnostic (CAD) techniques allow us to learn thecommon and hidden patterns from the input datasets andpredict the cancer status most accurately and efficiently. Wehave developed a multilayer feedforward neural network usinga scaled conjugate gradient backpropagation technique. Theproposed method produces an accuracy of 92.27%, sensitivity of94.8 %, and specificity of 90.2%. In conclusion, our approachusing the saliva analysis and Amino acid biomarkers in salivahas enabled us to reliably detect gastric cancer at very highaccuracy.
