The links between H. pylori infection and gastric and oesophageal cancer are either inconsistent or still poorly understood. In this paper, we present a machine learning approach that detects immuno- reactive protein bands from HelicoBlot assays of human serum samples related to these diseases. Feature encoding is applied via transfer learning by utilising existing computer vision networks on scanned HelicoBlo strips. XGBoost classifier is employed to predict the existence of relevant protein bands, and a 95% average detection AU- ROC is achieved. The proposed approach helps to reduce the amount of time required for man- ual annotation and the subjective bias associated with it. Moreover, the approach can be applied to large-scale epidemiological studies investigating the relationship between infectious disease and risk of cancer.