Cone beam computed tomography (CBCT) plays an important role in image-guided radiation therapy (IGRT), while having disadvantages of severe shading artifact caused by the reconstruction using scatter contaminated and truncated projections. The purpose of this study is to develop a deep convolutional neural network (DCNN) method for improving CBCT image quality. CBCT and planning computed tomography (pCT) image pairs from 20 prostate cancer patients were selected. Subsequently, each pCT volume was pre-aligned to the corresponding CBCT volume by image registration, thereby leading to registered pCT data (pCT_r). Next, a 39-layer DCNN model was trained to learn a direct mapping from the CBCT to the corresponding pCT_r images. The trained model was applied to a new CBCT data set to obtain improved CBCT (i-CBCT) images. The resulting i-CBCT images were compared to pCT_r using the spatial non-uniformity (SNU), the peak-signal-to-noise ratio (PSNR) and the structural similarity index measure (SSIM). The image quality of the i-CBCT has shown a substantial improvement on spatial uniformity compared to that of the original CBCT, and a significant improvement on the PSNR and the SSIM compared to that of the original CBCT and the enhanced CBCT by the existing pCT-based correction method. We have developed a DCNN method for improving CBCT image quality. The proposed method may be directly applicable to CBCT images acquired by any commercial CBCT scanner.

To determine the risk factors of severe radiation pneumonitis (RP) after stereotactic body radiation therapy (SBRT) for primary or secondary lung tumors. From January 2003 to March 2009, SBRT was performed on 117 patients (32 patients before 2005 and 85 patients after 2006) with lung tumors (primary = 74 patients and metastatic/recurrent = 43 patients) in our institution. In the current study, the results on cases with severe RP (grades 4-5) were evaluated. Serum Krebs von den Lungen-6 (KL-6) and serum Surfactant protein-D (SP-D) were used to predict the incidence of RP. A shadow of interstitial pneumonitis (IP) on the CT image before performing SBRT was also used as an indicator for RP. Since 2006, patients have been prescreened for biological markers (KL-6 & SP-D) as well as checking for an IP-shadow in CT. Grades 4-5 RP was observed in nine patients (7.7%) after SBRT and seven of these cases (6.0%) were grade 5 in our institution. A correlation was found between the incidence of RP and higher serum KL-6 & SP-D levels. IP-shadow in patient's CT was also found to correlate well with the severe RP. Severe RP was reduced from 18.8% before 2005 to 3.5% after 2006 (p = 0.042). There was no correlation between the dose volume histogram parameters and these severe RP patients. Patients presenting with an IP shadow in the CT and a high value of the serum KL-6 & SP-D before SBRT treatment developed severe radiation pneumonitis at a high rate. The reduction of RP incidence in patients treated after 2006 may have been attributed to prescreening of the patients. Therefore, pre-screening before SBRT for an IP shadow in CT and serum KL-6 & SP-D is recommended in the management and treatment of patients with primary or secondary lung tumors.

A proper echocardiographic study requires several video clips recorded from different acquisition angles for observation of the complex cardiac anatomy. However, these video clips are not necessarily labeled in a database. Identification of the acquired view becomes the first step of analyzing an echocardiogram. Currently, there is no consensus whether the mislabeled samples can be used to create a feasible clinical prediction model of ejection fraction (EF). The aim of this study was to test two types of input methods for the classification of images, and to test the accuracy of the prediction model for EF in a learning database containing mislabeled images that were not checked by observers. We enrolled 340 patients with five standard views (long axis, short axis, 3-chamber view, 4-chamber view and 2-chamber view) and 10 images in a cycle, used for training a convolutional neural network to classify views (total 17,000 labeled images). All DICOM images were rigidly registered and rescaled into a reference image to fit the size of echocardiographic images. We employed 5-fold cross validation to examine model performance. We tested models trained by two types of data, averaged images and 10 selected images. Our best model (from 10 selected images) classified video views with 98.1% overall test accuracy in the independent cohort. In our view classification model, 1.9% of the images were mislabeled. To determine if this 98.1% accuracy was acceptable for creating the clinical prediction model using echocardiographic data, we tested the prediction model for EF using learning data with a 1.9% error rate. The accuracy of the prediction model for EF was warranted, even with training data containing 1.9% mislabeled images. The CNN algorithm can classify images into five standard views in a clinical setting. Our results suggest that this approach may provide a clinically feasible accuracy level of view classification for the analysis of echocardiographic data.

In cervical cancer treatment, radiation therapy is selected based on the degree of tumor progression, and radiation oncologists are required to delineate tumor contours. To reduce the burden on radiation oncologists, an automatic segmentation of the tumor contours would prove useful. To the best of our knowledge, automatic tumor contour segmentation has rarely been applied to cervical cancer treatment. In this study, diffusion-weighted images (DWI) of 98 patients with cervical cancer were acquired. We trained an automatic tumor contour segmentation model using 2D U-Net and 3D U-Net to investigate the possibility of applying such a model to clinical practice. A total of 98 cases were employed for the training, and they were then predicted by swapping the training and test images. To predict tumor contours, six prediction images were obtained after six training sessions for one case. The six images were then summed and binarized to output a final image through automatic contour segmentation. For the evaluation, the Dice similarity coefficient (DSC) and Hausdorff distance (HD) was applied to analyze the difference between tumor contour delineation by radiation oncologists and the output image. The DSC ranged from 0.13 to 0.93 (median 0.83, mean 0.77). The cases with DSC <0.65 included tumors with a maximum diameter < 40 mm and heterogeneous intracavitary concentration due to necrosis. The HD ranged from 2.7 to 9.6 mm (median 4.7 mm). Thus, the study confirmed that the tumor contours of cervical cancer can be automatically segmented with high accuracy.

Purpose. Recently, Elekta has supplied volumetric modulated arc therapy (VMAT) in which multi-leaf collimator (MLC) shape, jaw position, collimator angle, and gantry speed vary continuously during gantry rotation. A quality assurance procedure for VMAT delivery is described. Methods and materials. A single-arc VMAT plan with 73 control points (CPs) and 5-degree gantry angle spacing for a prostate cancer patient has been created by ERGO + + treatment planning system (TPS), where MLC shapes are given by anatomic relationship between a target and organs at risk and the monitor unit for each CP is optimized based on given dose prescriptions. Actual leaf and jaw positions, gantry angles and dose rates during prostate VMAT delivery were recorded in every 0.25 seconds, and the errors between planned and actual values were evaluated. The dose re-calculation using these recorded data has been performed and compared with the original TPS plan using the gamma index. Results. Typical peak errors of gantry angles, leaf positions, and jaw positions were 3 degrees, 0.6 mm, and 1 mm, respectively. The dose distribution obtained by the TPS plan and the recalculated one agreed well under 2%-2 mm gamma index criteria. Conclusions. Quality assurance for prostate VMAT delivery has been performed with a satisfied result.