TSTaichiro Shiodera

Taichiro Shiodera

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Taichiro Shiodera has a total of 18 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Deep Shrinkage Convolutional Neural Network for Adaptive Noise Reduction

    by Kenzo Isogawa, Takashi Ida, Taichiro Shiodera, Tomoyuki Takeguchi

    The noise level of an image depends on settings of an imaging device. The settings can be used to select appropriate parameters for denoising methods. But denoising methods based on deep convolutional neural networks (deep-CNNs) do not have such adjustable parameters. Therefore, a deep-CNN whose training data contain limited levels of noise does not effectively restore images whose noise level is different from the training data. If the range of noise levels of training data is extended to solve the problem, the maximum performance of a produced deep-CNN is limited. To solve the trade-off, we propose a deep-CNN that is adjustable to the noise level of the input image immediately. We use soft shrinkage for activation functions of our deep-CNN. The soft shrinkage has thresholds proportional to the noise level given by the user. We also propose an optimization method for proportionality coefficients for the thresholds of soft shrinkage. Our method optimizes the coefficients for various noise levels simultaneously. In our experiment using a test set whose noise level is from 5 to 50, the proposed method showed higher PSNR than that in the case of the conventional method using only one deep-CNN, and PSNR comparable to that in the case of the conventional method using multiple noise-level-specific CNNs.


  • Open Access

    Automatic slice-alignment method in cardiac magnetic resonance imaging for evaluation of the right ventricle in patients with pulmonary hypertension

    by Kenichi Yokoyama, Shuhei Nitta, Shigehide Kuhara, Rieko Ishimura, Toshiya Kariyasu, Masamichi Imai, Toshiaki Nitatori, Tomoyuki Takeguchi, Taichiro Shiodera

    We propose a new automatic slice-alignment method, which enables right ventricular scan planning in addition to the left ventricular scan planning developed in our previous work, to simplify right ventricular cardiac scan planning and assess its accuracy and the clinical acceptability of the acquired imaging planes in the evaluation of patients with pulmonary hypertension. Steady-state free precession (SSFP) sequences covering the whole heart in the end-diastolic phase with ECG gating were used to acquire 2D axial multislice images. To realize right ventricular scan planning, two morphological feature points are added to be detected and a total of eight morphological features of the heart were extracted from these series of images, and six left ventricular planes and four right ventricular planes were calculated simultaneously based on the extracted features. The subjects were 33 patients (25 with chronic thromboembolic pulmonary hypertension and 8 with idiopathic pulmonary arterial hypertension). The four right ventricular reference planes including right ventricular short-axis, 4-chamber, 2-chamber, and 3-chamber images were evaluated. The acceptability of the acquired imaging planes was visually evaluated using a 4-point scale, and the angular differences between the results obtained by this method and by conventional manual annotation were measured for each view. The average visual scores were 3.9±0.4 for short-axis images, 3.8±0.4 for 4-chamber images, 3.8±0.4 for 2-chamber images, and 3.5±0.6 for 3-chamber images. The average angular differences were 8.7±5.3, 8.3±4.9, 8.1±4.8, and 7.9±5.3 degrees, respectively. The processing time was less than 2.5 seconds in all subjects. The proposed method, which enables right ventricular scan planning in addition to the left ventricular scan planning developed in our previous work, can provide clinically acceptable planes in a short time and is useful because special proficiency in performing cardiac MR for patients with right ventricles of various sizes and shapes is not required.


  • Automated flow quantification for spin labeling MR imaging

    by Taichiro Shiodera, Shuhei Nitta, Tomoyuki Takeguchi, Masao Yui, Yuichi Yamashita, Takao Yamamoto, Shinya Yamada

    Objective The Time-Spatial Labeling Inversion Pulse (Time-SLIP) technique enables tracing of regional fluid flows without the use of contrast medium. The objective of this study is to quantify automatically slow and complex fluid flows using the Time-SLIP technique. Materials and methods Series images were acquired with a 1.5-T MRI scanner using the Time-SLIP technique with half-Fourier fast spin-echo (FSE) and balanced steady-state free precession (bSSFP) sequences. In this method, labeled fluid regions in images were automatically detected based on image processing techniques for a given point. The flow velocity of the labeled fluid region was calculated using regression fitting for the region’s position. To evaluate our method, constant and non-constant laminar flows in a water phantom were studied. In addition, volunteer experiments were conducted to quantify the flow of cerebrospinal fluid. Results In the constant flow experiments the correlation factor r 2 between the flow velocity calculated from our method and the laminar peak velocity calculated from the volumetric flow rate was 0.9992 for the FSE sequence and 0.9982 for the bSSFP sequence. In the non-constant flow study, the flow velocity was calculated accurately for any period inversion time even when the flow velocity was changed, and the quantification error was negligible. In the volunteer experiments, r 2 between the flow velocity calculated by the proposed method and that obtained by manual annotation was 0.9383. Conclusion The experimental results showed that our proposed method can quickly and accurately provide information on flow velocities especially for slower and complex flows. Our method is, therefore, expected to be useful in diagnostic support systems.


  • Block Based Extra/Inter-Polating Prediction for Intra Coding

    by Taichiro Shiodera, Akiyuki Tanizawa, Takeshi Chujoh

    In this paper, we propose an efficient intra coding method. The method involves adding an extra/inter-polating prediction method to the conventional H.264/MPEG-4 AVC (H.264) intra prediction method. The proposal consists of two parts. One is to change the sub-block coding order in macroblock (MB). The bottom-right sub-block in MB is predicted using the reference pels on the upper and left side of the sub-block firstly. After that, the other sub-blocks are predicted using not only the reference pels on the upper and/or left side but also ones on the bottom and/or right side of each sub-block. The other part is to use an intra bidirectional prediction method which can use the interpolation of the prediction from upper/left pels and the prediction from bottom/right pels. Experimental results show that our method improves bit reduction by up to 7.7% at the same PSNR compared to H.264.