poly-DART: A discrete algebraic reconstruction technique for polychromatic X-ray CT

Application of MLP neural network to predict X-ray spectrum from tube voltage, filter material, and filter thickness used in medical imaging systems

The X-ray energy spectrum is crucial for image quality and dosage assessment in mammography, radiography, fluoroscopy, and CT which are frequently used for the diagnosis of many diseases including but not limited to patients with cardiovascular and cerebrovascular diseases. X-ray tubes have an electron filament (cathode), a tungsten/rubidium target (anode) oriented at an angle, and a metal filter (aluminum, beryllium, etc.) that may be placed in front of an exit window. When cathode electrons meet the anode, they generate X-rays with varied energies, creating a spectrum from zero to the electrons' greatest energy. In general, the energy spectrum of X-rays depends on the electron beam's energy (tube voltage), target angle, material, filter thickness, etc. Thus, each imaging system's X-ray energy spectrum is unique to its tubes. The primary goal of the current study is to develop a clever method for quickly estimating the X-ray energy spectrum for a variety of tube voltages, filter materials, and filter thickness using a small number of unique spectra. In this investigation, two distinct filters made of beryllium and aluminum with thicknesses of 0.4, 0.8, 1.2, 1.6, and 2 mm were employed to obtain certain limited X-ray spectra for tube voltages of 20, 30, 40, 50, 60, 80, 100, 130, and 150 kV. The three inputs of 150 Multilayer Perceptron (MLP) neural networks were tube voltage, filter type, and filter thickness to forecast the X-ray spectra point by point. After training, the MLP neural networks could predict the X-ray spectra for tubes with voltages between 20 and 150 kV and two distinct filters made of aluminum and beryllium with thicknesses between 0 and 2 mm. The presented methodology can be used as a suitable, fast, accurate and reliable alternative method for predicting X-ray spectrum in medical applications. Although a technique was put out in this work for a particular system that was the subject of Monte Carlo simulations, it may be applied to any genuine system.

Proposing an intelligent technique based on radial basis function neural network to forecast the energy spectrum of diagnostic X-ray imaging systems

In digital subtraction angiography (digital subtraction total cerebral angiography), cardiac and macrovascular cardiography, anorectal radiology, fluoroscopy, and computed tomography (CT), a prior knowledge to X-ray energy spectrum is crucial for assessing the image quality and also calculating patient X-ray dosage. The present investigation's main objective is to propose an intelligent technique for faster calculating X-ray energy spectrum of medical imaging systems with different exposure settings of tube voltage, filter material, and thickness based on limited specific spectra. In this study, Monte Carlo N Particle (MCNP) simulation code was initially used to generate some limited X-ray spectra for tube voltages of 20, 30, 40, 50, 80, 100, 130, and 150 kV for two different filters of beryllium and aluminum with thicknesses of 0. 4, 0.8, 1.2, 1.6 and 2 mm. Tube voltage, type, and thickness of filter were used as the 3 inputs of 150 Radial Basis Function Neural Network (RBFNN) to forecast point by point of the X-ray spectrum. After training, the RBFNNs could forecast most of the X-ray spectra for tube voltages in the range of 20-150 kV and two various filters of aluminum and beryllium with thicknesses in the range of 0-2 mm.

Deep learning reconstruction of digital breast tomosynthesis images for accurate breast density and patient-specific radiation dose estimation

The two-dimensional nature of mammography makes estimation of the overall breast density challenging, and estimation of the true patient-specific radiation dose impossible. Digital breast tomosynthesis (DBT), a pseudo-3D technique, is now commonly used in breast cancer screening and diagnostics. Still, the severely limited 3rd dimension information in DBT has not been used, until now, to estimate the true breast density or the patient-specific dose. This study proposes a reconstruction algorithm for DBT based on deep learning specifically optimized for these tasks. The algorithm, which we name DBToR, is based on unrolling a proximal-dual optimization method. The proximal operators are replaced with convolutional neural networks and prior knowledge is included in the model. This extends previous work on a deep learning-based reconstruction model by providing both the primal and the dual blocks with breast thickness information, which is available in DBT. Training and testing of the model were performed using virtual patient phantoms from two different sources. Reconstruction performance, and accuracy in estimation of breast density and radiation dose, were estimated, showing high accuracy (density <±3%; dose <±20%) without bias, significantly improving on the current state-of-the-art. This work also lays the groundwork for developing a deep learning-based reconstruction algorithm for the task of image interpretation by radiologists.