Search

In recent years, deep-learning-based image processing has emerged as a valuable tool for medical imaging owing to its high performance. However, the quality of deep-learning-based methods heavily relies on the amount of training data; the high cost of acquiring a large dataset is a limitation to their utilization in medical fields. Herein, based on deep learning, we developed a computed tomography (CT) modality conversion method requiring only a few unsupervised images. The proposed method is based on CycleGAN with several extensions tailored for CT images, which aims at preserving the structure in the processed images and reducing the amount of training data. This method was applied to realize the conversion of megavoltage computed tomography (MVCT) to kilovoltage computed tomography (kVCT) images. Training was conducted using several datasets acquired from patients with head and neck cancer. The size of the datasets ranged from 16 slices (two patients) to 2745 slices (137 patients) for MVCT and 2824 slices (98 patients) for kVCT. The required size of the training data was found to be as small as a few hundred slices. By statistical and visual evaluations, the quality improvement and structure preservation of the MVCT images converted by the proposed model were investigated. As a clinical benefit, it was observed by medical doctors that the converted images enhanced the precision of contouring. We developed an MVCT to kVCT conversion model based on deep learning, which can be trained using only a few hundred unpaired images. The stability of the model against changes in data size was demonstrated. This study promotes the reliable use of deep learning in clinical medicine by partially answering commonly asked questions, such as "Is our data sufficient?" and "How much data should we acquire?", Comment: 3+27 pages, 13 figures, version published in Medical Physics

Statistical iterative reconstruction is expected to improve the image quality of megavoltage computed tomography (MVCT). However, one of the challenges of iterative reconstruction is its large computational cost. The purpose of this work is to develop a fast iterative reconstruction algorithm by combining several iterative techniques and by optimizing reconstruction parameters. Megavolt projection data was acquired from a TomoTherapy system and reconstructed using our statistical iterative reconstruction. Total variation was used as the regularization term and the weight of the regularization term was determined by evaluating signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), and visual assessment of spatial resolution using Gammex and Cheese phantoms. Gradient decent with an adaptive convergence parameter, ordered subset expectation maximization (OSEM), and CPU/GPU parallelization were applied in order to accelerate the present reconstruction algorithm. The SNR and CNR of the iterative reconstruction were several times better than that of filtered back projection (FBP). The GPU parallelization code combined with the OSEM algorithm reconstructed an image several hundred times faster than a CPU calculation. With 500 iterations, which provided good convergence, our method produced a 512$\times$512 pixel image within a few seconds. The image quality of the present algorithm was much better than that of FBP for patient data. An image from the iterative reconstruction in TomoTherapy can be obtained within few seconds by fine-tuning the parameters. The iterative reconstruction with GPU was fast enough for clinical use, and largely improve the MVCT images., Comment: 11 pages, 4 figures

We have observed a "$K^-pp$"-like structure in the $d(\pi^+,K^+)$ reaction at 1.69 GeV/$c$. In this reaction $\Lambda(1405)$ hyperon resonance is expected to be produced as a doorway to form the $K^-pp$ through the $\Lambda^*p\rightarrow K^-pp$ process. However, most of the produced $\Lambda(1405)$'s would escape from deuteron without secondary reactions. Therefore, coincidence of high-momentum ($>$ 250~MeV/$c$) proton(s) in large emission angles ($39^\circ<\theta_{lab.}<122^\circ$) was requested to enhance the signal-to-background ratio. A broad enhancement in the proton coincidence spectra are observed around the missing-mass of 2.27 GeV/$c^2$, which corresponds to the $K^-pp$ binding energy of 95 $^{+18}_{-17}$ (stat.) $^{+30}_{-21}$ (syst.) MeV and the width of 162 $^{+87}_{-45}$ (stat.) $^{+66}_{-78}$ (syst.) MeV., Comment: 8 pages, 4 figures, submitted to PTEP

We have measured an inclusive missing-mass spectrum of the $d(\pi^+, K^+)$ reaction at the pion incident momentum of 1.69 GeV/$c$ at the laboratory scattering angles between 2$^\circ$ and 16$^\circ$ with the missing-mass resolution of 2.7 $\pm$ 0.1 MeV/$c^2$ (FWHM) at the missing mass of 2.27 GeV/$c^{2}$. In this Letter, we first try to understand the spectrum as a simple quasi-free picture based on several known elementary cross sections, considering the neutron/proton Fermi motion in deuteron. While gross spectrum structures are well understood in this picture, we have observed two distinct deviations; one peculiar enhancement at 2.13 GeV/$c^2$ is due to the $\Sigma N$ cusp, and the other notable feature is a shift of a broad bump structure, mainly originating from hyperon resonance productions of $\Lambda(1405)$ and $\Sigma(1385)^{+/0}$, by about 22.4 $\pm$ 0.4 (stat.) $^{+2.7}_{-1.7}$ (syst.) MeV/$c^2$ toward the low-mass side, which is calculated in the kinematics of a proton at rest as the target., Comment: 8 pages, 3 figures, submitted to PTEP

Multistage adaptive tests (MSTs) have gained increasing popularity in recent years. MST is a balanced compromise between linear test forms (i.e., paper-and-pencil testing and computer-based testing) and traditional item-level computer-adaptive testing (CAT). It combines the advantages of both. On one hand, MST is adaptive (and therefore more efficient than linear tests). On the other hand, unlike CAT, it allows test developers to review test forms before administration, and it allows examinees to review and revise answers. Despite the advantages of MST, there is little literature on the details of heuristic automated assembly of MST and on the investigation of MST in the context of classification tests. In this study, we designed a MST for a large-scale classification test and performed the automated test assembly using a heuristic method. We then compared the performance of the MST with that of a linear test form and a CAT using computer simulation. The automated test assembly was successful. In comparing MST and CAT, we did observe a trade-off in measurement accuracy and item bank usage. For classification purposes, however, MST provided classification accuracy as good as that from CAT, with more efficient item bank usage. (Contains 5 tables, 5 figures and 2 footnotes.)