- - - MSc, Department of Medical Physics and Biomedical Engineering, Faculty of Medicine, Shahid Beheshti University of Medical Sciences, Tehran, Iran
- - - PhD, Department of Medical Physics and Biomedical Engineering, Faculty of Medicine, Shahid Beheshti University of Medical Sciences, Tehran, Iran
- - - PhD candidate, Department of Medical Physics and Biomedical Engineering, Faculty of Medicine, Shahid Beheshti University of Medical Sciences, Tehran, Iran
- - - MD, Faculty of Medicine, Golestan University of Medical Sciences, Gorgan, Iran
- - - MD, Faculty of Medicine, Islamic Azad University, Tehran Medical Branch, Tehran, Iran

Background: Given the extensive use and preferred diagnostic method in common mammography tests for screening and diagnosis of breast cancer, there is concern about the increased dose absorbed by the patient due to the sensitivity of the breast tissue. Objective: This study aims to evaluate the entrance surface air kerma (ESAK) before irradiation to the patient through its estimation.Material and Methods: In this descriptive paper, firstly, a phantom was used to measure some data, including ESAK, Kvp, mAs, HVL, and type of filter/target. Secondly, the MultiLayer Perceptron (MLP) neural network model was trained with Levenberg-Marquardt (LM) backpropagation training algorithm and finally, ESAK was estimated. Results: Based on results obtained from the program in different neuron numbers, it was found that the number of ۳۵ neurons is the most optimal value, offering a regression coefficient of ۹۵.۷%. The Mean Squared Error (MSE) for all data was ۰.۴۳۷ mGy and accounting for ۴.۸% of the output range changes, predicting ۹۵.۲% accuracy in the present research. Conclusion: Using neural networks in ESAK prediction, the method proposed in the present research leads to the possible ESAK estimation of patients before X-Ray. The results suggested that the regression coefficient represented ۴.۳% difference between the kerma measured by solid-state dosimeter in the radiation field and the value predicted in the research. In comparison with the Monte-Carlo simulation method, this method has better accuracy.

Mammography, Neural Networks, Computer, Radiation Dosimeters