Oral cancer is one of the most commonly found cancers worldwide. Oral Epithelial Dysplasia (OED) is an Oral Potentially Malignant Disorder (OPMD) that can be characterized for preventive oral cancer screening. The standard for OED histological grading is conducted via the epithelial regions of tissue biopsies. However, this procedure is laborious, time-consuming, and subjective; consequently, it is prone to variability due to fatigue and limited expertise. Therefore, this study aims to explore the potential of using Convolutional Neural Network (CNN) and Transformer models for an automated epithelium segmentation algorithm directly from Whole Slide Images (WSIs). This approach can reduce the manual process and support pathologists in grading activities. Accordingly, candidate architectures based on CNN and Transformer are selected: UNet, ResNet50-UNet, VGG19-UNet, Swin-UNet, and MISSFormer. These models are trained using patch-based segmentation to mitigate the high computational cost caused by processing WSIs. The results indicate that UNet, optimized with the ADAM optimizer, demonstrates the best performance in patch-based segmentation with Intersection over Union (IoU) of 0.82 and Dice-Similarity Coefficient (DSC) of 0.87. Furthermore, this model achieves the highest IoU and DSC for tissue-level prediction, scoring 0.88 and 0.94, respectively. According to the experiment, overlapping and non-overlapping patching strategies perform similarly in most of the selected architectures. The latter approach, hence, is suggested for computational efficiency. These results can support enhancing automated epithelium segmentation to provide a reliable tool for assisting pathologists.
Stroke is a disease with a high mortality rate and is one of the leading causes of acquired disability in adults. Some patients experience some forms of disability after a stroke, such as long-lasting physical paralysis, speech impairment, and cognitive impairment. For post-stroke hemiplegia, studies have demonstrated that long-term and continuous resistance training can help patients regain some degree of motor ability. However, the fact is that some patients have not regained motor function after 3-6 months of rehabilitation training. The reasons for this are that it is difficult for patients to adhere to the training and to obtain appropriate training guidance.To address the problem of inadequate resource reserves for rehabilitation physicians, robotic systems were introduced into post-stroke rehabilitation training from the 1990s. However, in actual use, the motor assistance that robotic systems can provide is still judged by rehabilitation physicians based on the patient's status and degree of muscle damage, and the patient's motor initiative and attentional state cannot be fully explored.In this study, a hybrid brain-computer interface-based post-stroke rehabilitation training system was designed to infer patients' muscle rehabilitation status based on the level of corticomuscular coherence (CMC), so as to provide adaptive training guidance to patients during the long-term and continuous training process, promote patient initiative during the training process, and enhance the training effect. This study focuses on the correlation between corticomuscular coherence and voluntary muscle contraction, so as to infer the muscle contraction status of patients based on corticomuscular coherence. Finally the present study obtained a significant positive linear relationship between corticomuscular coherence and different force magnitudes in the alpha band.
We evaluated the effects of transcranial direct current stimulation (tDCS) on cognitive performances including the arithmetic performance, working memory, and electroencephalography (EEG) measurements of healthy participants when applied over multiple sessions spanning a five-day period. In a double-blind crossover randomized controlled study, participants received either active left tDCS or sham stimulation. Despite no significant changes in arithmetic performance, the tDCS group displayed enhanced reaction time and accuracy in a 2-back task compared with the control (sham) group, and EEG measurements revealed an increase in the alpha-band power spectral density (PSD) at the left temporal lobe for the treatment group over the experimental period, suggesting modulations in cortical activity. Further research is needed to elucidate the specific cognitive regions affected by tDCS and the underlying neurophysiological mechanisms.
Nowadays, people suffer from a variety of illnesses related to the body, including those that affect the lower body and make it difficult or impossible for patients to walk. The main goal of a wheelchair is to transport patients who cannot walk to their desired destination, either by themselves or with the help of a caregiver. The project's objective was to develop an effective solution for patients who are unable to operate the joystick on electronic wheelchairs. The smart wheelchair would be beneficial for patients who primarily reside in their rooms. This smart wheelchair will create a map of the room, allowing the patient to navigate it easily using only one finger to control it by simply clicking on the pre-created map.This project developed an electronic wheelchair that could be controlled and driven autonomously using a ROS (Robot Operating System) platform. Additionally, it utilized the SLAM (Simultaneous Localization and Mapping) method, which enables the construction of a map of an unknown environment while simultaneously localizing the wheelchair for navigation purposes. To assess the accuracy of map creation in an unknown environment, calibration was performed with and without loading weight to measure the distance and angle of the wheelchair's movement.The results of this project show that the developed electronic wheelchair successfully builds a map of an unknown environment and utilizes this map for navigation. Users can simply click on their desired destination on the map, and the wheelchair will autonomously navigate towards it. Moreover, the wheelchair can avoid obstacles while navigating. Although the smart wheelchair can perform well, there are some limitations due to the RPLidar. The RPLidar detects everything on the ground in parallel when creating the map and while navigating. Therefore, if any object is located below the RPLidar, it may not be able to detect it.
Medium-chain-length polyhydroxyalkanoate (MC-LPHA) is a biodegradable polyester with an ability to biodegrade and interact with the human body, making it a material of interest for various medical applications. As a versatile biopolymer, there is a growing trend towards its potential utilization in the field of medical elastomers. This study assessed the medical adhesive capabilities of MCL-PHA by evaluating its physical properties compared to commonly used medical adhesives. The results demonstrate that MCL-PHA possesses multiple characteristics aligning with those of existing medical adhesives. Specifically, MCL-PHA exhibits an adhesive strength to porcine skin of 50.2 kPa, closely resembling wound dressing adhesives and surpassing surgical adhesives. Furthermore, its shear resistance is comparable to surgical adhesive, measured at 27.6 kPa. The testing of material detachment from porcine skin indicates that MCL-PHA can be removed without causing harm. These findings illustrate the potential of MCL-PHA as a medical adhesive, indicating its properties which are consistent with both external wound dressings and surgical adhesives.
In several fields, such as microbiology research, medical diagnostics, and food safety evaluation, bacterial colony counting is extremely important. However, the method of manual counting is time-consuming, labor-intensive, and prone to human error. This research approached these problems by using MATLAB's image processing feature to automatically count the number of bacterial colonies on agar plates. This technique effectively detects bacterial colonies from photos of agar plates by using image analysis algorithms. The images of agar plates were captured while controlling the lighting and adjusting the size to achieve the highest possible image quality. This study encompassed 10 bacterial species, achieving an accuracy of approximately 80%. This level of precision underscores the reliability and effectiveness of our automated system.
This article deploys a 5-level shorten rectangular wave technique to measure lock-in electrical bio-impedance (EBI) in medical diagnosis. The new shorten rectangular EBI signal has better properties in eliminating odd harmonics compared to the conventional 3-level shorten rectangular wave technique. The results show that the measurement errors in the 3-component EBI are reduced about 0.3% for R, X, Z and 3% for Phase(Φ) when the 5-level signal is used instead 3-level signal.
Failure of glaucoma surgery occurred due to fibrosis formation. To prevent fibrosis formation, an anti-inflammation agent, such as dexamethasone (DEX), was utilized to inhibit inflammation during the wound healing process, leading to fibrosis reduction. In the present study, DEX is loaded into an ultrasonication-induced silk fibroin/hyaluronic acid (SF/HA) hydrogel as a sustained-release drug delivery system. The SF/HA hydrogel can be fabricated using ultrasonic induction resulting to homogeneous gel within 1 day after incubation at 37°C. For injectability test, the results confirmed that the hydrogel could be easily injected through the 30-gauge needle, particularly for hydrogel containing less concentration and higher HA content. Furthermore, The SF/HA hydrogel showed non-swelling ability, released DEX up to 60% along 30 days of incubation in balanced salt solution, and non-cytotoxic to fibroblasts. However, these properties of the SF/HA hydrogels are independent of their concentrations and SF/HA ratios. As aforementioned, the ultrasonication-induced SF/HA hydrogel could be utilized as an injected carrier of DEX for incorporation with postoperative glaucoma surgery to reduce inflammation and fibrosis formation.
To develop an experimental system that could be used for physiological experiments on voluntary finger movements in humans, we developed an experimental system that can realize passive finger movements by reproducing recorded users' own active movement trajectories using a haptic device called SPIDAR-GCC. The efficacy of the system was verified through high agreement of the finger trajectories between active and passive movements and no muscle activations during the passive movements. We further confirmed that high signal-to-noise ratio sensory evoked potentials (SEPs) derived from median nerve stimulation of the wrist was observed by electrical stimulation exceeding motor threshold without muscle fatigue throughout the experiment. As N20/P25 component of SEPs from 20 participants showed amplitude changes according to the movement conditions, these results suggested that the proposed experimental system has the capability to be utilized as a physiological experimental system for elucidating the neural control mechanisms of voluntary movements.
Optical Bone Densitometry has been introduced to the world as it serves screening task for osteoporosis which significantly save individual cost and time spent. In this report, we aim to diagnosis and analyze errors occur from this device in many perspectives included with determine acceptable range of percentage error, precision testing, device calibration, and effect of environmental light. Our findings show that there are some devices that have high percentage errors than acceptable value at 15% which environmental light significantly affect outcome, and it is suitable to select material for calibration which has similar properties to human bone as much as possible due to device ‘s principle.
Brain tumors, impacting a substantial global population annually, necessitate precise detection and classification for timely intervention and effective therapy. Though deep learning models have exhibited potential in medical image interpretation, a demand persists for enhanced accuracy and efficiency. This study introduces an optimized solution for brain tumor detection and classification via a concatenated EfficientNet-ConvNeXt model. This novel approach merges the power of EfficientNet and ConvNeXt—two formidable neural networks—to attain extraordinary precision in categorizing various brain tumor types, namely glioma, meningioma, pituitary tumor, and non-tumor. Experimental evaluations validate the model’s superiority over standalone architectures and existing deep learning techniques in terms of accuracy, sensitivity, and specificity. Demonstrating robustness against image quality fluctuations and variability in tumor types, the model exhibits strong potential for real-world clinical usage. Implementation of our proposed concatenated EfficientNet-ConvNeXt model resulted in substantial performance elevation, achieving an exceptional 99% predictive accuracy. These findings underscore our approach’s accuracy and efficiency, offering substantial aid to radiologists and clinicians in early-stage brain tumor detection and classification. The model’s predictive capabilities can considerably influence patient prognosis and therapy planning through enabling early intervention.
Artificial channels that transport substances across cell membranes are being actively studied for analytical techniques, drug delivery, and other applications. One example is carbon nanotubes (CNTs), tubular carbon materials with diameters in the nanometer range, whose hollow structure allows ions and small molecules to pass through easily. When single-walled CNTs are cut to lengths of several to several tens of nanometers (ultrashort CNTs, US-CNTs) and coated with phospholipids, they can spontaneously insert into cell membranes and artificial lipid bilayers. Furthermore, US-CNTs have attracted attention as artificial ion channels whose mechanical and chemical properties can be modified to control their permeability. The principle of US-CNTs forming channels in lipid bilayers has been investigated by molecular dynamics simulations, although this has seldom been examined experimentally. In this work, we estimated the channel-forming ability of US-CNTs by measuring the channel current signals of CNTs with different length distributions. The channel current signal pattern was found to differ depending on the length distribution, suggesting that length affects the ability of CNTs to form channels in membranes.
This research aims to use deep learning techniques to segment oral lesions in medical images for use as a preprocessing step in a classification model. Due to the complexity of oral lesions with undefined margins and dynamic shapes, and the limited amount of data certified by dentists, this approach was found to be underfitting and unsatisfactory. To improve the accuracy of the model, a new approach was proposed to segment interferences such as teeth from the images. This allows the model to better focus on the oral lesions. To achieve this goal, we implemented U-net models with different additional Convolutional Neural Networks (CNN) backbones, including DenseNet 121, EfficientNet B3, VGG 19, ResNet 18, SE-ResNet 18, ResNeXt 50, Inception V3, Mobilenet V2 and SE-ResNeXt 50. A segmentation model was trained with five classes of oral lesions: leukoplakia, pseudomembranous candidiasis, lichen planus, ulcer, and other white lesions. The results showed that DenseUNet and EfficientUNet achieved the highest validation and Intersection over Union (IoU) scores of 98% and 92%, respectively. Our proposed approach effectively segmented the interferences from the images, demonstrating the success of these models in handling the approach. Subsequently, a CNN model of DenseNet 121 was employed for classification. The training accuracy achieved 99.1%, while the validation and test accuracies reached 86.1% and 75.5%, respectively.
Middle ear infections are a prevalent health issue affecting individuals across all age groups. Timely and accurate detection is crucial for effective treatment and preventing complications. This study introduces an innovative smartphone-based machine learning system for detecting middle ear infections. The system combines an acoustic analysis module, using acoustic reflectometry theory, with a custom smartphone application. By connecting an earphone device to the smartphone, chirping noises are emitted and received in the ear, and the embedded microphone captures and analyzes the reflected sound waves using a machine learning algorithm. The system accurately categorizes the results into "no water," "monitor," and "water," identifying middle ear fluid, an important indicator of infection. Early detection of fluid buildup can lead to timely intervention, potentially preventing infections. Extensive testing was conducted on various subjects to validate the system's accuracy against a commercial device. The research represents a significant advancement, providing a non-invasive, accessible, and cost-effective solution for detecting middle ear infections in both home and clinical settings, promoting self-monitoring of ear health, and reducing associated complications.
This research aims to develop an advanced medical device designed to enhance the diagnostic quality of conventional colposcopy. The device utilizes cutting-edge technologies, including 3D image synthesis via stereoscopic imaging and polarized glasses as the primary focus of the study is to improve cervical cancer. The research scope encompasses enhancing spatial information, The research scope involves enhancing spatial information while allowing the doctor to maintain the advantage of near vision and enabling multi-angle imaging. The hardware of the colposcope is based on the design from Duke University's 2018 research. Our 3D Cervical Assessment - Fine-tuned Optics colposcope (3CA-FO) is capable of providing real-time 3D imaging with precise calibration, achieved through the utilization of the Embedded Mono Calibration for Heterogeneous Lenses technique. This ensures an instantaneous and high-quality 3D output response.
Cell patterning technology is used to understand single- cell morphology and cell-cell interactions and can also be used to capture cells and control their position. However, most research has focused on floating cells and cell-like particles rather than on adherent cells. In this study, we propose a technique that can control the position of adherent cells with an external field. This technique allows cell-cell networks of any shape to be constructed and may contribute to the construction of neuronal networks.
Erythemato-squamous diseases (ESD) are dermatological diseases that significantly impact the quality of life of an increasing number of patients worldwide. This study used a publicly available clinical dataset of 366 patients from the Department of Computer Engineering and Information Science, Bilkent University with 34 predictors contributing to the classification of ESDs. The data was curated to ensure unbiasedness and accuracy before applying principal component analysis and machine learning models to identify crucial factors in classifying the six main types of ESDs. 31 different machine learning models, including Tree, Linear Discriminant, Quadratic Discriminant, Naïve Bayes, SVM, KNN, Ensemble, Neural Network, and Kernel were trained, validated, and the classification accuracy was compared. The model that is the most adequate is the Fine KNN, which has the highest cross-validation classification accuracy at 100%. This model requires only eight predictors: itching, the Koebner phenomenon, follicular papules, fibrosis of the papillary dermis, spongiosis, inflammatory mononuclear infiltrate, bandlike infiltrate, and age.
Vascular bifurcation sites have been identified as preferential locations for intracranial aneurysm presence. The distal bifurcation of the basilar artery (BA) is known for its intricate geometry and frequent association with basilar aneurysms. Previous research on BA apex angles and BA aneurysms predominantly focused on tip aneurysms of the BA. Meanwhile, BA trunk aneurysms have not received much attention. Furthermore, limited attention has been given to the angles between the posterior cerebral artery (PCA) and the superior cerebellar artery (SCA). Our study aimed to address this lack of information and explored potential variations in BA angles across different types of BA aneurysms (basilar tip and trunk aneurysms). The study group comprised a cohort of thirty-three patients selected through a retrospective analysis of patient data from the AneuX project, a multicenter database on intracranial aneurysms. These patients were categorized into basilar trunk, basilar tip aneurysm, and non-BA aneurysm subgroups, each consisting of eleven patients. We investigated the angles among PCA, PCA-SCA, and SCA-BA in all patients. Significant differences were observed between the basilar trunk and tip aneurysm groups in the PCA and PCA-SCA angles. Notably, the angles within the non-BA group were found to fall within the range between the basilar tip and trunk groups. In conclusion, we propose a classification scheme for BA aneurysms and suggest considering PCA-SCA angles in exploring the correlation between BA angles and the presence of aneurysms.
Various surgical treatment options have been documented for the management of extra-articular distal humerus fractures. However, the optimal approach remains a subject of controversy, primarily due to the variability in individual fracture characteristics. In this context, the present study aims to investigate the influence of implants on the stability of extra-articular distal humerus fracture fixation and the associated failure mechanism through computational simulations. In this study, three-dimensional models of a locking compression plate (LCP) fixation and an intramedullary nail fixation for extra-articular distal humerus fracture were generated and finite element analysis was performed to compare the effectiveness of these fixations under loading conditions mimicking daily activities. The study focused on evaluating von-Mises stress and displacement on both the implants and bone fragments. The results revealed high-risk regions due to exceeded von-Mises stress, shedding light on the failure mechanism of fixation. To conclude, the results highlighted the superior overall performance of intramedullary nail fixation, particularly under shear force conditions. These findings and the ensuing discussion provide valuable insights and considerations for the further development of a novel fixation system.
This is the first report of sterile Thai silk fibroin (SF) solution (2-4% (w/v) in water) production at semi-commercial scale under the medical device management system conforming to standard ISO13485:2016. Thai silk cocoons (Bombyx mori, local strain, named Nangnoi Srisaket 1) was maintained, cultured and produced in controlled conditions at The Queen Sirikit Sericulture Center Nakhon Ratchasima, Thailand. The GMP (Medical Device) conformed protocol was optimized. The degummed silk fibers were dissolved in concentrated ionic liquid and purified in the clean room facility of the Department of Medical Sciences, Ministry of Public Health, Thailand. With minimal sterilization of the end product (4%w/w SF Solution), sterility and endotoxin test of the product indicated that the purified proteins were negative for bacteria, mycobacteria, yeast and fungi. The endotoxin levels of all inputs, intermediates and the product was <0.1 EU/ml conforming to the USP standard (for water for injection at the endotoxin level limits <0.25 EU/ml). This indicated that the Thai SF solution can be used for medical and health applications.
