A sensitive, non-invasive and specific tear biomarker sensor for diabetic retinopathy (DR) could be of great significance in clinical practice. Vascular endothelial growth factor (VEGF) was targeted as a tear biomarker for DR. In principle, we used hybridization chain reaction (HCR) and CeO2 nanoparticles as a cascade signal amplification strategy to construct a reusable aptasensor in search of VEGF. The report probes H1 and H2 modified with medium CeO2 nanoparticles were added in and HCR reaction. The first analysis method for a quantitative detection of VEGF was obtained using CeO2 nanoparticles to catalyze H2O2 to generate electrical signals. The fabricated biosensor exhibited an ultrasensitive detection of VEGF down to 0.27 fg mL−1, response range from 1 fg mL−1 to 0.1 ng mL−1. Subsequently the aptasensor carried out a second repeated assembly to obtain a second signal by strand displacement reaction (SDR). The ratio of the former and latter signals was used to carry out a second analysis method for quantifying VEGF. The second analysis method had detection limits as low as 7.39 fg mL-1, ranging from 10 fg mL−1 to 0.1 ng mL−1. The aptasensor could maintain above 99 % signal in five days and shows the acceptable reproducibility. For a practical tear biomarker detection, the analysis of the tears was carried out simultaneously using ELISA and the aptasensor which showed excellent precision. This non-invasive screening method provides a potential tool for the early diagnosis and monitoring of DR and may improve screening success among high-risk patients with diabetes.