As one of the most popular approaches in artificial intelligence, deep learning (DL) has attracted a lot of attention in the medical physics field over the past few years. The goals of this topical review article are twofold. First, we will provide an overview of the method to medical physics researchers interested in DL to help them start the endeavor. Second, we will give in-depth discussions on the DL technology to make researchers aware of its potential challenges and possible solutions. As such, we divide the article into two major parts. The first part introduces general concepts and principles of DL and summarizes major research resources, such as computational tools and databases. The second part discusses challenges faced by DL, present available methods to mitigate some of these challenges, as well as our recommendations.

Recent developments in radiotherapy therapy demand high computation powers to solve challenging problems in a timely fashion in a clinical environment. The graphics processing unit (GPU), as an emerging high-performance computing platform, has been introduced to radiotherapy. It is particularly attractive due to its high computational power, small size, and low cost for facility deployment and maintenance. Over the past few years, GPU-based high-performance computing in radiotherapy has experienced rapid developments. A tremendous amount of study has been conducted, in which large acceleration factors compared with the conventional CPU platform have been observed. In this paper, we will first give a brief introduction to the GPU hardware structure and programming model. We will then review the current applications of GPU in major imaging-related and therapy-related problems encountered in radiotherapy. A comparison of GPU with other platforms will also be presented.

Cone-beam CT (CBCT) has become the standard image guidance tool for patient setup in image-guided radiation therapy. However, due to its large illumination field, scattered photons severely degrade its image quality. While kernel-based scatter correction methods have been used routinely in the clinic, it is still desirable to develop Monte Carlo (MC) simulation-based methods due to their accuracy. However, the high computational burden of the MC method has prevented routine clinical application. This paper reports our recent development of a practical method of MC-based scatter estimation and removal for CBCT. In contrast with conventional MC approaches that estimate scatter signals using a scatter-contaminated CBCT image, our method used a planning CT image for MC simulation, which has the advantages of accurate image intensity and absence of image truncation. In our method, the planning CT was first rigidly registered with the CBCT. Scatter signals were then estimated via MC simulation. After scatter signals were removed from the raw CBCT projections, a corrected CBCT image was reconstructed. The entire workflow was implemented on a GPU platform for high computational efficiency. Strategies such as projection denoising, CT image downsampling, and interpolation along the angular direction were employed to further enhance the calculation speed. We studied the impact of key parameters in the workflow on the resulting accuracy and efficiency, based on which the optimal parameter values were determined. Our method was evaluated in numerical simulation, phantom, and real patient cases. In the simulation cases, our method reduced mean HU errors from 44 to 3 HU and from 78 to 9 HU in the full-fan and the half-fan cases, respectively. In both the phantom and the patient cases, image artifacts caused by scatter, such as ring artifacts around the bowtie area, were reduced. With all the techniques employed, we achieved computation time of less than 30 s including the time for both the scatter estimation and CBCT reconstruction steps. The efficacy of our method and its high computational efficiency make our method attractive for clinical use.

Purpose: In the treatment planning process of intensity‐modulated radiation therapy (IMRT), a human planner operates the treatment planning system (TPS) to adjust treatment planning parameters, for example, dose volume histogram (DVH) constraints’ locations and weights, to achieve a satisfactory plan for each patient. This process is usually time‐consuming, and the plan quality depends on planer’s experience and available planning time. In this study, we proposed to model the behaviors of human planners in treatment planning by a deep reinforcement learning (DRL)‐based virtual treatment planner network (VTPN), such that it can operate the TPS in a human‐like manner for treatment planning.Methods and Materials: Using prostate cancer IMRT as an example, we established the VTPN using a deep neural network developed. We considered an in‐house optimization engine with a weighted quadratic objective function. Virtual treatment planner network was designed to observe an intermediate plan DVHs and decide the action to improve the plan by changing weights and threshold dose in the objective function. We trained the VTPN in an end‐to‐end DRL process in 10 patient cases. A plan score was used to measure plan quality. We demonstrated the feasibility and effectiveness of the trained VTPN in another 64 patient cases.Results: Virtual treatment planner network was trained to spontaneously learn how to adjust treatment planning parameters to generate high‐quality treatment plans. In the 64 testing cases, with initialized parameters, quality score was 4.97 (±2.02), with 9.0 being the highest possible score. Using VTPN to perform treatment planning improved quality score to 8.44 (±0.48).Conclusions: To our knowledge, this was the first time that intelligent treatment planning behaviors of human planner in external beam IMRT are autonomously encoded in an artificial intelligence system. The trained VTPN is capable of behaving in a human‐like way to produce high‐quality plans.

A number of image-processing problems can be formulated as optimization problems. The objective function typically contains several terms specifically designed for different purposes. Parameters in front of these terms are used to control the relative importance among them. It is of critical importance to adjust these parameters, as quality of the solution depends on their values. Tuning parameters are a relatively straight forward task for a human, as one can intuitively determine the direction of parameter adjustment based on the solution quality. Yet manual parameter tuning is not only tedious in many cases, but also becomes impractical when a number of parameters exist in a problem. Aiming at solving this problem, this paper proposes an approach that employs deep reinforcement learning to train a system that can automatically adjust parameters in a human-like manner. We demonstrate our idea in an example problem of optimization-based iterative computed tomography (CT) reconstruction with a pixel-wise total-variation regularization term. We set up a parameter-tuning policy network (PTPN), which maps a CT image patch to an output that specifies the direction and amplitude by which the parameter at the patch center is adjusted. We train the PTPN via an end-to-end reinforcement learning procedure. We demonstrate that under the guidance of the trained PTPN, reconstructed CT images attain quality similar or better than those reconstructed with manually tuned parameters.