A key to the successful use of targeted cancer therapy is the ability to preselect patients who are likely to benefit from the treatment according to molecular markers. Assessment for predicting therapy response is mostly done using tumor biopsies. However, these might not truly represent all of the patient's malignant cells because of tumor heterogeneity and/or clonal evolution during disease progression. One potential strategy that can complement primary tumor biopsy is the analysis of circulating tumor cells (CTCs). In this study, we analyzed CTCs of patients with gastric cancer (GC) to find those who were likely to benefit from trastuzumab therapies. We developed an imaging-based method that enabled CTC identification simultaneously with evaluation of HER2 gene amplification (the 3D-IF-FISH method). Then we performed a study enrolling 101 GC patients in whom we analyzed CTCs by both 3D-IF-FISH and an FDA-approved CellSearch system. As compared with the CellSearch system, 3D-IF-FISH methods identified a higher number of patients whose primary tumors were HER2- but who had HER2+ CTCs, suggesting that the 3D-IF-FISH method is effective in preselecting patients for trastuzumab therapies. To demonstrate this, we performed an exploratory clinical study to evaluate the clinical benefits of trastuzumab treatment for advanced GC patients (n = 15) whose primary tumors were HER2-, but whose CTCs showed HER2 amplification. An interim evaluation after the first stage showed that these preselected patients had response rates comparable to those reported in the trastuzumab-plus-chemotherapy arm of the ToGA study. The present study offers a new, non-invasive strategy to select patients who are likely to benefit from trastuzumab-based therapies, despite their primary biopsy being HER2-negative. (UMIN ID: UMIN000008622).