A5105923667- Colorectal Cancer Screening and Detection
- Cardiac electrophysiology and arrhythmias
- Radiomics and Machine Learning in Medical Imaging
- ECG Monitoring and Analysis
- Gastric Cancer Management and Outcomes
- AI in cancer detection
- Cardiac Arrhythmias and Treatments
- Advanced Neural Network Applications
- Non-Invasive Vital Sign Monitoring
- Heart Rate Variability and Autonomic Control
- Cardiac pacing and defibrillation studies
- Advanced Image and Video Retrieval Techniques
- Endometrial and Cervical Cancer Treatments
- Atrial Fibrillation Management and Outcomes
- Cardiovascular Function and Risk Factors
- Gynecological conditions and treatments
- Optical Coherence Tomography Applications
- COVID-19 diagnosis using AI
- Cervical Cancer and HPV Research
- Photoacoustic and Ultrasonic Imaging
- Advanced Sensor and Energy Harvesting Materials
- EEG and Brain-Computer Interfaces
- Retinal Imaging and Analysis
- Machine Learning in Healthcare
- Bone and Joint Diseases
Tokyo Institute of Technology
2025
University of Aizu
2015-2024
Tokyo Medical and Dental University
2024
Harbin Institute of Technology
2024
Huazhong University of Science and Technology
2024
Tongji Hospital
2024
Institute of Science Tokyo
2024
Mayo Clinic in Florida
2024
Chung Shan Medical University
2023
Chang'an University
2022-2023
Encoder-decoder networks are state-of-the-art approaches to biomedical image segmentation, but have two problems: i.e., the widely used pooling operations may discard spatial information, and therefore low-level semantics lost. Feature fusion methods can mitigate these problems feature maps of different scales cannot be easily fused because downand upsampling change resolution map. To address issues, we propose INet, which enlarges receptive fields by increasing kernel sizes convolutional...
Facial action unit (AU) recognition is a crucial task for facial expressions analysis and has attracted extensive attention in the field of artificial intelligence computer vision. Existing works have either focused on designing or learning complex regional feature representations, delved into various types AU relationship modeling. Albeit with varying degrees progress, it still arduous existing methods to handle situations. In this paper, we investigate how integrate semantic propagation...
Introduction Blood pressure (BP) serves as a crucial parameter in the management of three prevalent chronic diseases, hypertension, cardiovascular and cerebrovascular diseases. However, conventional sphygmomanometer, utilizing cuff, is unsuitable for approach mobile health (mHealth). Methods Cuffless blood measurement, which eliminates need considered promising avenue. This method based on relationship between pulse arrival time (PAT) parameters BP. In this study, transit (PTT) was derived...
Cascade is a widely used approach that rejects obvious negative samples at early stages for learning better classifier and faster inference. This paper presents chained cascade network (CC-Net). In this CC-Net, there are many stages. Preceding placed shallow layers. Easy hard examples rejected layers so the computation deeper or wider not required. way, features classifiers latter handle more difficult with help of in previous It yields consistent boost detection performance on PASCAL VOC...
Obstructive sleep apnea (OSA) is a common chronic disorder that disrupts breathing during and associated with many other medical conditions, including hypertension, coronary heart disease, depression. Clinically, the standard for diagnosing OSA involves nocturnal polysomnography (PSG). However, this requires expert human intervention considerable time, which limits availability of diagnosis in public health sectors. Therefore, electrocardiogram (ECG)-based methods detection have been...
Radiofrequency catheter ablation (RFCA) is an effective therapy for atrial fibrillation (AF). However, it the problem of AF recurrence remains. This study investigates whether a deep convolutional neural network (CNN) can accurately predict in patients with who underwent RFCA, and compares CNN conventional statistical analysis.
It is promising to deploy CNN inference on local end-user devices for high-accuracy and time-sensitive applications. Model parallelism has the potential provide high throughput low latency in distributed inference. However, it non-trivial use model as original inherently tightly-coupled structure. In this article, we propose DeCNN, a more effective approach that uses decoupled structure optimize devices. DeCNN novel consisting of three schemes. Scheme-1 structure-level optimization. exploits...
Purpose: Endometrial thickness is one of the most important indicators in endometrial disease screening and diagnosis. Herein, we propose a method for automated measurement from transvaginal ultrasound images. Methods: Accurate relies on endometrium segmentation images that usually have ambiguous boundaries heterogeneous textures. Therefore, two-step was developed thickness. First, semantic based deep learning, to segment 2D Second, estimated segmented results, using largest inscribed circle...
This study aims to evaluate leg movement by integrating gait analysis with surface electromyography (sEMG) and accelerometer (ACC) data from the lower limbs. We employed a wireless, self-made, multi-channel measurement system in combination commercial GaitUp Physilog® 5 shoe-worn inertial sensors record walking patterns muscle activations of 17 participants. approach generated comprehensive dataset comprising 1452 samples. To accurately predict parameters, machine learning model was...
The electrocardiogram (ECG) is widely used for cardiovascular disease diagnosis and daily health monitoring. Before ECG analysis, quality screening an essential but time-consuming experience-dependent work technicians. An automatic assessment method can reduce unnecessary time loss to help cardiologists perform diagnosis. This study aims develop system search qualified ECGs interpretation. proposed consists of data augmentation parts. For augmentation, we train a conditional generative...
Abstract Retinal segmentation is a prerequisite for quantifying retinal structural features and diagnosing related ophthalmic diseases. Canny operator recognized as the best boundary detection so far, often used to obtain initial of retina in segmentation. However, traditional susceptible vascular shadows, vitreous artifacts, or noise interference segmentation, causing serious misdetection missed detection. This paper proposed an improved automatic boundaries. The algorithm solves problems...
Flexible endoscopic evaluation of swallowing (FEES) is considered the gold standard in diagnosing oropharyngeal dysphagia. Recent advances deep learning have led to a resurgence artificial intelligence-assisted computer-aided diagnosis (AI-assisted CAD) for variety applications. AI-assisted CAD would be remarkable benefit providing medical services populations with inadequate access dysphagia experts, especially aging societies. This paper presents an named FEES-CAD aspiration and...
Photoplethysmography (PPG) signal is potentially suitable in atrial fibrillation (AF) detection for its convenience use and similarity physiological origin to electrocardiogram (ECG). There are a few preceding studies that have shown the possibility of using peak-to-peak interval PPG (PPIp) AF detection. However, as generalized model, accuracy an detector should be pursued on one hand; other hand, generalizability paid attention view individual differences manifestation even same arrhythmia...
Capsule endoscopy is a common method for detecting digestive diseases. The location of capsule endoscope should be constantly monitored through visual inspection the endoscopic images by medical staff to confirm examination’s progress. In this study, we proposed computer-aided analysis (CADx) localization endoscope. At first, classifier based on Swin Transformer was classify each frame videos into stomach, small intestine, and large respectively. Then, K-means algorithm used correct outliers...
Background and Study Aims Small polyps are occasionally missed during colonoscopy. This study was conducted to validate the diagnostic performance of a polyp‐detection algorithm alert endoscopists unrecognized lesions. Methods A computer‐aided detection (CADe) developed based on convolutional neural networks using training data from 1991 still colonoscopy images 283 subjects with adenomatous polyps. The CADe evaluated validation dataset including 50 short videos 1–2 (3.5 ± 1.5 mm, range 2–8...
Automatic polyp detection is reported to have a high false-positive rate (FPR) because of various polyp-like structures and artifacts in complex colon environment. An efficient polyp's computer-aided (CADe) system should sensitivity low FPR (high specificity). Convolutional neural networks been implemented colonoscopy-based automatic achieved performance improving rate. However, environments caused excessive false positives are going prevent the clinical implementation CADe systems. To...
Background and study aims Colorectal cancers (CRC) with deep submucosal invasion (T1b) could be metastatic lesions. However, endoscopic images of T1b CRC resemble those mucosal CRCs (Tis) or superficial (T1a). The aim this was to develop an automatic computer-aided diagnosis (CAD) system identify based on plain images. Patients methods In two hospitals, 1839 non-magnified from 313 (Tis 134, T1a 46, 56, beyond 37) sessile morphology were extracted for training. A CAD trained the data...
The understanding of long-range pixel–pixel dependencies plays a vital role in image segmentation. use CNN plus an attention mechanism still has room for improvement, since existing transformer-based architectures require many thousands annotated training samples to model spatial dependencies. This paper presents smooth branch (SAB), novel architecture that simplifies the biomedical segmentation small datasets. SAB is essentially modified operation implements subnetwork via reshaped feature...
Recently many wearable devices have been developed for casual health management. ECG is an important tool monitoring of cardiac arrhythmias. Wearable are able to collect electrocardiogram (ECG) data in a convenient and low-cost way. should sufficient quality experts analyze make clinical decision. Identifying low signals advance will significantly facilitate diagnosis. In this paper, we use 1D convolutional neural network (CNN) classify single-lead with duration 10 s as "acceptable" or...
Cascade is a widely used approach that rejects obvious negative samples at early stages for learning better classifier and faster inference. This paper presents chained cascade network (CC-Net). In this CC-Net, the cascaded stage aided by classification scores in previous stages. Feature chaining further proposed so feature current uses features as prior information. The ConvNet classifiers of multiple are jointly learned an end-to-end network. way, latter handle more difficult with help It...