Diabetic retinopathy is a sequela of diabetes. Biomedical research endeavours to find fast, easy and sensitive diagnostic tools, which can be used to identify such diseases at an early stage, in order to provide preventive therapeutic measures. Our work proposes two experimental methodologies to discern the difference between blood serum from healthy subjects and that of diabetes/early stage diabetes retinopathy patients. In the first approach we establish Raman spectroscopy, which probes constituent molecules in the serum, as a reliable analytical technique for this purpose. It is demonstrated that other than individual Raman spectral analysis, chemometric analyses of the data, obtained from Raman measurements, can be used to discriminate healthy and diabetic serum samples. Further, we propose a novel technique, in which the geometry of the fractal patterns formed by gold nanoparticles, in the presence of serum samples, is used in finding unique diagnostic signatures. As-prepared gold colloidal particles form directional fractals with four-fold symmetry. We have shown that the fractal structures formed by gold nanoparticles are distinctly different when incubated with serum samples. Most importantly, we observe unique arc-fractals by gold particles in presence of diabetes retinopathy serum samples. We quantify the fractal geometry by measuring the persistence length, which can be used as a biomarker in pathological reporting systems. We believe that our findings will motivate the medical researchers to employ the proposed methodologies for diagnosing diabetes/early stage diabetic retinopathy on larger number of subjects in clinical trials.