Skin has various physiological functions and three-dimensional detection of microcirculation within the skin is essential to evaluate and diagnose the skin physiological functions. In this study, a three-dimensional imaging technique of the skin physiological function using optical coherence tomography-based angiography (OCTA) is proposed. OCTA is a method to three-dimensionally visualize vascular networks within the tissue by detecting instantaneous variances in OCT signals that correlate with blood flow velocity. The proposed method gives a stimulus to the skin, continuously detects responses to the stimulus in vascular networks and blood flow velocities with OCTA, and evaluates the skin physiological functions. In this presentation, to test the feasibility I made a validation experiment where an alcohol stimulus was given to the skin and responses to the alcohol stimulus were detected. The utility of the method was investigated by making a comparison between the experimental results and alcohol tolerance of the subjects.

Skin wrinkles are associated with changes in the rheological properties of skin tissue. In this paper, we use Optical Coherence Tomography (OCT) to visualize skin tissue at micr-scale, and Optical Coherence Straingraphy (OCSA) to visualize skin tissue rheology. In order to investigate the validity of this method, suction tests were conducted on the forearm. The OCSA method was applied to the creep recovery time after the suction was released. The epidermis was more elastic than the dermis layer with a shorter creep recovery time by time series of strain velocity distribution. Therefor rheological properties of the skin can be visualized on a micro scale and that skin mechanics can be evaluated.

Hydraulic permeability κ estimation method of human subcutaneous adipose tissue (SAT) has been proposed by integrating poroelastic-transport model (pe-TM) to wearable electrical impedance tomography (w-EIT). In-vivo experiments were conducted by applying external compressive pressure -P on human calf boundary to induce interstitial fluid flow and ion movement in SAT. pe-TM predicted the ion concentration distribution c_mod by coupling poroelastic and transport model to describe the hydrodynamics and transport phenomena inside SAT. w-EIT measured the time-difference conductivity distribution δγ in SAT resulted from the ion movement. κ was estimated by applying an iterative curve-fitting between c_mod predicted from pe-MTM and experimental ion concentration distribution c_exp derived from w-EIT. As a result, κ of SAT was range from 1.45-1.63 x 10^-11 m/Pa.s, which were correspondent to the other soft tissues κ (such as mice subcutaneous, rabbit aorta, etc.) from literature which were in range of 1-10 x 10^-11 m/Pa.s

We have been developing heart valve-like tissues (Biovalve) using in-body tissue architecture. Previously, we fabricated a novel prosthetic valve with a membranous tissue (Biosheet) as a valve cusp and an elliptical valve annulus. Because of isotropic mechanical properties of the valve, it was difficult to satisfy the performance of both opening and closing. The purpose of this study is to clarify the influence of mechanical property of anisotropic cusp on valve performance.The valve models were fabricated from a combination of polyurethane sheets and Ni-Ti wires. In-Vitro simulation was performed to evaluate valve performance and to observe its opening-closing behavior. The experimental conditions were set at the physiological condition of healthy adult male. According to ISO5840, temporal waveform of left atrial pressure, left ventricular pressure and mitral regurgitation were measured, and regurgitation rate, mean pressure gradient and effective orifice area were calculated to evaluate the valve performance.

Force sensors such as load cells are used to measure the force applied to the skin when an object comes into contact with it. However, it is very difficult to measure the distributed stress inside the skin with these sensors. To solve this problem, we propose a method to measure the stress inside the skin by using the color change when pressure is applied to the skin. In order to investigate the effectiveness of this method, we measured the color change of the skin and the distribution of the stress in the skin using the finite element method when the cylindrical stimulus was pressed. This method is expected to make it possible to measure body pressure without contact and to include internal stress.

Fluctuations with properties called long time correlation and fractal nature have been observed in many biological systems.Recently, we proposed a model for fluctuations of two-dimensional trajectories that is described as a composite of fluctuations that have different fractal properties in each of the two different directions.Besides, we introduced an analysis method called oriented fractal scaling component analysis (OFSCA).OFSCA can extract the original two independent components.In OFSCA, a 2D trajectory is projected in a specific direction, and the angle is estimated from the relationship between the direction and the estimated scaling exponent.OFSCA can be applied to fluctuations that follow a normal distribution.In this study, we measured the center of foot pressure sway with eyes open and analyzed with OFSCA.The subjects were three healthy males. From the results, non-orthogonal components with different scaling properties were confirmed and were suggested that such non-orthogonality was related to postural control centered on the dominant foot.

Static posturography tests are conducted to assess the balance ability of patients. Existing descriptive statistics yield higher index values for older people than healthy younger ones. It is important to reveal the mechanism of the postural control system and utilize the system parameters for balance ability assessment. For this purpose, a mathematical modeling of the center of pressure was conducted. We introduce linear stochastic differential equations to reproduce the center of pressure trajectories and verified them in terms of the descriptive statistics reproducibility. The model equation comprises viscosity, stiffness, and stochastic terms to express the center of pressure movement during the posturography test. The experimental results show that the mathematical model has high descriptive statistics reproducibility, and the balance ability can be assessed using a small number of model parameters with a clear physical meaning.

Bruxism is a symptom in which one grinds, gnashes, or clenches one's teeth unconsciously. It may cause the teeth to be chipped, which can result in temporomandibular joint disorder. Pseudo clenching is masseter muscles' excessive tension without tooth contact, which is difficult to distinguish from bruxism. This study was performed to detect bruxism and pseudo clenching automatically by machine learning. Single-stream or multi-stream Hidden Markov Models (HMM) for each of the three classes (bruxism, pseudo clenching, and others) were trained with Mel Frequency Cepstral Coefficients (MFCC) calculated from EMG and acoustic signals from 12 healthy adults. The averaged F measure, calculated with cross-validation, was used to evaluate the model. The F measure with EMG only was 77.2±8.1%, that with acoustics only was 60.1±16.0% and that with both features was 81.7±8.5%.These results indicated that a rather good performance of bruxism detection was realized, but further improvements are needed.

Surface electromyogram (EMG) has been utilized as control signals for robotic prosthetic hands and rehabilitation devices because it reflects human motion intentions. The variance of the EMG signal has been known to be accompanied by variability due to the uncertainty of muscle activity. The authors have previously formulated a scale mixture model that can consider the uncertainty in the variance of EMG signals. This paper proposes an EMG pattern classification method incorporating a Bayesian-extended scale mixture model, thereby allowing the accurate motion classification with considering uncertainty in the EMG variance. The parameters of the proposed method can be learned via variational Bayesian inference. The EMG analysis experiments for three subjects showed that the proposed method can classify subject's motions with high accuracy of > 90%. The results of classification experiments on several public EMG datasets demonstrated that the proposed method outperforms the general classifiers in all conditions.

The human-like appearance and movements of a myoelectric prosthetic hand are important for upper limb amputees, and many studies have been conducted so far. However, there is no attempt to introduce the involuntary physiological phenomenon, called tremor, into the control of a prosthetic hand. To realize more human-like movements of a prosthetic hand, this study proposes an artificial tremor generation method based on the variance distribution of electromyogram (EMG) signals and introduces it to prosthetic hand control. The proposed artificial tremor generation method can reproduce the action tremor component generated during voluntary movements and the physiological tremor component derived from involuntary mechanical reflexes, respectively. In the experiment, we calculated the power spectrum from the joint angle of the prosthetic hand and compared it with the actual tremor measured from a human body. The results showed that the generated artificial tremor has characteristics similar to the actual human tremor.

Pulse wave velocity (PWV) of artery contains important information such as blood pressure and arterial stiffness. The aim of this study is to estimate PWV of arteries from the difference of pulse arrival time with plethysmogram at peripheral small region. However, it is difficult to calculate PWV by the ratio of distance and difference of pulse arrival time between two close points. Considering the stiffness parameter beta, the square of PWV at periphery is proportional to the peripheral blood pressure. In proposed method, PWV was estimated by correction of the difference of pulse arrival time at hand region using the characteristics between square of PWV and hand elevation. The results of verification experiments showed the estimated PWV had a close correlation to arterial PWV. Therefore it was suggested that PWV may be able to be estimated from the plethysmogram of the hand region.

Magnetocardiography (MCG) is a non-invasive method of measuring the magnetic field generated by the electrical heart activity. The source estimation by MCG is useful technics for clinical applications. Several spatial filter methods (minimum norm, beamformer, LORETA, eLORETA, etc) have been proposed to estimate the source of the brain activity. However, when these methods are applied to source estimation of the heart activity, deep sources tend to more expand the estimated solutions than shallow sources, even for the same intensity. In this study, we aimed to suppress the expanding of the estimated solution due to the depth of the source by using multi-sensor plane method. We performed the experiment with eLORETA to evaluate the proposed method and to compared two methods of making spatial filters. As a result, the estimation accuracy was different for each spatial filter, and the expansion of the deep source could be suppressed by the proposed method.

Magnetocardiograms (MCGs) have become increasingly relevant for clinical research, due to its potential to detect early stages of heart disease. However, if the current spatial filter method is applied to the MCG, the estimated solution of the signal source at the deep position tends to be wider than it at the shallow position. One of the reasons is that the information of the sensors located at the ends is lost because the sensor plane is set bigger than the analysis area usually. Therefore, in this study, we proposed the setting that the analysis area larger than the sensor plane and performed the simulation to confirm the amount of information of each sensor can be used equally. As a result, the position and size of the estimated solutions were similar to the area shown in CT images. And GOF (Goodness of fit) was about 0.996.

Heart rate variability (HRV) shows a long-range correlation meaning a slow decay of the autocorrelation. The long-range correlation can be characterized by the scaling exponent, which is estimated from logarithmic plots of the power spectrum or the fractal analysis. In the HRV scaling behavior, a crossover phenomenon has also been reported. However, the crossover phenomenon has not been fully understood. To analyze such crossover phenomena, we develop a methodology based on the detrending moving-average analysis (DMA), which removes non-stationary trend using a high-order Savitzky-Golay filter. Because most biological signals contain non-stationary trend, the trend removal procedure is essential for the accurate scaling estimation. However, the trend removal procedure in high-order DMA induces time-scale distortions. In this study, we introduced a new method to correct such time-scale distortions, and analyzed the HRV crossover phenomenon. In our presentation, the physiological significance of the crossover phenomenon in HRV will be discussed.

The purpose of this study was to analyze heart rate variability on functional food using Empirical Mode Decomposition (EMD), which is used for nonlinear analysis, and to compare EMD results with those obtained by Fast Fourier Transform (FFT). Two experimental sessions of Gamma-Aminobutyric Acid (GABA) or placebo intake were conducted in a healthy man. Electrocardiogram (ECG) was measured before and after the intake of GABA or placebo. After obtaining the time series of RR intervals from ECG, both analyses were performed to calculate each index of autonomic nerve activity. The results showed that the LF/HF ratio was not a difference in both intake by FFT analysis, but by EMD analysis, which was a small tendency for LF/HF ratio in intake of GABA compared with placebo. The EMD analysis suggested that it could obtain sufficient changes of autonomic nerve activity than the FFT analysis on functional food.

It is worth to monitor sleep quality on a daily basis and to solve sleep disorder problems, because sleep is essential for the maintenance of vital activities. It has been reported that the appearance rate of each sleep stage changes depending on the progress of sleep. However, it has not been fully understood how heart rate variability (HRV) changes depending on the progress of sleep. The purpose of this study is to clarify the relationship between the appearance rate of sleep stages and HRV during sleep. In this study, we compared temporal profiles of HRV indices under two conditions, sleep at home and sleep on a semi-reclined car seat. Our result showed that the changes of power at low frequency (HF.nu) of HRV differ in the two sleep environments. It suggested that the temporal profile analysis of HRV is useful for the sleep quality evaluation.

Assessment of mortality risk of patients with congestive heart failure (CHF) using heart rate variability (HRV) is clinically important. Previous studies proposed lambda25sec (non-Gaussian distribution index) and SDAMPLF (variance of the instantaneous amplitude of low frequency (LF)-band HRV) were effective in predicting the mortality risk in CHF. In this study, we propose the information entropy of the instantaneous amplitude of the LF-band HRV (IEAMPLF). We analyzed the 24-hour HRV data from 108 CHF patients. The accuracy of the logistic regression model which predicts the mortality events was same when using IEAMPLF instead of lambda25sec as one of the explanatory variables, and it was higher than the case using SDAMPLF. This implies that the prediction accuracy can be maintained by using IEAMPLF without the assumption of multiplicative log-normal process.

Background: Moxibustion is an alternative medicine performed by burning moxa at a specific part of the body. However, there has been no quantitative analysis of whether moxibustion induces physiological responses in previous studies. In this study, we investigated the thermal effects of moxibustion on cardiovascular responses. Methods: Twenty healthy volunteers participated in this study. Moxibustion treatment was applied to the lower leg (Zusanli acupoint). Heart rate(HR), blood pressure(BP) and skin temperature(ST) were measured continuously for 2 min at rest and 6 min at moxibustion. Results: HR significantly decreased (64.3±7.5 to 62.3±1.3bpm, p=0.005) when ST reached a maximum (45.0±10.1°C) by moxibustion. There was no significant change in BP, and the bradycardic effect was also observed when ST was continuously maintained with a 38°C heat stimulator at specific sites in the body. Conclusion: Regional heat stimulation by moxibustion provides fundamental evidence for effective bradycardia response.

Biometric authentication such as face recognition and fingerprint recognition have been pervasive. However, there are many attacks against the biometric authentication. One of them is a "presentation attack" using artificial biometrics to break the authentication. Meanwhile, several approaches using photoplethysmogram (PPG) which can be recorded optically for authentication have been proposed. PPG-based authentication may be available in the near future, because PPG have fewer restrictions in measurement sites and postures than other biometrics, and smartwatches with PPG recording function have been pervasive. Therefore, the prediction of attacks against PPG-based authentication and countermeasures are required. We proposed a presentation attack using the advantage of PPG in performing measurements on various sites. The attack records PPG stealthily and utilizes it for identity spoofing. We investigated the feasibility of the attack and showed that the attack may occur. We investigate countermeasures using the unique information of the measurement sites against the attack.

In this paper, the prediction and infection prevention system for the Cogh in SARS-COV-2 has been invented. By the use of the visual images of the surface of the human necks, the motion of the Cough was realized and prediction of the Cough was embodied. Active partition will prevent infection.

Due to the pandemic of COVID-19, infrared thermography (IRT) was applied for screening of potential infection in many scenes. Measuring facial skin temperature alone using IRT is easily affected by ambient temperature; such as fever-based screening is less sensitive in detecting infection. Therefore, our previous work proposed an infectious disease screening system based on multiple vital signs using CCD camera and IRT (T.Negisi & G.Sun, Sensors, 2020). In this paper, automatically track the nostrils using thermography alone was newly developed and implemented into the system based on Histograms of Oriented Gradients (HOG) and Support Vector Machine (SVM) method, thereby enable robust measurement of respiratory rate. The Bland Altman Plot exhibited close agreement for non-contact IRT and contact-type belt in the measurement of respiration rate with upper and lower 95% limits of agreement (without motion: -2.7bpm and 2.7bpm, with motion: -1.8bpm and 3.0bpm, respectively).

An image processing method for detecting gas-trapping by small airway obstruction (termed functional small airway disease; fSAD) has been proposed by non-rigid image registration technique between inspiratory and expiratory 3D-CT data set(Nat. Med, 2012). The fSAD is defined as a voxel with HU values of -950 or more on inspiratory CT and less than -856 on expiratory CT. Although this definition presumes that expiratory voxels are accurately registered to corresponding inspiratory voxels, there have been no statements regarding registration accuracy in any published papers. We generated virtual inspiratory and expiratory 3D-CT data sets in computer, where emphysema voxels were randomly distributed in the lung parenchyma with fractal size distribution, and found that even a tiny mis-registration generated plenty of false fSAD voxels. We concluded that the proposed method for fSAD was not valid.

Automatic and long-term monitoring of respiratory is in great demand for lung diseases. It gets required greater in these years due to COVID-19 pandemic to reduce medical staff fatigue for checking patient conditions frequently for long time. Kobayashi et al., in our team, developed a device measuring respiratory condition by quantizing the displacement between the 6th and 8th ribs. We introduce long short-term memory (LSTM) neural network to classify patient respiratory signals into the two states of normal and low-functional respirations. The signals were checked by a medical doctor manually for classified into the two states. In the process, they were transformed to frequency-domain spectra with complex-valued wavelet transform, and then quantized the respiratory wavelet spectra due to the large number of spectra patterns. After that, the LSTM learned and classified the processed respiratory signals. The experimental results showed the feasibility to detect the two states.

Previous studies have shown that time-dependent cardiorespiratory responses before and after the onset of dynamic exercise at various intensities depend on the relationship between exercise training and its exercise intensity. However, it is not clear when and to what extent such differences in training intensity for the development of short- and long-term physiological adaptations affect cardiorespiratory function during exercise. The subjects were 16 college student athletes who were divided into 95% and 80% intensity training groups. After HIT, maximal oxygen uptake increased significantly in both groups. After HIT, maximal heart rate during step exercise increased significantly only in the 95% training group after HIT after 6 weeks, and an increase in heart rate immediately after the start of exercise was observed after HIT. We concluded that HIT intensity-dependent adaptive changes in cardiorespiratory function during exercise appear mainly in the cardiovascular system after 6 weeks.

Flap ischemia and consecutive flap loss is an innate complication in reconstructive free flap surgery. With the development of machine learning, time series analysis based flap failure prediction has become possible. With the laser doppler flowmetry pocket device (PocketLDF) by JMS it has become possible to measure skin perfusion every second. In this trial skin perfusion in addition to blood pressure, pulse and respiratory rate were measured in 3 patients and flap failure prediction based on the Auto-Regressive Moving Average (ARMA) model and the Long Short Term Memory (LSTM) network was conducted.Accurate perfusion prediction with the ARMA model for stationary processes and with the LSTM network for non-stationary processes was possible. In comparison with real time observation by the attending doctor, the ARMA model was able to predict flap ischemia ahead of time.

In Parkinson's disease (PD), the relationship between cortical thinning and various physical and mental symptoms is not fully understood. Here, we attempted to predict PD symptoms from cortical thinning patterns in PD patients. We evaluated the motor and non-motor symptoms of 181 PD patients treated at Kyoto University Hospital using neurological tests, neuropsychological tests, and questionnaires. In addition, head MRI was also recorded, and T1-weighted images (MPRAGE) were obtained. First, we drew a dendrogram based on the Spearman correlations, which evaluates the similarity of individual differences in behavioral performance among tasks, and clinical tasks close to known domains were naturally classified. Furthermore, cortical thickness was determined for T1-weighted images using FreeSurfer (ver. 6) by dividing cortex into 180 unilateral regions (360 bilateral regions) based on the HCP-MMP1 atlas. We predicted the clinical-task performances based on combinations of cortical thickness in all cortical regions using a LightGBM algorithm.

Magnetoencephalography (MEG) is an essential tool for non-invasive focus localization of epilepsy patients. However, little is known about the clinical value of MEG sharp transients (STs) when epileptic spikes are not found. At our hospital, 53 patients underwent comprehensive presurgical evaluations including head MRI, video EEG monitoring and MEG within the last 5 years. 15 out of 53 patients had only MEG STs. ECDs of STs were clustered in 6 patients. In 3 of them, clusters overlapped with structural lesions. Some dipoles overlapped with structural lesion in 1 patient without ECD cluster. 4 patients with clustered ECD and 5 patients without ECD cluster underwent surgical resection of structural lesions and all became seizure free. MEG sharp transients are less specific and difficult to localize compared with spikes, but good prognosis can be achieved by integrating with specific findings from other modalities.

In Japan of a super-aging society, the number of patients suffering from fragility fracture of the pelvis (FFP) with osteoporosis is increasing. It can make patients bedridden and the complication. Physicians diagnose the fracture in 3D CT images, which is hard and time-consuming to find FFP. This paper proposes a novel method of boring survey based fracture detection (BSFD), to automatically detect FFP in 3D CT images. Firstly, the bone surface of the pelvis is extracted from CT images. Then, it bores the quadratic prism for the internal bone area with CT values. The 3D convolutional neural network can predict a probability of fracture, and apply to the whole pelvis. The method was evaluated by using 110 elderly subjects with pelvic fractures. The AUC was 0.84 for training subjects and 0.77 for evaluation subjects.In addition, it is useful for physicians to display the 3D distribution of fracture possibilities.

The purpose of this study was to evaluate metatarsal torsion in 3D in feet with and without hallux valgus. We captured computerized tomography images of 10 normal feet and 10 hallux valgus feet. A foot of a random participant in the normal feet was selected as the reference foot. After reconstruction of 3D images, the areas 35% proximal and 35% distal to each metatarsal in each participant were superimposed on the corresponding areas of the reference foot. The torsion angle of the first metatarsal was defined as the angle of rotation between the proximal and distal axes of the metatarsal of the evaluated foot relative to the reference foot. Internal torsion was significantly more pronounced in the hallux valgus group (p < 0.01). There were no significant differences in second, third fourth and fifth metatarsal. We suggest that sufficient correction to restore supination is considered necessary when performing surgery.

Post-mortem imaging (Autopsy imaging: Ai) has been performed at many institutions in Japan, because we have the highest number of computed tomography (CT) scanners per population in Japan. Currently, there are many reports on organ segmentation of clinical CT images, but there are still few reports on segmentation of Ai. The purpose of this study is to confirm whether the method of lung segmentation in clinical CT is acceptable to Ai or not, and to segment the lung region in Ai precisely. We applied 3D U-Net to segment the lung region in CT images. Two training datasets were prepared: [A] clinical CT images (55 cases), [B] Clinical CT images (41 cases) and Post-mortem CT images (14 cases). As a result, the model trained by [A] indicated poor segmentation in a case with marked post-mortem changes. The model trained by [B] could segment the lung region more accurately.

The overlooking bladder tumors during transurethral resection of bladder tumor (TURBT) causes high intravesical recurrence rate. The conventional White light imaging(WLI) and Narrow-band imaging (NBI) were subjective and poor reproducibility due to depending on doctor's experience and skills. Recently, image recognition using artificial intelligence (AI) has been applied for diagnostic imaging and achieving a good performance result, however the feasibility of cystoscopy with AI is still unknown. In this study, we constructed an object detection system based on the AI architecture. We prospectively obtained cystoscopic images from patients who underwent TURBT, and divided these images as the training data and test data. The sensitivity, specificity and Positive predictive value of WLI and NBI were 84.6%, 76.9% and 90.8%, and 78.1%, 97.1% and 95.8%, respectively. Our system has the possibility to improve detection rates of the bladder tumors during cystoscopy, and might be beneficial to reduce recurrence rate of bladder tumors.

The general movements (GMs) assessment is used in the early diagnosis of neonatal disorders. Although this method is highly reliable, the result may vary because it relies on visual examination. In this paper, we propose a video-based classification method of GMs using a deep neural network. In the proposed method, spatial and motion features of infants' movements are extracted from video images using a two-stream convolutional neural network. The extracted features are then concatenated into a single feature vector and input to a recurrent neural network, thereby allowing the classification of the type of GMs. The experiment was conducted for 100 infants (normal GMs: 35 writhing movements, 38 fidgety movements; abnormal GMs: 27 poor repertoire of GMs). The results demonstrated that the proposed method outperformed the conventional GMs evaluation system relying on domain-dependent knowledge. Therefore, the evaluation of infants' spatial and temporal features may be effective for automatic GMs classification.

Although deep learning has made remarkable achievements in medical image diagnosis, with accuracy exceeding that of medical specialists in some fields, those medical images are all formatted with specific imaging equipment and methods. In the field of oral surgery, the most popular images are with various single-lens reflex cameras no-standardized in terms of shooting range, angle, conditions, light intensity, number of pixels, etc. Since they are much noisier than the well-formatted medical images, it is more difficult to improve the diagnostic accuracy. In this research, we verified the influence of the shooting equipment, and examined the difference in accuracy when we used 1,000 images with a single-lens reflex camera 1,000 fluorescence images taken with a fluorescence observation equipment, respectively, and trained them with a model using CNN. We used 1,000 images taken with a single-lens reflex camera and 1,000 images with fluorescence observation equipment.

This study discusses mandibular-shape features specified on two-dimensional (2D) cephalograms and three-dimensional (3D) CT volumes. The feature points were specified on cephalograms with a house-developed software. Then, the points on frontal and lateral cephalograms were projected into the 3D space, which we had calibrated the imaging coordinate systems, and given 3D geometries. The calibration corrected geometrical incorrespondence between ear pads on the frontal and lateral images. The 3D point reconstruction was done considering projection scaling. In addition, 3D feature points were specified on the CT volumes. Then, we compared these three-type feature point sets for each mandibular landmark and tried to give them some clinical meanings. Comparison among the three point sets resulted that the 3D reconstruction with projection worked well and correctly provided the 3D geometries of feature points. As well, the difference between the 2D and 3D points showed some trends relative to the curvature of bone surfaces.

Cardiac rehabilitation has been reported to improve not only short-term prognosis but also long-term prognosis for the patients with cardiovascular diseases including heart failure. Its importance is recognized in daily clinical practice. When rehabilitation is performed in the acute phase, changes of vital signs including arrhythmia and reduced oxygen saturation often occur. However, it is difficult for medical professionals to grasp vital signs change simultaneously while observing the walking situation of the patient. In this study, we monitor the patient's biological information (oxygen saturation, pulse rate and ECG), storing it on a server over an internet connection, and developed a system that enables medical professionals to easily grasp vital signs change using a smart glass connected to the internet. By this system, it becomes possible to provide a safer cardiac rehabilitation environment.

Transesophageal Doppler (TED) velocity in the descending thoracic aorta (DA) is used to track changes in cardiac output (CO). However, CO tracking by this method is hampered by substantial change in aortic cross-sectional area (CSA) or proportionality between blood flow to the upper and lower body. To overcome this, we have developed a new method of TED CO monitoring. Our method estimates CO (CO_est) from DA velocity signal first by compensating changes in the aortic CSA, and by compensating changes in the blood flow proportionality through a machine learning of the relation between the CSA-adjusted CO and a reference CO (CO_ref). In 8 anesthetized dogs, we compared CO_est with CO_ref under diverse hemodynamic conditions. Between CO_est and CO_ref, concordance rate in four-quadrant plot analysis was 93%, while angular concordance rate in polar plot analysis was 94%. In conclusion, our method allows reliable TED CO monitoring by utilizing machine learning approach.

Collecting adequate cases of minor demential disease such as DLB or FTLD is problematic to CAD for dementia. We have investigated the applicability of CycleGAN for PET images (Kimura, ANM, 2020). We will evaluate the performance quantitatively. Because CycleGAN can generate real images, we set ROIs onto same regions both in real and synthesized images, and we look into the consistency of the voxel values. Moreover, we compare performance of a logistic regression to distinguish healthy controls (HC) and Alzheimer's disease (AD) where a part of training data is a synthesized by CycleGAN. As a result, the synthesized voxel values coincide with the real images (p<0.05). If only 3 cases utilized becomes worse to 76%, but it can be recovered to 91% with the synthesized images that is similar when only real cases are applied. We conclude that CycleGAN is applicable to generate a training data to CAD for AI.

53 mood disorder patients (mean 39 years) with 8 or more points on the Hamilton Depression Rating Scale and 58 healthy subjects (mean 36 years) were measured RR intervals and accelerations with wearable heart rate sensors (myBeat WHS-1, Union Tool Co., Ltd.) from 15:00 to 3 consecutive days. The average of heart rate variability and activity magnitude during waking and sleeping time period estimated from the accelerations were calculated, and those of patients were compared with controls by Mann Whitney U test. Compared with the controls, the patients had shorter RR intervals during both time periods. In the sleep time, the patients had less HF as index of the parasympathetic nervous system than controls. And the patients activity was less than the controls during the waking time periods. It was suggested that heart rate variability and activity magnitude during sleep and waking time periods could be objective indicators of depression.

In recent years, endoscopic surgery has been performed on various areas in the human body and there is a greater need for minimally invasive surgery which causes little burden on the body. The purpose of this research is to develop non-planar micro components with internal 3D wiring and electronic parts suitable for mounting to endoscopes for enabling higher performance and multifunctionality. For the fabrication of non-planar components, the photolithography technology was applied, and a lamination of permanent photoresist called SU-8 film (Nippon Kayaku Co., Ltd.) was used. In this research, non-planar micro components which can emit light from the side were fabricated and evaluated. It is expected that the burden on patients and doctors can be reduced by the widespread use of endoscopes and catheters equipped with various non-planar microscopic components.

Prostate cancer is the second leading cause of cancer death in men. At present, the methods for classifying early cancer on MRI images are mainly focused on single image modality and with low robustness. Therefore, this paper focuses on the method of classifying prostate cancer grade on multi-modality MRI images and maintaining robustness. In this paper, we propose a novel and effective multi-modal convolutional neural network for discriminating prostate cancer clinical severity grade, i.e. robust multimodal feature disentanglement attention net(RMDANet), and greatly improve the accuracy and robustness. T2-weighted(T2) and Diffusion-weighted imaging(DWI) are mainly used in this article. Experiments were conducted on the ProstateX dataset and augmented with hospital data, By comparing with other baseline methods, multi-modal dual input methods, SOTA methods, the AUC values obtained by the proposed model in this paper after the test set are higher than those of other classical models, the AUC value reached 0.835.

Purpose: We proposed a method for predicting the outcome of extracorporeal shock wave lithotripsy (ESWL) treatment using CT texture feature analysis. Method: 171 cases of CT images with a urinary stone are used in this study. 97 of them are stone-free (SF) cases who were successfully treated with ESWL. The proposed SF case classification method is based on CT texture features analysis within the stone region. The image feature set includes stone volume, six statistics of the CT intensity, and 12 features describing intensity gradient concentration to the center of the stone. A support vector machine with radial basis function kernel is used for SF identification. Result: The five-fold cross validation showed the SF classification accuracy of 56.8% on average. The area under the ROC curve was 0.612. Conclusion: These results confirmed the usefulness of the proposed method. The future work is to improve the performance by adding various features.

To prevent Covid-19 pandemic, infection screening system with high accuracy is greatly anticipated. Body temperature measurement at airport quarantine is easily affected by antipyretic medications and will result in low accuracy. Hence our group has developed non-contact infection screening system by measuring body temperature, heart rate and respiration rate.Proposed system measures body temperature by estimating prefrontal cortex temperature from facial surface near inner corner of an eye and heart rate is detected by luminance changes on face obtained with CCD camera. Using thorax- six-partitioning independent component analysis, Time-of-Flight sensor determined respiratory rate without the influence of body movements. Test with six healthy students showed correlation coefficient with reference was 0.86 and mean squared error was 1.98 bpm.Our system with Time-of-Flight sensor is capable of diagnosing chronic obstructive pulmonary disease which is the second leading cause of severity in COVID-19, and will be operated at hospital soon.

We report on the development of KindAI-COVID, a system using deep-learning to differentiate COVID-19 pneumonia using chest CT images. Our system's feature is to differentiate COVID-19 pneumonia for each CT slice to collect many training images for deep-learning. First, the range of slices needed for diagnosis is specified by an experienced radiologist. Next, the lung fields are extracted using U-net. The images are differentiated using fine-tuned GoogLeNet pre-trained on the ImageNet. In this study, CT data of 39 patients with COVID-19 pneumonia and 91 patients with non-COVID-19 pneumonia taken at Kindai University Hospital are included. The accuracies were 83.1% and 87.2% per slice and per examination, respectively. Our system was able to differentiate COVID-19 pneumonia with relatively high accuracy. We believe that limiting the target to only the slices needed for diagnosis and training deep-learning on many slices were effective.

Early detection of COVID-19-derived and other pneumonia is crucial to avoid disease spreading. Hence, we developed real-time pneumonia monitoring system and applied in clinical setting. System measures respiratory rates (RR) and heart rates (HR) and HF of heart rate variability which stands for parasympathetic nervous activity using microwave radars located beneath bed mattress. Nurse monitor displays, RR, HR, cardiopulmonary arrest, suspect of pneumonia, and suspect of arterial fibrillation responsible for brain infarction in real-time. In order to compensate drastic individual differences of vital signs in elderly patients, we adopted Mahalanobis's squared distance, a non-Euclidian distance determined by RR, HR and HF of the past 10 hours divided into anterior (before pneumonia pathogenesis) and posterior 5-hour periods (after pneumonia pathogenesis). Two out of 20 inpatients (8 males and 12 females, 74 +/- 14 years at Genki-kai Yokohama Hospital) developed pneumonia during 5-days validation, with sensitivity of 100% and NPV of 100%.

In dialysis, arteriovenous fistula is generally indispensable. However, it often causes stenosis and occlusion, For its daily management, we have proposed an optical transillumination imaging technique and proved its fundamental feasibility. In clinical practice, the optical conditions of the forearm differ from patient to patient. The sensitivity to detect the boundary between the blood and the blood vessel wall should be improved. To solve these problems, we introduced a differential principle. Two transillumination images were taken at two near-infrared wavelengths in which the inter-wavelength absorption difference is much larger for the blood than for the blood vessel wall. In the difference image, we could enhance the boundary with less individual difference. In experiments, it was confirmed that we could measure the blood vessel inner diameter as small as 2.0 mm with less than 20% error.

Due to the labor shortage at Long-Term Care Health Facilities, the number of fall accidents is increasing at night. Therefore, to make up for labor shortages and assist with operations, it is necessary to construct a monitoring system that can detect the getting up action of the people requiring nursing care. So, in this study, we proposed a monitoring system that can detect the getting up using the infrared camera. As a verification result, we found that by using the near-infrared reflective sheet, it became easy to detect and track the actions and the features of the actions could be obtained. In addition, we proposed an algorithm that is a novel and specific to the body movement that proceeds to bed fall.

One of the early symptoms of musculoskeletal disorders is called "locomotive syndrome" (Locomo). There is a qualitative index called "Locomomo-Degree Test," which can to determine whether a person has Locomo or not, but has issues in safety and simplicity. As a preliminary step to establish an automatic and quantitative evaluation method of Locomo degree, we tried to determine whether a person has Locomo or not by using acceleration sensor and a machine learning model. To investigate a new index to discriminate Locomo from the characteristics of human movement and to obtain the independent variables for machine learning, one-legged stand test and alternate one-legged stand test were conducted for elderly persons. A statistical method, multivariate logistic regression analysis and a machine learning method, Gradient Boosting Decision Tree (GBDT), were used to predict the presence or absence of locomotion. As a result, the performance of GBDT exceeded that of our previous study.

<Purpose>To detect bone metastases on FDG-PET/CT images with fewer false positives (FPs), we propose a method based on cascade voxel classification by two types of unsupervised AI anomaly detection.<Methods>Firstly, anomaly detection using Mahalanobis distance from normal bone voxels is adopted at bone voxels to extract voxels with abnormal CT value and SUV coarsely. Secondly, nonlinear anomaly detection using one-class support vector machine (OCSVM) is applied to the coarsely extracted voxels. The OCSVM uses seven local texture features to enhance metastases-like patterns. Finally, local maxima of the enhanced image are detected as the metastasis candidates.<Results>An experiment with ten images including 19 metastases showed the result with 89.5 % sensitivity and 59.5 FPs per case. The number of FPs was 93.9 % lower than when OCSVM was used only.<Conclusion>This result confirmed the effectiveness of the proposed method.

Molecular-targeted anticancer drugs are administrated for any patients in a fixed-dose. The plasma concentration considerably varies among individuals, inducing serious adverse events in some occasions. Recent studies showed relevance of the plasma drug level to the efficacy and toxicity. Nevertheless, the measurement at clinical sites has not yet been fully achieved, owing to lack of the rapid and easy method. To address this issue, we developed a strategy with an electrochemical sensor composed of conductive diamond. Initially, rat plasma mixed in advance with pazopanib was tested. Linear concentration-dependent response was observed along the therapeutic window. Each measurement took ~35 s, and the process was completed in ~10 min. Next, from rats orally administered with pazopanib, and blood of < 60 μL were longitudinally sampled multiple times. Measured T_max was ~4 hours, as described in the literature. Finally, we constructed a portable, palm-size device. These approaches accelerate tailored medicine for cancer.

This study aims at non-invasive venous pressure measurement for early diagnosis of chronic venous insufficiency of the lower extremities. Focusing on lower limb venous hypertension in patients, we are investigating the development of a pen-type compression device for non-invasive venous pressure measurement and a method for extracting vascular region. As a previous study, the ultrasonic probe with pressure measuring device was used to observe the state of the subcutaneous surface in the vein and measure the pressure when the vein was compressed. But the device is very expensive. We have developed a pen-type compression device to compress blood vessels. In order to analyze the venous pressure, we developed a real time blood vessel extraction method. A system was created and the external pressure measured by the pen-type compression device was synchronized with the ultrasonic video. We also confirmed the stability of the system using this non-invasive venous pressure measurement method.