Category: Nevin Manimala

Sensors (Basel). 2023 May 7;23(9):4543. doi: 10.3390/s23094543.

ABSTRACT

The paper was devoted to the application of saliency analysis methods in the performance analysis of deep neural networks used for the binary classification of brain tumours. We have presented the basic issues related to deep learning techniques. A significant challenge in using deep learning methods is the ability to explain the decision-making process of the network. To ensure accurate results, the deep network being used must undergo extensive training to produce high-quality predictions. There are various network architectures that differ in their properties and number of parameters. Consequently, an intriguing question is how these different networks arrive at similar or distinct decisions based on the same set of prerequisites. Therefore, three widely used deep convolutional networks have been discussed, such as VGG16, ResNet50 and EfficientNetB7, which were used as backbone models. We have customized the output layer of these pre-trained models with a softmax layer. In addition, an additional network has been described that was used to assess the saliency areas obtained. For each of the above networks, many tests have been performed using key metrics, including statistical evaluation of the impact of class activation mapping (CAM) and gradient-weighted class activation mapping (Grad-CAM) on network performance on a publicly available dataset of brain tumour X-ray images.

PMID:37177747 | DOI:10.3390/s23094543

Sensors (Basel). 2023 May 6;23(9):4522. doi: 10.3390/s23094522.

ABSTRACT

This work relates to the quality of the electrocardiogram (ECG) signal of an elderly person, transmitted using optical wireless links. The studied system uses infrared signals between an optical transmitter located on the person’s wrist and optical receivers placed on the ceiling. As the elderly person moves inside a room, the optical channel is time-varying, affecting the received ECG signal. To assess the ECG quality, we use specific signal quality indexes (SQIs), allowing the evaluation of the spectral and statistical characteristics of the signal. Our main contribution is studying how the SQIs behave according to the optical transmission performance and the studied context in order to determine the conditions required to obtain excellent quality indexes. The approach is based on the simulation of the whole chain, from the raw ECG to the extraction process after transmission until the evaluation of SQIs. This technique was developed considering optical channel modeling, including the mobility of the elderly. The obtained results show the potential of optical wireless communication technologies for reliable ECG monitoring in such a context. It has been observed that excellent ECG quality can be obtained with a minimum SNR of 11 dB for on-off keying modulation.

PMID:37177726 | DOI:10.3390/s23094522

Sensors (Basel). 2023 May 6;23(9):4518. doi: 10.3390/s23094518.

ABSTRACT

The relentless expansion of communications services and applications in 5G networks and their further projected growth bring the challenge of necessary spectrum scarcity, a challenge which might be overcome using the concept of cognitive radio. Furthermore, an extremely high number of low-power devices are introduced by the concept of the Internet of Things (IoT), which also requires efficient energy usage and practically applicable device powering. Motivated by these facts, in this paper, we analyze a wirelessly powered underlay cognitive system based on a realistic case in which statistical channel state information (CSI) is available. In the system considered, the primary and the cognitive networks share the same spectrum band under the constraint of an interference threshold and a maximal tolerable outage permitted by the primary user. To adopt the system model in realistic IoT application scenarios in which network nodes are mobile, we consider the randomly moving cognitive user receiver. For the analyzed system, we derive the closed-form expressions for the outage probability, the outage capacity, and the ergodic capacity. The obtained analytical results are corroborated by an independent simulation method.

PMID:37177722 | DOI:10.3390/s23094518

Sensors (Basel). 2023 May 5;23(9):4504. doi: 10.3390/s23094504.

ABSTRACT

The present manuscript introduces an investigation of the structural and functional changes in the optic nerve in patients undergoing glaucoma treatment by comparing optical coherence tomography (OCT) measurements and RETeval system parameters. For such a purpose, 140 eyes were examined at the Ophthalmology Clinic of the “Elpis” General Hospital of Athens between October 2022 and April 2023. A total of 59 out of 140 eyes were from patients with early glaucoma under treatment (case group), 63 were healthy eyes (control group) and 18 were excluded. The experimental measurements were statistically analyzed using the SPSS software package. The main outcomes are summarized below: (i) there was no statistical difference between the right and left eye for both groups, (ii) statistical differences were found between age interval subgroups (30-54 and 55-80 years old) for the control group, mainly for the time response part of the RETeval parameters. Such difference was not indicated by the OCT system, and (iii) a statistical difference occurred between the control and case group for both OCT (through the retinal nerve fiber layer-RNFL thickness) and the RETeval parameters (through the photopic negative response-PhNR). RNFL was found to be correlated to b-wave (ms) and W-ratio parameters. In conclusion, the PhNR obtained by the RETeval system could be a valuable supplementary tool for the objective examination of patients with early glaucoma.

PMID:37177707 | DOI:10.3390/s23094504

Sensors (Basel). 2023 May 5;23(9):4487. doi: 10.3390/s23094487.

ABSTRACT

Background: Near-infrared spectroscopy (NIRS) relative concentration signals contain ‘noise’ from physiological processes such as respiration and heart rate. Simultaneous assessment of NIRS and respiratory rate (RR) using a single sensor would facilitate a perfectly time-synced assessment of (cerebral) physiology. Our aim was to extract respiratory rate from cerebral NIRS intensity signals in neonates admitted to a neonatal intensive care unit (NICU). Methods: A novel algorithm, NRR (NIRS RR), is developed for extracting RR from NIRS signals recorded from critically ill neonates. In total, 19 measurements were recorded from ten neonates admitted to the NICU with a gestational age and birth weight of 38 ± 5 weeks and 3092 ± 990 g, respectively. We synchronously recorded NIRS and reference RR signals sampled at 100 Hz and 0.5 Hz, respectively. The performance of the NRR algorithm is assessed in terms of the agreement and linear correlation between the reference and extracted RRs, and it is compared statistically with that of two existing methods. Results: The NRR algorithm showed a mean error of 1.1 breaths per minute (BPM), a root mean square error of 3.8 BPM, and Bland-Altman limits of agreement of 6.7 BPM averaged over all measurements. In addition, a linear correlation of 84.5% (p < 0.01) was achieved between the reference and extracted RRs. The statistical analyses confirmed the significant (p < 0.05) outperformance of the NRR algorithm with respect to the existing methods. Conclusions: We showed the possibility of extracting RR from neonatal NIRS in an intensive care environment, which showed high correspondence with the reference RR recorded. Adding the NRR algorithm to a NIRS system provides the opportunity to record synchronously different physiological sources of information about cerebral perfusion and respiration by a single monitoring system. This allows for a concurrent integrated analysis of the impact of breathing (including apnea) on cerebral hemodynamics.

PMID:37177691 | DOI:10.3390/s23094487

Sensors (Basel). 2023 Apr 30;23(9):4422. doi: 10.3390/s23094422.

ABSTRACT

Gait and balance are an intricate interplay between the brain, nervous system, sensory organs, and musculoskeletal system. They are greatly influenced by the type of footwear, walking patterns, and surface. This exploratory study examines the effects of the Infinity Walk, pronation, and footwear conditions on brain effective connectivity patterns. A continuous-wave functional near-infrared spectroscopy device collected data from five healthy participants. A highly computationally efficient connectivity model based on the Grange causal relationship between the channels was applied to data to find the effective relationship between inter- and intra-hemispheric brain connectivity. Brain regions of interest (ROI) were less connected during the barefoot condition than during other complex walks. Conversely, the highest interconnectedness between ROI was observed while wearing flat insoles and medially wedged sandals, which is a relatively difficult type of footwear to walk in. No statistically significant (p-value <0.05) effect on connectivity patterns was observed during the corrected pronated posture. The regions designated as motoric, sensorimotor, and temporal became increasingly connected with difficult walking patterns and footwear conditions. The Infinity Walk causes effective bidirectional connections between ROI across all conditions and both hemispheres. Due to its repetitive pattern, the Infinity Walk is a good test method, particularly for neuro-rehabilitation and motoric learning experiments.

PMID:37177624 | DOI:10.3390/s23094422

Sensors (Basel). 2023 Apr 30;23(9):4404. doi: 10.3390/s23094404.

ABSTRACT

This work investigates the effectiveness of deep neural networks within the realm of battery charging. This is done by introducing an innovative control methodology that not only ensures safety and optimizes the charging current, but also substantially reduces the computational complexity with respect to traditional model-based approaches. In addition to their high computational costs, model-based approaches are also hindered by their need to accurately know the model parameters and the internal states of the battery, which are typically unmeasurable in a realistic scenario. In this regard, the deep learning-based methodology described in this work was been applied for the first time to the best of the authors’ knowledge, to scenarios where the battery’s internal states cannot be measured and an estimate of the battery’s parameters is unavailable. The reported results from the statistical validation of such a methodology underline the efficacy of this approach in approximating the optimal charging policy.

PMID:37177604 | DOI:10.3390/s23094404

Sensors (Basel). 2023 Apr 28;23(9):4357. doi: 10.3390/s23094357.

ABSTRACT

Fifth Generation (5G) signals using the millimeter wave (mmWave) spectrums are highly vulnerable to blockage due to rapid variations in channel link quality. This can cause the devices or User Equipment (UE) to suffer from connection failure. In a dual connectivity (DC) network, the channel’s intermittency issues were partially solved by maintaining the UE’s connectivity to primary (LTE advanced stations) and secondary (5G mmWave stations) simultaneously. Even though the dual-connected network performs excellently in maintaining connectivity, its performance drops significantly due to the inefficient handover from one 5G mmWave station to another. The situation worsens when UE travels a long distance in a highly dense obstacle environment, which requires multiple ineffective handovers that eventually lead to performance degradation. This research aimed to propose an Adaptive TTT Handover (ATH) mechanism that deals with unpredictable 5G mmWave wireless channel behaviors that are highly intermittent. An adaptive algorithm was developed to automatically adjust the handover control parameters, such as Time-to-Trigger (TTT), based on the current state of channel condition measured by the Signal-to-Interference-Noise Ratio (SINR). The developed algorithm was tested under a 5G mmWave statistical channel model to represent a time-varying channel matrix that includes fading and the Doppler effect. The performance of the proposed handover mechanism was analyzed and evaluated in terms of handover probability, latency, and throughput by using the Network Simulator 3 tool. The comparative simulation result shows that the proposed adaptive handover mechanism performs excellently compared to conventional handovers and other enhancement techniques.

PMID:37177560 | DOI:10.3390/s23094357

Sensors (Basel). 2023 Apr 27;23(9):4346. doi: 10.3390/s23094346.

ABSTRACT

The use of artificial intelligence (AI) to detect phishing emails is primarily dependent on large-scale centralized datasets, which has opened it up to a myriad of privacy, trust, and legal issues. Moreover, organizations have been loath to share emails, given the risk of leaking commercially sensitive information. Consequently, it has been difficult to obtain sufficient emails to train a global AI model efficiently. Accordingly, privacy-preserving distributed and collaborative machine learning, particularly federated learning (FL), is a desideratum. As it is already prevalent in the healthcare sector, questions remain regarding the effectiveness and efficacy of FL-based phishing detection within the context of multi-organization collaborations. To the best of our knowledge, the work herein was the first to investigate the use of FL in phishing email detection. This study focused on building upon a deep neural network model, particularly recurrent convolutional neural network (RNN) and bidirectional encoder representations from transformers (BERT), for phishing email detection. We analyzed the FL-entangled learning performance in various settings, including (i) a balanced and asymmetrical data distribution among organizations and (ii) scalability. Our results corroborated the comparable performance statistics of FL in phishing email detection to centralized learning for balanced datasets and low organizational counts. Moreover, we observed a variation in performance when increasing the organizational counts. For a fixed total email dataset, the global RNN-based model had a 1.8% accuracy decrease when the organizational counts were increased from 2 to 10. In contrast, BERT accuracy increased by 0.6% when increasing organizational counts from 2 to 5. However, if we increased the overall email dataset by introducing new organizations in the FL framework, the organizational level performance improved by achieving a faster convergence speed. In addition, FL suffered in its overall global model performance due to highly unstable outputs if the email dataset distribution was highly asymmetric.

PMID:37177549 | DOI:10.3390/s23094346

Sensors (Basel). 2023 Apr 26;23(9):4303. doi: 10.3390/s23094303.

ABSTRACT

Temperature-controlled closed-loop systems are vital to the transportation of produce. By maintaining specific transportation temperatures and adjusting to environmental factors, these systems delay decomposition. Wireless sensor networks (WSN) can be used to monitor the temperature levels at different locations within these transportation containers and provide feedback to these systems. However, there are a range of unique challenges in WSN implementations, such as the cost of the hardware, implementation difficulties, and the general ruggedness of the environment. This paper presents the novel results of a real-life application, where a sensor network was implemented to monitor the environmental temperatures at different locations inside commercial temperature-controlled shipping containers. The possibility of predicting one or more locations inside the container in the absence or breakdown of a logger placed in that location is explored using combinatorial input-output settings. A total of 1016 machine learning (ML) models are exhaustively trained, tested, and validated in search of the best model and the best combinations to produce a higher prediction result. The statistical correlations between different loggers and logger combinations are studied to identify a systematic approach to finding the optimal setting and placement of loggers under a cost constraint. Our findings suggest that even under different and incrementally higher cost constraints, one can use empirical approaches such as neural networks to predict temperature variations in a location with an absent or failed logger, within a margin of error comparable to the manufacturer-specified sensor accuracy. In fact, the median test accuracy is 1.02 degrees Fahrenheit when using only a single sensor to predict the remaining locations under the assumptions of critical system failure, and drops to as little as 0.8 and 0.65 degrees Fahrenheit when using one or three more sensors in the prediction algorithm. We also demonstrate that, by using correlation coefficients and time series similarity measurements, one can identify the optimal input-output pairs for the prediction algorithm reliably under most instances. For example, discrete time warping can be used to select the best location to place the sensors with a 92% match between the lowest prediction error and the highest similarity sensor with the rest of the group. The findings of this research can be used for power management in sensor batteries, especially for long transportation routes, by alternating standby modes where the temperature data for the OFF sensors are predicted by the ON sensors.

PMID:37177507 | DOI:10.3390/s23094303