Accurate and robust estimation the scatterer size from Optical Coherence Tomography (OCT) images has the potential to provide a diagnostically useful biomarker of disease. In the past, Mie Theory with curve fitting, the autocorrelation of the spectrum and the bandwidth of the correlation of the derivative (COD) have been explored for this purpose. However, these approaches are very challenging to apply to scatterer sizes below 4 μm due to their inherent lack of accuracy or, in the case of COD, the limitations imposed by Mie Theory itself. On the other hand, the Fractal Dimension (FD) has been used in the analysis of OCT images to examine the structural variations of biological tissues. In this study, we propose the use of fractal analysis to robustly and accurately estimate the size of scatterers as small as 0.1 μm in diameter. The box counting method was used to define the statistical characteristics of the FD, first calculated for individual neighborhoods and, subsequently, for the entire image. Using a prudently selected subset of these features, the scatterer size of microsphere phantoms was estimated with a mean error of 32.8 %. The proposed method will, of course, have to be tested further both on an expanded phantom set but also on human normal and disease tissue. However, given the preliminary results presented in this study, this approach has the potential to be further developed and to perform in vivo scatterer size estimation.