Raman spectroscopy and machine learning for early detection of gastric cancer and Helicobacter pylori with gastric juice

Gastric cancer is a leading cause of cancer-related mortality and highlights the need for early detection of gastric cancer and Helicobacter pylori (HP) infection, which is a major risk factor. Early non-invasive and convenient diagnostic tools capable of capturing the dynamic molecular alterations during carcinogenesis and HP infection is needed. In this study, we use Raman spectroscopy and machine learning algorithms to detect different gastric lesions and HP infection condition with gastric juice samples. 133 patients from Peking University Third Hospital were involved and categorized into groups based on histopathological diagnosis: early gastric cancer (EGC), dysplasia (DYS), intestinal metaplasia (IM), and chronic superficial gastritis (CSG), with further classification based on HP infection. The stacked machine learning model demonstrated high diagnostic performance, achieving 90% accuracy, 90% sensitivity, and 97% specificity in distinguishing pathological stages, along with 96% accuracy, 96% sensitivity, and 96% specificity in HP detection. The multilayer perceptron (MLP) model based on gastric juice Raman spectroscopy showed excellent discrimination capability, with an AUC of 0.98 for differentiating controls from patients with DYS and EGC. Additionally, Raman spectroscopy achieved an AUC of 0.95 in distinguishing control gastric mucosa from precancerous lesions (IM, DYS) and EGC. The approach offers a rapid, accurate, and minimally invasive diagnostic tool, demonstrating significant potential for clinical application in rapid and accurate detection of precancerous lesions, early gastric cancer, and HP infection.