Medical Engineering & Physics最新文献

Epilepsy is a chronic disease caused by repeated abnormal discharge of neurons in the brain. Accurately predicting the onset of epilepsy can effectively improve the quality of life for patients with the condition. While there are many methods for detecting epilepsy, EEG is currently considered one of the most effective analytical tools due to the abundant information it provides about brain activity. The aim of this study is to explore potential time-frequency and channel features from multi-channel epileptic EEG signals and to develop a patient-specific seizure prediction network. In this paper, an epilepsy EEG signal classification algorithm called Channel Recurrent Criss-cross Attention Network (CRCANet) is proposed. Firstly, the spectrograms processed by the short-time fourier transform is input into a Convolutional Neural Network (CNN). Then, the spectrogram feature map obtained in the previous step is input into the channel attention module to establish correlations between channels. Subsequently, the feature diagram containing channel attention characteristics is input into the recurrent criss-cross attention module to enhance the information content of each pixel. Finally, two fully connected layers are used for classification. We validated the method on 13 patients in the public CHB-MIT scalp EEG dataset, achieving an average accuracy of 93.8 %, sensitivity of 94.3 %, and specificity of 93.5 %. The experimental results indicate that CRCANet can effectively capture the time-frequency and channel characteristics of EEG signals while improving training efficiency.

Cranioplasty is the surgical repair of a bone defect in the skull resulting from a previous operation or injury, which involves lifting the scalp and restoring the contour of the skull with a graft made from material that is reconstructed from scans of the patient's own skull. The paper introduces a 3D printing technology in creating molds, which are filled with polymethyl methacrylate (PMMA) to reconstruct the missing bone part of the skull. The procedure included several steps to create a 3D model in an STL format, conversion into a G-code which is further used to produce the mold itself using a 3D printer. The paper presents our initial experience with 5 patients who undergone cranioplasty utilizing 3D printed molds. Making a patient-specific model is a very complex process and consists of several steps. The creation of a patient-specific 3D model loading of DICOM images obtained by CT scanning, followed by thresholding-based segmentation and generation of a precise 3D model of part of the patient's skull. Next step is creating the G-code models for 3D printing, after which the actual models are printed using Fused Deposition Modeling printer and PLA material. All surgeries showed good match of the missing bone part and part created using 3D printed mold, without additional steps in refinement. In such a way, 3D printing technology helps in creating personalized and visually appealing bone replacements, at a low cost of the final product.

Stereotactic Radiosurgery (SRS) for brain tumors using Medical Linear Accelerator (LINAC) demands high precision and accuracy. A specific Quality Assurance (QA) is essential for every patient undergoing SRS to protect nearby non-cancerous cells by ensuring that the X-ray beams are targeted according to tumor position. In this work, a water-filled generic anthropomorphic head phantom consisting of two removable parts with eccentric holes was developed using Additive Manufacturing (AM) process for performing QA in SRS. In the patient specific QA, the planned radiation dose using Treatment Planning System (TPS) was compared with the dose measured in the phantom. Also, the energy consistency of radiation beams was tested at 200 MU for different energy beams at the central and eccentric holes of the phantom using an ionization chamber. Experimentally examined results show that planned doses in TPS are reaching the target within a 5% deviation. The ratio of the dose delivered in the eccentric hole to the dose delivered to the central hole shows variations of less than 2% for the energy consistency test. The designed, low-cost water-filled anthropomorphic phantom is observed to improve positioning verification and accurate dosimetry of patient-specific QA in SRS treatment.

Infrared thermography (IRT) is a well-known imaging technique that provides a non-invasive displaying of the ocular surface temperature distribution. Currently, compact smartphone-based IRT devices, as well as special software for processing thermal images, have become available. The study aimed to determine the possible use of smartphone-based IRT devices for real-time ocular surface thermal imaging. This study involved 32 healthy individuals (64 eyes); 10 patients (10 eyes) with proliferative diabetic retinopathy (PDR) and absolute glaucoma; 10 patients (10 eyes) with PDR, who underwent vitreoretinal surgery. In all cases, simultaneous ocular surface IRT of both eyes was performed. In healthy individuals, the ocular surface temperature (OST) averaged 34.6 ± 0.8 °C and did not differ substantially between the paired eyes, in different age groups, and after pupil dilation. In our study, high intraocular pressure was accompanied by a decrease in OST in all cases. After vitreoretinal surgery in cases with confirmed subclinical inflammation, the OST was higher than the baseline and differed from that of the paired eye by more than 1.0 °C. These results highlight that smartphone-based IRT imaging could be useful for the non-invasive detection of OST asymmetry between paired eyes due to increased intraocular pressure or subclinical inflammation.

As the number of patients with cardiovascular diseases (CVDs) increases annually, a reliable and automated system for detecting electrocardiogram (ECG) abnormalities is becoming increasingly essential. Scholars have developed numerous methods of arrhythmia classification using machine learning or deep learning. However, the issue of low classification rates of individual classes in inter-patient heartbeat classification remains a challenge. This study proposes a method for inter-patient heartbeat classification by fusing dual-channel squeeze-and-excitation residual neural networks (SE-ResNet) and expert features. In the preprocessing stage, ECG heartbeats extracted from both leads of ECG signals are filtered and normalized. Additionally, nine features representing waveform morphology and heartbeat contextual information are selected to be fused with the deep neural networks. Using different filter and kernel sizes for each block, the SE-residual block-based model can effectively learn long-term features between heartbeats. The divided ECG heartbeats and extracted features are then input to the improved SE-ResNet for training and testing according to the inter-patient scheme. The focal loss is utilized to handle the heartbeat of the imbalance category. The proposed arrhythmia classification method is evaluated on three open-source databases, and it achieved an overall F1-score of 83.39 % in the MIT-BIH database. This system can be applied in the scenario of daily monitoring of ECG and plays a significant role in diagnosing arrhythmias.

The neural control of human quiet stance remains controversial, with classic views suggesting a limited role of the brain and recent findings conversely indicating direct cortical control of muscles during upright posture. Conceptual neural feedback control models have been proposed and tested against experimental evidence. The most renowned model is the continuous impedance control model. However, when time delays are included in this model to simulate neural transmission, the continuous controller becomes unstable. Another model, the intermittent control model, assumes that the central nervous system (CNS) activates muscles intermittently, and not continuously, to counteract gravitational torque. In this study, a delayed reinforcement learning algorithm was developed to seek optimal control policy to balance a one-segment inverted pendulum model representing the human body. According to this approach, there was no a-priori strategy imposed on the controller but rather the optimal strategy emerged from the reward-based learning. The simulation results indicated that the optimal neural controller exhibits intermittent, and not continuous, characteristics, in agreement with the possibility that the CNS intermittently provides neural feedback torque to maintain an upright posture.

Computed tomography (CT) imaging is frequently employed in a variety of musculoskeletal research applications. Although research studies often use imaging protocols developed for clinical applications, lower dose protocols are likely possible when the goal is to reconstruct 3D bone models. Our purpose was to describe the dose-accuracy trade-off between incrementally lower-dose CT scans and the geometric reconstruction accuracy of the humerus, scapula, and clavicle. Six shoulder specimens were acquired and scanned using 5 helical CT protocols: 1) 120 kVp, 450 mA (full-dose); 2) 120 kVp, 120 mA; 3) 120 kVp, 100 mA; 4) 100 kVp, 100 mA; 5) 80 kVp, 80 mA. Scans were segmented and reconstructed into 3D surface meshes. Geometric error was assessed by comparing the surfaces of the low-dose meshes to the full-dose (gold standard) mesh and was described using mean absolute error, bias, precision, and worst-case error. All low-dose protocols resulted in a >70 % reduction in the effective dose. Lower dose scans resulted in higher geometric errors; however, error magnitudes were generally <0.5 mm. These data suggest that the effective dose associated with CT imaging can be substantially reduced without a significant loss of geometric reconstruction accuracy.

Epilepsy claims the lives of many people, so researchers strive to build highly accurate diagnostic models. One of the limitations of obtaining high accuracy is the scarcity of Electroencephalography (EEG) data and the fact that they are from different devices in terms of the channels number and sampling frequency. The paper proposes universal epilepsy diagnoses with high accuracy from electroencephalography signals taken from any device. The novelty of the proposal is to convert VEEG video into images, separating some parts and unifying images taken from different devices. The images were tested by dividing the video into labeled frames of different periods. By adding the spatial attention layer to the deep learning in the new model, classification accuracy increased to 99.95 %, taking five seconds/frame. The proposed has high accuracy in detecting epilepsy from any EEG without being restricted to a specific number of channels or sampling frequencies.

In addition to human donor bones, bone models made of synthetic materials are the gold standard substitutes for biomechanical testing of osteosyntheses. However, commercially available artificial bone models are not able to adequately reproduce the mechanical properties of human bone, especially not human osteoporotic bone.

To overcome this issue, new types of polyurethane-based synthetic osteoporotic bone models have been developed. Its base materials for the cancellous bone portion and for the cortical portion have already been morphologically and mechanically validated against human bone. Thus, the aim of this study was to combine the two validated base materials for the two bone components to produce femur models with real human geometry, one with a hollow intramedullary canal and one with an intramedullary canal filled with synthetic cancellous bone, and mechanically validate them in comparison to fresh frozen human bone.

These custom-made synthetic bone models were fabricated from a computer-tomography data set in a 2-step casting process to achieve not only the real geometry but also realistic cortical thicknesses of the femur. The synthetic bones were tested for axial compression, four-point bending in two planes, and torsion and validated against human osteoporotic bone.

The results showed that the mechanical properties of the polyurethane-based synthetic bone models with hollow intramedullary canals are in the range of those of the human osteoporotic femur. Both, the femur models with the hollow and spongy-bone-filled intramedullary canal, showed no substantial differences in bending stiffness and axial compression stiffness compared to human osteoporotic bone. Torsional stiffnesses were slightly higher but within the range of human osteoporotic femurs.

Concluding, this study shows that the innovative polyurethane-based femur models are comparable to human bones in terms of bending, axial compression, and torsional stiffness.

Sleep is an integral and vital component of human life, contributing significantly to overall health and well-being, but a considerable number of people worldwide experience sleep disorders. Sleep disorder diagnosis heavily depends on accurately classifying sleep stages. Traditionally, this classification has been performed manually by trained sleep technologists that visually inspect polysomnography records. However, in order to mitigate the labor-intensive nature of this process, automated approaches have been developed. These automated methods aim to streamline and facilitate sleep stage classification. This study aims to classify sleep stages in a dataset comprising subjects with insomnia, PLM, and sleep apnea. The dataset consists of PSG recordings from the multi-ethnic study of atherosclerosis (MESA) cohort of the national sleep research resource (NSRR), including 2056 subjects. Among these subjects, 130 have insomnia, 39 suffer from PLM, 156 have sleep apnea, and the remaining 1731 are classified as good sleepers. This study proposes an automated computerized technique to classify sleep stages, developing a machine-learning model with explainable artificial intelligence (XAI) capabilities using wavelet-based Hjorth parameters. An optimal biorthogonal wavelet filter bank (BOWFB) has been employed to extract subbands (SBs) from 30 seconds of electroencephalogram (EEG) epochs. Three EEG channels, namely: Fz_Cz, Cz_Oz, and C4_M1, are employed to yield an optimum outcome. The Hjorth parameters extracted from SBs were then fed to different machine learning algorithms. To gain an understanding of the model, in this study, we used SHAP (Shapley Additive explanations) method. For subjects suffering from the aforementioned diseases, the model utilized features derived from all channels and employed an ensembled bagged trees (EnBT) classifier. The highest accuracy of 86.8%, 87.3%, 85.0%, 84.5%, and 83.8% is obtained for the insomniac, PLM, apniac, good sleepers and complete datasets, respectively. Using these techniques and datasets, the study aims to enhance sleep stage classification accuracy and improve understanding of sleep disorders such as insomnia, PLM, and sleep apnea.