Biomedical Engineering: Applications, Basis and Communications

Cancer is the second cause of mortality worldwide and it has been identified as a perilous disease. Among all types of cancers, Breast Cancer (BC) is a substantial research subject in the medical imaging area, because it is a serious disease and primary reason for death in women. Proper diagnosis helps patients to get adequate treatment, enhancing the probability of surviving. Because of the poor contrast and unclear structure of tumor cells in the images, automatic segmenting of breast tumors remains a difficult task. However, the identification and interpretation of breast lesions is challenging even for expert radiologists. To address these limitations, an efficient mechanism for BC detection and classification using histopathological images is proposed, which employs a DenseNet-based Chronological Circle Inspired Optimization Algorithm (CCIOA). Deep Learning (DL) approaches are used in the suggested BC classification scheme to precisely segment and identify the BC. The segmentation is done using ResuNet++, and an efficient optimization method called Invasive Water Ebola Optimization (IWEO) is used to fine-tune the DL network’s parameters. Furthermore, DenseNet is utilized for BC detection, while CCIOA is used for DenseNet training. The CCIOA-DenseNet is evaluated using the metrics of accuracy, True Positive Rate (TPR), and True Negative Rate (TNR). Experiment results show that the CCIOA-DenseNet attained better accuracy of 0.971, TPR of 0.966, and TNR of 0.954.

This research presents a method to select the best parameters of an extreme learning machine (ELM) for estimating grasping force from two-channel surface electromyography (sEMG) signals. Its advantages compared to the use of multi-channel sEMG signals include faster computing, lower electrode costs, and easier usage. The proposed method is appropriate for certain applications where time is very important but accuracy can be slightly compromised. The study recorded sEMG signals from 20 healthy volunteers, 10 males and 10 females, aged 22-52 years. The recorded sEMG signals were used in testing the proposed optimization method for grasping force estimation. It was found that the proposed method for optimizing the number of nodes, gain, and factor values would make the force estimation more efficient. The results show that the optimum number of nodes, gain, and factor values is 4, 0.05, and 1.0045, respectively. The resulting root mean square error and correlation coefficient values in force estimation from the ELM optimization method were 2.5642 and 0.9287, respectively, when the computing time was only 0.0038 s. These results show the feasibility of the proposed method for estimating force using only two-channel sEMG signals. As a result, real-time implementation, such as force estimation in robotic hand control for rehabilitation using sEMG signals, can be achieved efficiently.

Chronic Venous Insufficiency (CVI) is a venous incompetence condition that leads to improper blood circulation from the lower limbs towards the heart. This occurs as a result of blood pooling in the veins of the leg, resulting in twisted, dilated, and tortuous veins. Aging, obesity, prolonged standing or sitting, and lack of mobility are all important causes of the occurrence of this chronic disease. The cost of CVI diagnosis and treatment is extremely high. Infrared thermographic image analysis is used for early detection and reduces the cost of diagnosis. Deep learning (DL) techniques play an important role in early prediction and may aid clinicians in diagnosing CVI. An automated classification model will assist the physician in making a precise diagnosis of the abnormal vein and treating the patient according to the severity of the condition. There is a need for a model that can perform successful classification without the need for pre-processing when compared to the traditional machine learning (ML) methods that depend on ideal manual feature extraction to achieve optimal outcomes. In this research, we recommend the customized DenseNet-121 architecture for CVI detection and compare it with other advanced DL models to determine its efficacy. DenseNet-121 and other pre-trained convolutional neural network models, including EfficientNetB0 and Inception_v3, were trained using a transfer learning strategy. The experimental findings indicate that the proposed modified DenseNet-121 model outperformed other classical methods. The reported results provide evidence of the robustness of the suggested method in addition to the high accuracy that it possessed, as shown by the overall testing accuracy of 97.4%. Thus, this study can be considered as a non-invasive and cost-effective approach for diagnosing chronic venous insufficiency condition in lower extremity.

Lung sound (LS) signals are vital for diagnosing pulmonary disorders. However, heart sound (HS) interferes with the analysis of LS, leading to the misdiagnosis of lung disorders. To address this issue, we propose an Enhanced Variational Mode Decomposition (E-VMD) technique to remove HS interference from LSs effectively. The E-VMD method automatically determines the mode number for signal decomposition based on the characteristics of variational mode functions (VMFs) such as normalized permutation entropy, kurtosis index, extreme frequency domain, and energy loss coefficient. The performance of the proposed denoising technique was evaluated using six performance metrics: Signal-to-noise ratio (SNR), root mean square error (RMSE), normalized mean absolute error (nMAE), correlation coefficient factor (CCF), CPU${}_{\text{time}}$, and CPU${}_{\text{usage}}$. In comparison to other denoising methods such as empirical mode decomposition (EMD), ensemble empirical mode decomposition (EEMD), complementary ensemble empirical mode decomposition (CEEMD), singular spectrum analysis (SSA), and variational mode decomposition (VMD), the new E-VMD method demonstrates superior denoising outcome. The proposed method was evaluated using LS recorded from the outpatient department of Thoracic Medicine at Thanjavur Medical College and Hospital, Thanjavur. The obtained performance measures are as follows: RMSE: 0.02103 ± 0.00054, SNR: 28.52464 ± 0.00253, nMAE: 0.00009 ± 0.00056, CCF: 0.9962, CPU${}_{\text{usage}}$: 34.586, and CPU${}_{\text{time}}$: 0.452 s. These results affirm the adaptability and robustness of the proposed method, even in the existence of HS noise. This method improves denoising accuracy and computational efficacy, making it a useful tool for improving the analysis of LS signals and assisting in medical diagnostics. This technique utilizes an electronic stethoscope, a common clinical device used by healthcare professionals for detecting lung disease.

Epilepsy seizures are caused by abnormal, excessive, or synchronized neuronal activity in the brain, which is difficult to treat and is extremely stubborn. Therefore, studying the activity of epilepsy can greatly contribute to its diagnosis and treatment. The original signal is decomposed into IMFs and residual by ensemble empirical mode decomposition (EEMD), and then the first three intrinsic mode functions (IMF) are selected to replace the original signal, and the nonlinear and non-stationary problems of the original signal are solved. The Least Squares Support Vector Machine (LSSVM) was used as the classifier, its parameters (gam and sig2) are optimized by Particle Swarm Optimization (PSO). The experiment used the EEG database published by the University of Bonn (UoB) to realize the classification of normal, interictal and ictal periods. When PSO was employed, the recognition accuracy of the test set was 93.33%, with a classification time of 0.035 s and the Information Transfer Rate (ITR) of 3.77 bpm in training 70 classes with 100 samples each. In contrast, without PSO, the recognition accuracy of the test set was 92%, with a classification time of 0.039 s and the ITR of 2.88 bpm without PSO in training 70 classes with 100 samples each. The experimental results show that EEMD and LSSVM can effectively implement the three-classification problem and provide an effective means for the onset prediction of epilepsy patients.

Image de-noising is an essential tool for removing unwanted signals from an image. In Computed Tomography (CT) images, the image quality is degraded by the absorption of X-rays and quantum noise, which is generated due to the excitement of X-ray photons. Removal of noise and preservation of information in the CT images becomes a challenge for an imaging algorithm design. During the algorithm design selection of dataset is an important aspect for deducing results. The dataset used in this research comprises of 60 CT scan images of liver cancer archived from the arterial contrast enhanced phase. In this phase the cancer cells appear more intense as compared to the healthy liver tissue due to the absorption of contrast enhancing reagent. The experimentation for appropriate noise removal filter selection is done by testing the images using Mean, Median and Weiner Filters. The filter selected should give an image output which has minimal randomness, sharper boundaries and no blur. The de-noised image will provide a better visibility of the disease to the radiologist and physician. The performance parameters used for the assessment of various filters used in the study include visual assessment, entropy and signal to noise ratio (SNR) of the images. Median filter gives an accuracy of 96%, mean filter is 76.2% accurate with respect to original information and Weiner filters has an accuracy of 79.7%.

Magnetic Resonance Imaging (MRI) provides detailed information about soft tissues, which is essential for disease analysis. However, the presence of Rician noise in MR images introduces uncertainties that challenge medical practitioners during analysis. The objective of our research paper is to introduce an innovative dual-channel deep learning (DL) model designed to effectively denoize MR images. The methodology of this model integrates two distinct pathways, each equipped with unique normalization and activation techniques, facilitating the creation of a wide range of image features. Specifically, we employ Group Normalization in combination with Parametric Rectified Linear Units (PRELU) and Local Response Normalizations (LRN) alongside Scaled Exponential Linear Units (SELU) within both channels of our denoizing network. The outcomes of our proposed network exhibit clinical relevance, empowering medical professionals to conduct more efficient disease analysis. When evaluated by experienced radiologists, our results were deemed satisfactory. The network achieved a noteworthy improvement in performance metrics without requiring retraining. Specifically, there was a (6.65 ± 0.03)% enhancement in Peak Signal-to-Noise Ratio (PSNR) values and a (3.9 ± 0.02)% improvement in Structural Similarity Index (SSIM) values. Furthermore, when evaluated on the dataset on which the network was initially trained, the increase in PSNR and SSIM values was even more pronounced, with a (7.5 ± 0.02)% improvement in PSNR and a (4.3 ± 0.01)% enhancement in SSIM. Evaluation metrics, such as SSIM and PSNR, demonstrated a notable enhancement in the results obtained using our network. The statistical significance of our findings is evident, with p-values consistently less than 0.05 (p < 0.05). Importantly, our network demonstrates exceptional generalizability, as it performs remarkably well on different datasets without the need for retraining.

Atherosclerosis-generated coronary artery disease (CAD) causes the anomaly of autonomic activity. The assessment of autonomic balance under disease conditions is of clinical interest. Thus, this work aimed to evaluate the CAD using Poincare-derived stress score (SS) and sympathetic/sympathetic-parasympathetic (S/PS). A total of 60 male (50–55 years) volunteers including CAD ($N$=30) and control ($N$=30) participated in this work. Digital lead-II electrocardiogram (ECG) was recorded for 10 min in a supine position and filtered to remove noises. A total of 10 tachogram samples were computed from each subject for a 5 min ECG signal duration with a shift of 30 s on the processed signal using Acqknowledge 4.0 (Biopac Systems Inc., USA). The heart rate variability (HRV) parameters were computed from the obtained tachogram. The computed Poincare plot parameters were used to derive the SS and S/PS ratios. The regression model was employed to observe the correlation of SS and S/PS to the HRV parameters. The obtained results revealed reduced HRV with higher sympathetic and reduced parasympathetic activity under CAD than the control subjects. The autonomic dysfunction was observed with significantly higher values of SS and S/PS in CAD than in the control subjects. The S/PS was observed to be positively associated with LF/HF and SD2/SD1 parameters. While SS was found to be positively correlated to the LF and LF/HF parameters in CAD and control subjects. The obtained linear strong correlation of SS and S/PS with control and CAD subjects suggested the use of SS and S/PS as excellent sympathetic activity and sympathovagal balance markers.