AI-enhanced X-ray spectrum reconstruction for medical imaging system.

For the purpose of assessing image quality and calculating patient X-ray dosage in radiology, computed tomography (CT), fluoroscopy, mammography, and other fields, it is necessary to have prior knowledge of the X-ray energy spectrum. The main components of an X-ray tube are an electron filament, also known as the cathode, and an anode, which is often made of tungsten or rubidium and angled at a certain angle. At the point where the electrons generated by the cathode and the anode make contact, a spectrum of X-rays with energies spanning from zero to the maximum energy value of the released electrons is created. Typically, the energy distribution of X-rays depends on various parameters, including the energy of the electron beam (tube voltage) and the angle of the anode. As a result, the X-ray energy spectrum is specific to the configuration of each tube and imaging system. This study aims to develop an efficient method for rapidly determining the X-ray energy spectrum of medical imaging systems across a broad range of tube voltages and anode angles using a limited set of specific spectra. The investigation began by simulating seven different anode angles between 12° and 24° using the Monte Carlo N Particle (MCNP) method. The X-ray spectra were generated for tube voltages of 20, 30, 40, 50, 60, 70, 80, 100, 130, and 150 kV. In order to make point-by-point X-ray spectrum predictions, 150 Radial Basis Function Neural Networks (RBFNNs) were trained using tube voltage and anode angle as inputs. The RBFNNs were trained to anticipate the X-ray spectra for different target angles and tube voltages between 20 and 150 kV. This research only used Monte Carlo simulations to represent one system; however, the approach shown here is generalizable to any real-world system.