In diagnostic radiology, the air kerma is an essential parameter. Radiologists consider the air kerma, when calculating organ doses and dangers to patients. The intensity of the radiation beam is represented by the air kerma, which is the value of energy wasted by a photon as it travels through air. Because of the heel effect in X-ray sources, air kerma varies throughout the field of medical imaging systems. One possible contributor to this discrepancy is the X-ray tube's voltage. In this study, an approach has been proposed for predicting the air kerma anywhere inside the field of X-ray beams utilized in medical diagnostic imaging systems. As a first step, a diagnostic imaging system was modelled using the Monte Carlo N-Particle platform. We used a tungsten target and aluminum and beryllium filters of varying thicknesses to recreate the X-ray tube. The air kerma has been measured in different parts of the conical X-ray beam that is working at 30, 50, 70, 90, 110, 130, and 150 kV. This gives enough data for training neural networks. The voltage of the X-ray tube, filter type, filter thickness, and the coordinates of each point used to calculate the air kerma were all inputs to the MLP neural network. The MLP architecture, known for its significant advancements in research and expanding applications, was trained to predict the quantity of air kerma as its output. Specifically, by considering X-ray tube filters of varying thicknesses, the trained MLP model demonstrated its capability to accurately predict the air kerma at every point within the X-ray field for a range of X-ray tube voltages typically used in medical diagnostic radiography (30–150 kV).