Objective: To design and validate an artificial neural network (ANN) model for differentiating cervical smears with endometrial carcinoma from those with benign endometrial cells. Study Design: A total of 39 liquid-based cervical smears (28 benign and 11 histologically proven endometrial carcinoma cases) were included. Cytological features were studied for each case and included the number of cell clusters, size of the clusters, three-dimensional clustering, singly dispersed endometrial cells, nucleoli, nuclear margin irregularity, chromatin pattern, necrosis, leukophagocytosis, and micronucleus score. Semiquantitative grading was done for each variable. An artificial neural network with the architecture 17-3-1 (17 input, 3 hidden nodes, and 1 output node) was designed using appropriate neurointelligence software. Results: The cases were divided into training (n=27), validation (n=6), and test (n=6) set by the ANN program. After adequate training the ANN program was able to diagnose all cases with benign endometrial cells and endometrial carcinoma in the test set as well as in the training set. In the validation set all 3 benign cases were correctly diagnosed, and only 1 out of 3 endometrial carcinoma cases was misdiagnosed as benign. Conclusion: We have created a successful ANN model for the cytodiagnosis of endometrial carcinoma, which may be useful in the future. © Science Printers and Publishers, Inc.