Background: The highly increase of Alzheimer's disease among human lead to increasing the demand of finding a reliable way to diagnose its symptoms at the early stages. Recent researches in this area demonstrate that the signal complexity analysis of the electroencephalogram can be useful in prognosis the development of this illness form mild cognitive impairment to Alzheimer's disease. The focus of this study was on approximate entropy and proposing an effective approach for using this criterion to diagnose the mild cognitive impairment.Methods: In this research, the electroencephalograms of 16 normal subjects and 11 patients were used. The signals were captured based on 10-20 international system for 30 minutes. In the preprocessing phase, the artefacts were eliminated both by visually inspection by a specialist physician and using band pass filter. In the processing phase, different scenarios were considered and applied to define the different parameters of approximate entropy. Finally, the results were analyzed using t-test to optimize the define protocol of the entropy and find the appropriate channels for diagnosing the disease.Findings: A protocol for extracting the complexity based on approximate entropy was determined, in which the difference of the entropy of normal subjects and patients were more remarkable. By using this protocol, the number of appropriate channels for diagnosing the disease increased (P < 0.05). These results also showed decreasing the gray matter volume in the patients with mild cognitive impairment.Conclusion: Using the entropy measurements for different channels of patients with mild cognitive impairment, demonstrate that the amount of complexity of signals decreased.