In this paper we present a study done in a sample of 123 diabetic patients. The photomotor reflex of the patients, as a reaction of an external and well-controlled light stimulation, was digitally measured to acquire variables and characteristics like pupil time of response, pupil diameter over the time and mathematical behavior of the photomotor response. These variables were added into a data base in order to computationally analyze them. Variables like gender, edge, weight, capillary blood glucose (CB) and glycated hemoglobin (GH), which were previously recollected on each patient, were also included in the data base for the analysis.

The mathematical behavior of the photomotor response of each patient was statistically characterized to get an approximation of the signal to a sigmoidal function. From this characterization a set of 4 values of the signal along the time were also considered for the data base.

A clustering based in the k-means algorithm was applied to the resulted data base in order to get specific groups with similar characteristics. The results indicate that, even all the patients where previously diagnosed with diabetes, there are differences between them and 4 groups of patients were detected. It means that not every patient with diabetes must receive the same medical treatment. In this way, we get an innovative model for the recognition of specific characteristics in diabetic patients for improving the quality of medical attention.