Cancer nanomedicine involving nanotechnology-based drugs and in vivo imaging agents is an active field of nanoscience that provides new ways of enhancing therapeutic and diagnostic efficacy. Translating cancer nanomedicine mainly comes from rational and scalable design of nanoparticles to achieve versatile properties including specific size because nanomaterials whose properties confer unique advantages can only optimize clinical impact. Here, a facile scalable synthesis of highly monodisperse small silica nanoparticles was developed by screening various alkaline buffer solutions as catalysts. The size of silica nanoparticles ranging from 7 to 30 nm was finely controlled by varying the reaction temperature. Moderate sized silica nanoparticles in the range of 30 to 50 nm and large sized silica nanoparticles (>100 nm) were readily synthesized by in situ adding tetraethylorthosilicate (TEOS) and applying the Stöber method in the reaction solution using small silica nanoparticles as the seeds, respectively. Having shown the ability to precisely synthesize size controlled silica nanoparticles with a process compatible with good manufacturing practices, we performed in vivo fluorescence imaging and immunofluorescence analysis of sentinel lymph nodes (SLNs) with the synthesized nanoparticles having different sizes to investigate the size effect for effective identification of SLNs. The synthesized nanoparticles with a diameter of 12 nm showed effective SLN uptake within 10 min after intradermal injection both in noninvasive and in intraoperative imaging mode and were localized evenly inside the SLN, whereas the 120 nm sized nanoparticles failed to identify the SLN with noninvasive imaging at 10 min post-injection and distributed only in the medulla region not in the superficial cortex of the SLN. Taken together, a new facile scalable synthesis technique to achieve fine size controlling capability from very small silica nanoparticles (7 nm) was developed and it made possible to investigate the optimal size of nanoparticles for efficient SLN mapping which is still controversial.