Convolutional LSTM model for cine image prediction of abdominal motion.

Objective.In this study, we tackle the challenge of latency in magnetic resonance linear accelerator (MR-Linac) systems, which compromises target coverage accuracy in gated real-time radiotherapy. Our focus is on enhancing motion prediction precision in abdominal organs to address this issue. We developed a convolutional long short-term memory (convLSTM) model, utilizing 2D cine magnetic resonance (cine-MR) imaging for this purpose.Approach.Our model, featuring a sequence-to-one architecture with six input frames and one output frame, employs structural similarity index measure (SSIM) as loss function. Data was gathered from 17 cine-MRI datasets using the Philips Ingenia MR-sim system and an Elekta Unity MR-Linac equivalent sequence, focusing on regions of interest (ROIs) like the stomach, liver, pancreas, and kidney. The datasets varied in duration from 1 to 10 min.Main results.The study comprised three main phases: hyperparameter optimization, individual training, and transfer learning with or without fine-tuning. Hyperparameters were initially optimized to construct the most effective model. Then, the model was individually applied to each dataset to predict images four frames ahead (1.24-3.28 s). We evaluated the model's performance using metrics such as SSIM, normalized mean square error, normalized correlation coefficient, and peak signal-to-noise ratio, specifically for ROIs with target motion. The average SSIM values achieved were 0.54, 0.64, 0.77, and 0.66 for the stomach, liver, kidney, and pancreas, respectively. In the transfer learning phase with fine-tuning, the model showed improved SSIM values of 0.69 for the liver and 0.78 for the kidney, compared to 0.64 and 0.37 without fine-tuning.Significance. The study's significant contribution is demonstrating the convLSTM model's ability to accurately predict motion for multiple abdominal organs using a Unity-equivalent MR sequence. This advancement is key in mitigating latency issues in MR-Linac radiotherapy, potentially improving the precision and effectiveness of real-time treatment for abdominal cancers.

Validity of the line-pair bar-pattern method in the measurement of the modulation transfer function (MTF) in megavoltage imaging.

The measurement of the modulation transfer function (MTF) of an imaging device is a common requirement in evaluating radiographic detector performance. Such measurements are considered mandatory in detector development research, and may also be carried out as part of routine quality assurance (QA) checks of image quality. Traditionally, MTF measurement has been performed by imaging either a narrow slit or a sharp edge in order to generate a line spread function, whose Fourier transform provides the MTF on a near-continuous frequency domain. Much less commonly employed is the method of square-wave line-pair modulations, in which the modulation response to bar resolution targets contained in a bar pattern is used to estimate the MTF at discrete spatial frequencies. While the slit and edge methods offer advantages of accuracy and a well-know standardized protocol for measurement based on several decades of development, their major limitation is the difficult and time-consuming experimental setup that is necessary to ensure accurate measurements. On the other hand, the bar pattern offers the advantage of a quick, simple, and easy measurement without the need for a complex experimental setup, with the main disadvantages of the technique being a pseudo-normalization that may lead to an overestimated MTF, and corrections for removing higher-order frequency harmonics that require interpolating between discrete spatial frequencies. Therefore, bar patterns are traditionally used for qualitative imaging applications like detector QA in terms of relative and arbitrarily defined spatial resolution metrics, while slit and edge methods are preferred for quantitative MTF measurements. Compared to diagnostic x rays, MTF measurements using megavoltage x rays are further complicated by low x-ray attenuation and excessive Compton scattering. In this work, a method to measure the MTF of megavoltage x-ray detectors based on imaging square-wave line pairs with improved near-zero-frequency normalization was developed as an adaptation to previously reported methods. Monte Carlo simulations were used to identify an improved normalization condition with which the accuracy of the MTF determined from line-pair modulations could be enhanced considerably compared to previously used techniques. Slit, edge, and bar-pattern measurements were performed to obtain the MTF of commercial megavoltage imaging devices including portal film and electronic portal imaging devices. A comparison of the MTF measurements from the three techniques was used to ascertain the validity of the proposed bar-pattern method for accurate and reliable measurement of MTF for megavoltage imagers. Statistical analyses revealed no significant differences between the bar-pattern method and the standard slit and edge techniques, indicating very good agreement (mean difference within +/- 3%). These results indicated the potential for line-pair bar patterns to be used more effectively than in the past for traditional QA imaging as well as for quantitative MTF measurement in detector development research.

Effect of recombination in a high quantum efficiency prototype ionization-chamber-based electronic portal imaging device.

The quantum efficiency (QE) of an imaging detector can be increased by utilizing a thick, high-density detection medium to increase the number of quantum interactions. However, image quality is more accurately described by the detection quantum efficiency (DQE). If a significant fraction of the increase in the number of detected quanta from a thick, dense detector were to result in useful imaging signal, this represents a favorable case where enhanced QE leads to increased DQE. However, for ionization-type detectors, one factor that limits DQE is the recombination between ion pairs that acts as a secondary quantum sink due to which enhancement in QE may not result in higher DQE depending on the extent of the signal loss from recombination. Therefore, an analysis of signal loss mechanisms or quantum sinks in an imaging system is essential for validating the overall benefit of high QE detectors. In this paper, a study of ion recombination as a secondary quantum sink is presented for a high QE prototype ion-chamber-based electronic portal imaging device (EPID): the kinestatic charge detector (KCD). The KCD utilizes a high pressure noble gas (krypton or xenon at 100 atm) and an arbitrarily large detector thickness (of the order of centimeters), resulting in a high QE imager. Compared with commercial amorphous silicon flat panel imagers that provide DQE(0) approximately 0.01, the KCD has much higher DQE. Studies indicated that DQE(0) = 0.20 for 6.1 cm thick, 100 atm (rho = 3.4 g/cm3) xenon chamber, and DQE(0)=0.34 for a 9.1 cm thick chamber. A series of experiments was devised and conducted to determine the signal loss due to recombination for a KCD chamber. The measurements indicated a fractional recombination loss of about 14% for a krypton chamber and about 18% for a xenon chamber under standard operating conditions (100 atm chamber pressure and 1275 V/cm electric field intensity). A theoretical treatment of the effect of recombination on imaging signal-to-noise ratio was applied to quantify the loss in DQE. These calculations indicated that recombination had a limited effect (<2%) on DQE under standard operating conditions. This was validated by good agreement between experimentally measured DQE and that obtained using Monte Carlo simulations that did not account for recombination.