生体医工学
Online ISSN : 1881-4379
Print ISSN : 1347-443X
ISSN-L : 1347-443X
最新号
選択された号の論文の6件中1~6を表示しています
研究
  • 鶴 智太, 徳井 亜加根, 青柳 有香, 井手 菜帆花, 大屋 祐輔, 守本 祐司
    2022 年 60 巻 1 号 p. 1-7
    発行日: 2022/03/10
    公開日: 2022/05/20
    ジャーナル フリー

    Among the three types of wearable sensors (acceleration, gyro, and magnetic) used to measure joint angles, it is not clear which sensor provides the most accurate measurement of thigh tilt angle during daily walking, which is important for assessing walking speed loss. The purpose of this study was to compare the errors of the three types of sensors due to physical motion. 12 healthy male subjects wearing the sensors and markers were subjected to perform 3 patterns of foot-stepping and 2 patterns of walking. For the foot-stepping experiment, we calculated the deviation (error) from the data obtained from each of the three types of sensors for each 10% of the walking cycle, assuming that the data values obtained from the video analysis were true values. For the walking experiment, we evaluated the homology between the data patterns obtained from each of the three types of sensors. The results showed that the error of the accelerometer was the largest in the case of fast foot-stepping. In the walking experiment, the data pattern obtained from the gyro sensor was not homologous to the data pattern obtained from the other two sensors. These results indicate that the magnetic sensor is the least sensitive to the effects of body motion.

  • 岸本 和昌, 竹村 匡正, 杉山 治, 小島 諒介, 八上 全弘, 南部 雅幸, 藤井 清孝, 黒田 知宏
    2022 年 60 巻 1 号 p. 8-15
    発行日: 2022/03/10
    公開日: 2022/05/20
    ジャーナル フリー

    Accidents may occur in hospitals when the medical staff fail to notice the alarm ringing at a distance or in a closed room. In many hospitals, patient monitoring devices are connected to the hospital information system through a network, but some medical devices cannot be connected because they do not produce any external output. If the staff can detect the alarm ringing in a hospital room from some distance, they can provide more efficient and proactive medical care. In this study, alarm sounds were collected using a monaural microphone, and a machine learning classifier was constructed using deep neural networks. The classifier was evaluated using a simulation dataset of polyphonic alarm sounds, superimposed with the environmental sounds of a hospital ward. Data were collected from four devices, and two training datasets were created with a logarithm spectrogram using Mel filter bank (MFB) and custom filter bank (CFB). In addition, two classifiers were developed for 16 classes based on a combination of the four devices. One classifier was trained on MFB and the other on CFB. The classifiers evaluated the simulation dataset with a signal-to-noise ratio (SNR) of 30, 20, 10, and 0 dB. The classifier trained on CFB had a micro F1 score of 72.7% and an area-under-the-ROCcurve of 0.963 at an SNR of 0 dB. This micro F1 score was 4.5 points higher than that of the score of the classifier trained on MFB. In addition, the misidentification rate of the environmental sounds (class without all devices) was 1.2%. Therefore, the classifier could not reliably distinguish between the alarm sound and environmental sounds, but the possibility as a notification system was presented.

  • 西澤 優希奈, 大北 陽一, 木村 亜紀子, 五味 文, 井村 誠孝
    2022 年 60 巻 1 号 p. 16-24
    発行日: 2022/03/10
    公開日: 2022/05/20
    ジャーナル フリー

    Cyclotropia is a condition in which one eyes is rotated about the anteroposterior axis. Cyclotropia affects the visual function;but, it is difficult to detect any external changes because of the rotational symmetry of the eyeballs in relation to the anteroposterior axis. Moreover, it takes time to receive an adequate medical diagnosis. To measure the degree of cyclotropia, a major amblyoscope that can display separate optotypes to the left and right eyes is widely used. However, a major amblyoscope is a highly specialized inspection equipment and is not suitable for daily measurements. In this study, we propose a method to quantify cyclodeviation by monocular vision using common devices such as smartphones and tablets that can be used for the routine testing of cyclotropia. In the proposed method, a tilted bar as an optotype is presented on the device to a monocularly blinded patient, and the patient is requested to make the tilted bar horizontal by rotating the device. The deviation between the optotype and the horizontal is measured when the horizontal is maintained based on the patient’s sense of equilibrium, and the cyclodeviation is measured as the sum of the deviations of both eyes. We implemented the proposed method as an application running on smartphones and tested the application on 47 patients with paralytic strabismus and 10 normal subjects. In patients with acquired paralytic strabismus, the correlation between the results of the proposed method and those measured with a major amblyoscope was 0.66, implying that the measurement results were good. The ROC analysis based on the results of all subjects for the classification of paralytic strabismus patients and normal subjects resulted in an AUC of 0.89, indicating that the implementation of the proposed method had a moderate predictive ability for cyclodeviation.

  • 三村 祐輝, 阿部 光, 齊藤 直, 杉本 俊之, 高橋 龍尚, 柳田 裕隆
    2022 年 60 巻 1 号 p. 25-31
    発行日: 2022/03/10
    公開日: 2022/05/20
    ジャーナル フリー

    In ultrasound systems, there is a trade-off between image quality and the number of transducers (elements) placed in the array probe, and to achieve a high resolution, the array probe must be created with additional elements. In 3D/4D echo, a significant number of elements are required compared to 2D echo, because imaging is performed using a two-dimensional array probe. As a result, 3D/4D echo inevitably has the problem of increasing equipment costs. To solve this problem, it is necessary to develop a method to reconstruct images with higher resolution using fewer elements than conventional imaging methods. The L1-norm minimization method has attracted much attention in recent years because of its ability to reconstruct high-resolution images even from incomplete data. In this study, we investigated whether high quality images can be reconstructed by minimizing the L1 norm for an object with an orientation, height, and depth of the imaging area (3D object). We also used image quality evaluation methods (MSE: Mean Squared Error, PSNR: Peak Signal to Noise Ratio, and SSIM: Structural SIMilarity) to quantitatively evaluate the obtained images. The results confirmed that L1-norm minimization is capable of reconstructing high-quality images with high evaluation values in all image quality evaluation methods compared to conventional methods.

  • 森 亮介, 簑 弘幸
    2022 年 60 巻 1 号 p. 32-37
    発行日: 2022/03/10
    公開日: 2022/05/20
    ジャーナル フリー

    θ-γ phase-amplitude coupling (PAC) is observed in neural networks of the hippocampus, in which the amplitude of high-frequency oscillations, γ wave, is modulated by the phase of low-frequency oscillations—θ wave. Several studies have shown that the degree of PAC is closely related to memory formation. In contrast, the stochastic resonance (SR) phenomenon, in which the addition of random synaptic noise can enhance the transmission of sub-threshold synaptic stimuli, is also observed in the hippocampus. Although PAC reportedly appears in a specific status of brain activity, it is still unclear if and how random synaptic noise could contribute to generating θ-γ PAC. In this study, we investigated whether or not θ-γ PAC could be generated by an addition of random synaptic noise in a hippocampal neural network model with sub-threshold excitable synaptic stimuli using the modulation index (MI) through computer simulations. The results of computer simulations showed that the typical resonance curve of resonance phenomena, i.e., MI was maximized at a specific amplitude of synaptic noise. Therefore, it is concluded that θ-γ PAC can be generated by the vibrational mode transition, induced by the random synaptic noise, in the hippocampal neural network model with mixed sub-and supra-threshold stimuli.

  • 外川 龍之介, 中尾 光之, 虫明 元, 片山 統裕
    2022 年 60 巻 1 号 p. 38-45
    発行日: 2022/03/10
    公開日: 2022/05/20
    ジャーナル フリー

    The primary visual cortex has a topographic projection from the retina, called the cortical retinotopic map. Since retinotopic maps help evaluate visual responses and estimate retinal defects, they may play an essential role in research on visual reconstruction using animals. So far, several methods have been proposed to estimate the retinotopic map. Since the methods use the synchronous averaging method, it should be assumed that the eye position does not change during measurement. To satisfy this condition, fixation of the eyeball and anesthesia are necessary; however, these are stressful to the animal. Hence, measurements in unanesthetized and rest states are desirable since they are less stressful. However, eye movement occurs in these states. Therefore, we have developed a model-based retinotopic map estimation method without using the synchronous averaging process. This method is robust against changes in eye position during measurement. In the method, we approximated the retina and the shape of the receptive field of the cortex to be plane and Gauss functions, respectively. However, there has been a concern that the error increases in the peripheral part of the retina. This study attempted to overcome this problem by modeling the retina as a sphere and replacing the receptive field function with the Von Mises-Fisher function on the sphere. We compared the horizontal retinotopic maps of unanesthetized and rest mice estimated by Kalatsky’s method and the proposed method. The single-eye visual field of the azimuth retinotopic map estimated by the proposed method was approximately±90°, which was almost the same as that observed in mice. On the other hand, the visual field estimated by Kalatsky’s method was approximately±40°, which was narrower than these. Next, we compared the maps by the proposed method using planar and spherical retinal models. We found that a field of view of about ±60° was estimated by the planar model, narrower than the spherical model. These results suggest that the proposed method with the spherical retinal model is excellent for estimating the retinotopic maps of mice with eye movement.

feedback
Top