Biomedical Physics & Engineering Express最新文献

Attention Deficit Hyperactivity Disorder (ADHD) is a highly prevalent childhood disorder, and related research has been increasing in recent years. However, it remains a challenging issue to accurately identify individuals with ADHD. The research proposes a method for ADHD detection using Recursive Feature Elimination-Genetic Algorithm (RFE-GA) for the feature selection of EEG data. Firstly, this study employed Transfer Entropy (TE) to construct brain networks from the EEG data of the ADHD and Normal groups, conducting an analysis of effective connectivity to unveil causal relationships in the brain's information exchange activities. Subsequently, a dual-layer feature selection method combining Recursive Feature Elimination (RFE) and Genetic Algorithm (GA) was proposed. Using the global search capability of GA and the feature selection ability of RFE, the performance of each feature subset is evaluated to find the optimal feature subset. Finally, a Support Vector Machine (SVM) classifier was employed to classify the ultimate feature set. The results revealed the control group exhibited lower connectivity strength in the left temporal alpha and beta bands, but higher frontal connectivity strength compared to the ADHD group. Additionally, in the gamma frequency band, the control group had higher top lobe connectivity strength than the ADHD group. Through the RFE-GA feature selection method, the optimized feature set was more concise, achieving classification accuracies of 91.3%, 94.1%, and 90.7% for the alpha, beta, and gamma frequency bands, respectively. The proposed RFE-GA feature selection method significantly reduced the number of features, thereby improving classification accuracy. .

Cervical cancer remains a major global health challenge, accounting for significant morbidity and mortality among women. Early detection through screening, such as Pap smear tests, is crucial for effective treatment and improved patient outcomes. However, traditional manual analysis of Pap smear images is labor-intensive, subject to human error, and requires extensive expertise. To address these challenges, automated approaches using deep learning techniques have been increasingly explored, offering the potential for enhanced diagnostic accuracy and efficiency. This research focuses on improving cervical cancer detection from Pap smear images using advanced deep-learning techniques. Specifically, we aim to enhance classification performance by leveraging Transfer Learning (TL) combined with an attention mechanism, supplemented by effective preprocessing techniques. Our preprocessing pipeline includes image normalization, resizing, and the application of Histogram of Oriented Gradients (HOG), all of which contribute to better feature extraction and improved model performance. The dataset used in this study is the Mendeley Liquid-Based Cytology (LBC) dataset, which provides a comprehensive collection of cervical cytology images annotated by expert cytopathologists. Initial experiments with the ResNet model on raw data yielded an accuracy of 63.95%. However, by applying our preprocessing techniques and integrating an attention mechanism, the accuracy of the ResNet model increased dramatically to 96.74%. Further, the Xception model, known for its superior feature extraction capabilities, achieved the best performance with an accuracy of 98.95%, along with high precision (0.97), recall (0.99), and F1-Score (0.98) on preprocessed data with an attention mechanism. These results underscore the effectiveness of combining preprocessing techniques, TL, and attention mechanisms to significantly enhance the performance of automated cervical cancer detection systems. Our findings demonstrate the potential of these advanced techniques to provide reliable, accurate, and efficient diagnostic tools, which could greatly benefit clinical practice and improve patient outcomes in cervical cancer screening.

Biological cells, characterized by complex and dynamic structures, demand precise models for comprehensive understanding, especially when subjected to external factors such as electric fields (EF) for manipulation or treatment. This interaction is integral to technologies like pulsed electric fields (PEF), inducing reversible and irreversible structural variations. Our study explores both simplified and sophisticated equivalent circuit models for biological cells under the influence of an external EF, covering diverse cell structures from single- to double-shell configurations. The paper highlights challenges in circuit modeling, specifically addressing the incorporation of reversible or irreversible pores in the membrane during external EF interactions, emphasizing the need for further research to refine technical aspects in this field. Additionally, we review a comparative analysis of the performance and applicability of the proposed circuit models, providing insights into their strengths and limitations. This contributes to a deeper insight of the complexities associated with modeling biological cells under external EF influences, paving the way for enhanced applications in medical and technological domains in future. .

Alzheimer's disease (AD) is a progressive neurological disorder. It is identified by the gradual shrinkage of the brain and the loss of brain cells. This leads to cognitive decline and impaired social functioning, making it a major contributor to dementia. While there are no treatments to reverse AD's progression, spotting the disease's onset can have a significant impact in the medical field. Deep learning (DL) has revolutionized medical image classification by automating feature engineering, removing the requirement for human experts in feature extraction. DL-based solutions are highly accurate but demand a lot of training data, which poses a common challenge. Transfer learning (TL) has gained attention for its knack for handling limited data and expediting model training. This study uses TL to classify AD using T1-weighted 3D Magnetic Resonance Imaging (MRI) from the Alzheimer's Disease Neuroimaging (ADNI) database. Four modified pre-trained deep neural networks (DNN), VGG16, MobileNet, DenseNet121, and NASNetMobile, are trained and evaluated on the ADNI dataset. The 3-1-3 weight averaging technique and fine-tuning improve the performance of the classification models. The evaluated accuracies for AD classification are VGG16: 98.75%; MobileNet: 97.5%; DenseNet: 97.5%; and NASNetMobile: 96.25%. The receiver operating characteristic (ROC), precision-recall (PR), and Kolmogorov-Smirnov (KS) statistic plots validate the effectiveness of the modified pre-trained model. Modified VGG16 excels with area under the curve (AUC) values of 0.99 for ROC and 0.998 for PR curves. The proposed approach shows effective AD classification by achieving high accuracy using the 3-1-3 weight averaging technique and fine-tuning.

Periodic discharges (PDs) are pathologic patterns of epileptiform discharges repeating at regular intervals, commonly detected in the human electroencephalogram (EEG) signals in patients who are critically ill. The frequency and spatial extent of PDs are associated with the tendency of PDs to cause brain injury, existing automated algorithms do not quantify the frequency and spatial extent of PDs. The present study presents an algorithm for quantifying frequency and spatial extent of PDs. The algorithm quantifies the evolution of these parameters within a short (10-14 second) window, with a focus on lateralized and generalized periodic discharges. We test our algorithm on 300 'easy', 300 'medium', and 240 'hard' examples (840 total epochs) of periodic discharges as quantified by interrater consensus from human experts when analyzing the given EEG epochs. We observe 95.0% agreement with a 95% confidence interval (CI) of [94.9%, 95.1%] between algorithm outputs with reviewer clincal judgement for easy examples, 92.0% agreement (95% CI [91.9%, 92.2%]) for medium examples, and 90.4% agreement (95% CI [90.3%, 90.6%]) for hard examples. The algorithm is also computationally efficient and is able to run in 0.385 ± 0.038 seconds for a single epoch using our provided implementation of the algorithm. The results demonstrate the algorithm's effectiveness in quantifying these discharges and provide a standardized and efficient approach for PD quantification as compared to existing manual approaches.

Purpose. Dual layer MLC (DMLC) has have been adopted in several commercial products and one major challenge in DMLC usage is leaf sequencing for intensity-modulated radiation therapy (IMRT). In this study we developed a leaf sequencing algorithm for IMRT with an orthogonal DMLC.Methods and Materials. This new algorithm is inspired by the algorithm proposed by Dai and Zhu for IMRT with single layer MLC (SMLC). It iterately determines a delivery segment intensity and corresponding segment shape for a given fluence matrix and leaves residual fluence matrix to following iterations. The segment intensity is determined according to complexities of residual fluence matrix when segment intensity varies from one to highest level in the matrix. The segment intensity and corresponding segment shape that result least complexity was selected. Although the algorithm framework is similar to Dai and Zhu's algorithm, this new algorithm develops complexity algorithms along with rules for determining segment leaf settings when delivered with orthogonal DMLC. This algorithm has been evaluated with 9 groups of randomly generated fluence matrices with various dimensions and intensity levels. Sixteen fluence matrices generated in Pinnacle system for two clinical IMRT examples were also used for evaluation. Statistical information of leaf sequences generated with this algorithm for both the random and clinical matrices were compared to the results of two typical algorithms for SMLC: that proposed by Dai and Zhu and that proposed by Bortfled.Results. Compared to the SMLC delivery sequences generated with Dai and Zhu's algorithm, the proposed algorithm for orthogonal DMLC delivery reduces the average number of segments by 27.7% ∼ 41.8% for 9 groups of randomly generated fluence matrices and 10.5% ∼ 41.7% for clinical ones. When comparing MU efficiency between different algorithms, it is observed that the proposed algorithm performs better than the optimal efficiency of SMLC delivery when dealing with simple fluence matrices, but slightly worse when handling complex ones.Conclusion. This new algorithm generates leaf sequences for orthogonal DMLC delivery with high delivery efficiency in terms of number of leaf segments. This algorithm has potential to work well with orthogonal DMLC for improving efficiency or quality of IMRT.

Low-frequency sonophoresis has emerged as a promising minimally invasive transdermal drug delivery method. However, effectively inducing cavitation on the skin surface with a compact, low-frequency ultrasound transducer poses a significant challenge. This paper presents a modified design of a low-frequency ultrasound transducer capable of generating ultrasound cavitation on the skin surfaces. The transducer comprises a piezoelectric ceramic disk and a bowl-shaped acoustic resonator. A conical slit structure was incorporated into the modified transducer design to amplify vibration displacement and enhance the maximum sound pressure. The FEM-based simulation results confirmed that the maximum sound pressure at the resonance frequency of 78 kHz was increased by 1.9 times that of the previous design. Ultrasound cavitation could be experimentally observed on the gel surface. Moreover, 3 min of ultrasound treatment significantly improved the caffeine permeability across an artificial membrane. These results demonstrated that this transducer holds promise for enhancing drug permeation by generating ultrasound cavitation on the skin surface.

Echocardiography is one the most commonly used imaging modalities for the diagnosis of congenital heart disease. Echocardiographic image analysis is crucial to obtaining accurate cardiac anatomy information. Semantic segmentation models can be used to precisely delimit the borders of the left ventricle, and allow an accurate and automatic identification of the region of interest, which can be extremely useful for cardiologists. In the field of computer vision, convolutional neural network (CNN) architectures remain dominant. Existing CNN approaches have proved highly efficient for the segmentation of various medical images over the past decade. However, these solutions usually struggle to capture long-range dependencies, especially when it comes to images with objects of different scales and complex structures. In this study, we present an efficient method for semantic segmentation of echocardiographic images that overcomes these challenges by leveraging the self-attention mechanism of the Transformer architecture. The proposed solution extracts long-range dependencies and efficiently processes objects at different scales, improving performance in a variety of tasks. We introduce Shifted Windows Transformer models (Swin Transformers), which encode both the content of anatomical structures and the relationship between them. Our solution combines the Swin Transformer and U-Net architectures, producing a U-shaped variant. The validation of the proposed method is performed with the EchoNet-Dynamic dataset used to train our model. The results show an accuracy of 0.97, a Dice coefficient of 0.87, and an Intersection over union (IoU) of 0.78. Swin Transformer models are promising for semantically segmenting echocardiographic images and may help assist cardiologists in automatically analyzing and measuring complex echocardiographic images.