Strain Parameters Research Articles

The effects of myocardial bridging on two-dimensional myocardial strain during dobutamine stress echocardiography.

Myocardial bridging (MB) is a common anatomic variant in coronary arteries with unclear functional significance. We evaluated regional myocardial strain by speckle tracking during dobutamine stress echocardiography (DSE) in patients with MB in the left anterior descending coronary artery (LAD). We studied 11 patients with MB in the LAD and no obstructive coronary artery disease (CAD), 7 patients without MB, but obstructive CAD in the LAD, and 12 controls without MB or obstructive CAD. MB was defined as either > 1mm (superficial) or > 2mm (deep) intramyocardial course of the LAD in coronary CT angiography. Regional longitudinal, radial and circumferential strains and strain rates as well as post-systolic strain index (PSI) were measured at rest, peak stress, and early recovery (1min after stress). Strain parameters during DSE were similar in the myocardium distal to MB and other myocardial regions of the same patients as well as the LAD territory in controls. However, patients with obstructive CAD showed impaired LS and strain rate as well as increased PSI at peak stress. None of the MB was associated with systolic compression in invasive coronary angiography and strain parameters were similar between superficial and deep MB. Stress myocardial blood flow by positron emission tomography correlated with LS and RS at peak stress in the myocardium distal to MB (r = -0.73, p = 0.03, and r = 0.64, p = 0.04, respectively). Myocardial strain is not reduced during DSE in patients with MB in the LAD and no significant systolic compression.

Deep Convolutional Backbone Comparison for Automated PET Image Quality Assessment.

Pretraining deep convolutional network mappings using natural images helps with medical imaging analysis tasks; this is important given the limited number of clinically-annotated medical images. Many two-dimensional pretrained backbone networks, however, are currently available. This work compared 18 different backbones from 5 architecture groups (pretrained on ImageNet) for the task of assessing [18F]FDG brain Positron Emission Transmission (PET) image quality (reconstructed at seven simulated doses), based on three clinical image quality metrics (global quality rating, pattern recognition, and diagnostic confidence). Using two-dimensional randomly sampled patches, up to eight patients (at three dose levels each) were used for training, with three separate patient datasets used for testing. Each backbone was trained five times with the same training and validation sets, and with six cross-folds. Training only the final fully connected layer (with ~6,000-20,000 trainable parameters) achieved a test mean-absolute-error of ~0.5 (which was within the intrinsic uncertainty of clinical scoring). To compare "classical" and over-parameterized regimes, the pretrained weights of the last 40% of the network layers were then unfrozen. The mean-absolute-error fell below 0.5 for 14 out of the 18 backbones assessed, including two that previously failed to train. Generally, backbones with residual units (e.g. DenseNets and ResNetV2s), were suited to this task, in terms of achieving the lowest mean-absolute-error at test time (~0.45 - 0.5). This proof-of-concept study shows that over-parameterization may also be important for automated PET image quality assessments.

The Effects of Normobaric Hypoxia on the Acute Physiological Responses to Resistance Training: A Narrative Review.

Allsopp, GL, Britto, FA, Wright, CR, and Deldicque, L. The effects of normobaric hypoxia on the acute physiological responses to resistance training: a narrative review. J Strength Cond Res 38(11): 2001-2011, 2024-Athletes have used altitude training for many years as a strategy to improve endurance performance. The use of resistance training in simulated altitude (normobaric hypoxia) is a growing strategy that aims to improve the hypertrophy and strength adaptations to training. An increasing breadth of research has characterized the acute physiological responses to resistance training in hypoxia, often with the goal to elucidate the mechanisms by which hypoxia may improve the training adaptations. There is currently no consensus on the overall effectiveness of hypoxic resistance training for strength and hypertrophy adaptations, nor the underlying biochemical pathways involved. There are, however, numerous interesting physiological responses that are amplified by performing resistance training in hypoxia. These include potential changes to the energy system contribution to exercise and alterations to the level of metabolic stress, hormone and cytokine production, autonomic regulation, and other hypoxia-induced cellular pathways. This review describes the foundational exercise physiology underpinning the acute responses to resistance training in normobaric hypoxia, potential applications to clinical populations, including training considerations for athletic populations. The review also presents a summary of the ideal training parameters to promote metabolic stress and associated training adaptations. There are currently many gaps in our understanding of the physiological responses to hypoxic resistance training, partly caused by the infancy of the research field and diversity of hypoxic and training parameters.

Robust Symbol Detection Based on Quaternion Neural Networks in Wireless Polarization-Shift-Keying Communications.

Quaternion neural networks (QNNs) form a class of neural networks constructed with quaternion numbers. They are suitable for processing 3-D features with fewer trainable free parameters than real-valued neural networks (RVNNs). This article proposes symbol detection in wireless polarization-shift-keying (PolSK) communications by employing QNNs. We demonstrate that quaternion plays a crucial role in the symbol detection of PolSK signals. Existing artificial-intelligence communication studies mainly focus on RVNN-based symbol detection in digital modulations having constellations in complex plane. However, in PolSK, information symbols are represented as the state of polarization, which can be mapped on the Poincare sphere and thus its symbols have a 3-D data structure. Quaternion algebra offers a unified representation to process 3-D data with rotational invariance and, therefore, it keeps the internal relationship among three components of a PolSK symbol. Hence, we can expect that QNNs learn the distribution of received symbols on the Poincare sphere with higher consistency to detect the transmitted symbols more efficiently than RVNNs. We compare PolSK symbol detection accuracy of two types of QNNs, RVNN, existing methods such as least-square and minimum-mean-square-error channel estimations, as well as detection knowing perfect channel state information (CSI). Simulation results including symbol error rate show that the proposed QNNs outperform the existing estimation methods and that they reach better results with two to three times fewer free parameters than the RVNN. We find that QNN processing will bring practical use of PolSK communications.

Early Detection of Left Ventricular Dysfunction With Machine Learning-Based Strain Imaging in Aortic Stenosis Patients.

Aortic stenosis (AS) is a common cardiovascular condition where early detection of left ventricular (LV) dysfunction is essential for timely intervention and optimal management. Current echocardiographic measurements, such as ejection fraction (EF), are insensitive to minor changes in LV function, and strain imaging is typically limited to the global longitudinal strain (GLS) parameter due to robustness issues. This study introduces a novel, fully automatic algorithm to enhance the detection of LV dysfunction in AS patients using multiple strain imaging parameters. We applied supervised machine-learning techniques to classify data from 82 severe AS patients, 96 chest pain subjects, and 319 healthy volunteers. Our model significantly outperformed EF and GLS in distinguishing AS patients from healthy volunteers (area under the curve [AUC] = 0.97 vs. 0.88 and 0.82, respectively). It also surpassed EF and GLS in differentiating AS patients from chest pain subjects (AUC = 0.95 vs. 0.90 and 0.55, respectively). This novel, clinically interpretable model leverages the potential of strain imaging to enhance diagnostic accuracy and guide clinical decision-making in LV dysfunction, thereby improving clinical practice.

Clinical Prompt Learning With Frozen Language Models.

When the first transformer-based language models were published in the late 2010s, pretraining with general text and then fine-tuning the model on a task-specific dataset often achieved the state-of-the-art performance. However, more recent work suggests that for some tasks, directly prompting the pretrained model matches or surpasses fine-tuning in performance with few or no model parameter updates required. The use of prompts with language models for natural language processing (NLP) tasks is known as prompt learning. We investigated the viability of prompt learning on clinically meaningful decision tasks and directly compared this with more traditional fine-tuning methods. Results show that prompt learning methods were able to match or surpass the performance of traditional fine-tuning with up to 1000 times fewer trainable parameters, less training time, less training data, and lower computation resource requirements. We argue that these characteristics make prompt learning a very desirable alternative to traditional fine-tuning for clinical tasks, where the computational resources of public health providers are limited, and where data can often not be made available or not be used for fine-tuning due to patient privacy concerns. The complementary code to reproduce the experiments presented in this work can be found at https://github.com/NtaylorOX/Public_Clinical_Prompt.

Comprehensive analysis of recurrence factors in cryoballoon AF ablation: integrating clinical, biomarkers, and echocardiographic parameters.

Atrial fibrillation (AF) poses substantial challenges in cardiovascular diseases, impacting patient health and economic burdens. Understanding the mechanical effects of AF on the left atrium (LA) and assessing the influence of treatment modalities on LA functions are critical. This study aims to assess the efficacy of echocardiographic and biochemical parameters in predicting AF recurrence following second generation cryoballoon ablation (CB-2). Ninety-two patients with symptomatic AF, treated with CB-2 at Istanbul University-Cerrahpaşa, Faculty of Medicine, Department of Cardiology, were prospectively examined from January 2021 to July 2023. The study endeavors to develop a predictive model for AF recurrence, investigating the relationship between echocardiographic measurements and serum biomarkers with recurrence. The follow-up duration for echocardiographic assessments and biochemical analyses was systematically documented. The study revealed a significant enhancement in LA mechanical functions during echocardiographic follow-ups three months post-procedure. Specifically, LA strain parameters emerged as significant predictors of recurrence (LAsr: 95%CI 1.004-1.246, p = 0.047; LAsct: 95%CI 1.040-1.750, p = 0.024). Biochemical analyses demonstrated a correlation between elevated PRO-BNP levels and an increased risk of recurrence (95%CI 1.000-1.003, p = 0.012). Moreover, specific biomarkers such as MYBPHL, which demonstrated increased levels post-procedure, were deemed indicative of atrial damage, suggesting potential additional atrial substrate modification beyond PVI. Consequently, improvements in LA function post-cryoballoon ablation and biochemical markers have surfaced as potential indicators for predicting AF recurrence. This study elucidates the effectiveness of CB-2 in treating AF and its impact on LA functions. Notably, LA strain measurements and PRO-BNP levels have emerged as reliable indicators for predicting recurrence. Beyond clinical implications, our research establishes a foundation for a deeper understanding of the role of CB-2 in AF management and factors associated with recurrence.

Effect of substrate temperature on the physical and sensing properties of nanostructured Fe2O3 thin films

Using Chemical Bath Deposition (CBD) Method and various substrate temperatures, Fe2O3 films were successfully deposited. The produced film thickness was around (320 nm). Using X-ray diffraction, researchers may examine the polycrystalline structure of Fe2O3 thin films. These nanofilms contain strong peaks at 2θ =32.21, suggesting a preferred orientation along the (110) plane, and the grain size increases with substrate temperature, according to XRD tests. When the base temperature was raised from 350 to 450 o C, the strain parameter decreased from 31.35 to 28.43. AFM testing of the surface morphology of the deposition of material yields excellent homogenous coatings. The findings show that the average particle size of the nanoparticles ranges from (69.8 to 32.7) nm. SEM images show Fe2O3 films at (350, 400, 450) °C. Increased temperature reduces grain size, influencing morphology variations. The absorbance increases with substrate temperatures and decreases rapidly at short wavelengths, which correspond to the energy gap. The transmittance increases with increasing wavelength range. It decreases with rising substrate temperatures. The band gap values vary from 2.17 eV to 2.06 eV by increasing the substrate temperatures from 350 to 450 o C. It was discovered that the band gap reduces as the temperature of the Fe2O3 substrate increases. In addition, the optical constants for all films, including the absorption coefficient, the refractive index, and the extinction coefficient, were computed. Fe2O3 film's resistance over time at 350, 400, and 450°C for 300 ppm NO2 demonstrates oxidation effect and temperature sensitivity. Sensitivity decreases with higher base temperature due to charge carrier recombination, affecting NO2 response.

The Immediate Effects of a Standardized Kettlebell Swing Protocol on Lumbar Paraspinal Muscle Function: A Randomized Controlled Trial.

Hanney, WJ, Perez, A, Collado, G, Palmer, AC, Wilson, AT, Richardson, RM, and Kolber, MJ. J Strength Cond Res 38(11): 1854-1859, 2024-Kettlebell swings (KBSs) are commonly used to target the lumbar erector spinae and lower body musculature. This exercise exhibits distinct loading properties that requires cyclical contraction of the trunk extensors and posterior chain, potentially explaining its novel influence on muscle contractility. Tensiomyography (TMG) is a reliable, noninvasive, passive technique that may be used to examine muscular fatigue produced by exercises such as KBSs. The purpose of this randomized control trial was to determine the extent of muscle fatigue in the lumbar erector spinae musculature following the performance of a previously published high-intensity interval KBS protocol. Forty-one adults between the ages of 18 and 45 years were recruited. Inclusion criteria included subjects with no recent history of low back pain and clearance by the physical activity readiness questionnaire. Subjects were randomly allocated to either a KBS group ( n = 21) or a control group (CON; n = 20) who only performed the unloaded warm-up. Subjects were assessed at baseline, postintervention, and 24-hours postintervention for bilateral erector spinae fatigue, measured by 5 TMG parameters (Dm, Tc, Tr, Td, and Ts). The results were evaluated through a 2 × 3 (group × time) repeated-measures analysis of variance. The level of significance was set at p ≤ 0.05. There was no significant difference in lumbar erector spinae fatigue, measured by the 5 TMG parameters ( p ≥ 0.079), following the interval KBS protocol in comparison with the CON group at 3 assessment periods. A high-intensity interval KBS protocol failed to produce significant differences in erector spinae fatigue compared with the control group that did not perform a KBS. These findings warrant further investigation into muscle fatigue produced with higher intensity protocols and possibly suggest, depending on the programming goals, the need for an alternate KBS training parameters.

FINE VESSEL SEGMENTATION WITH REFINEMENT GATE IN DEEP LEARNING ARCHITECTURES

Automated vessel segmentation is essential in diagnosing eye-related disorders and monitoring progressive retinal diseases. State-of-the-art methods have achieved excellent results in this field, but very few have considered the post-processing of feature maps. As a result, there is often a lack of small and fine vessels or discontinuities in segmented vessels. To address this issue, this study introduces a novel post-processing technique called the refinement gate, which works with a deep learning model during training. The refinement gate enhances contextual information to extract important features from feature maps better. The proposed technique is applied with U-net architecture and placed after every convolution block in the encoder path. Visual and statistical comparisons demonstrate the robustness of the proposed method using three publicly available datasets, namely: the DRIVE DB, the STARE DB, and CHASE_DB1 datasets, showing significant improvements to segment weak and tiny vessels. The reported results confirm the potential of the model to be used as a segmentation tool in the medical field. This study is the first to propose such a gating mechanism without additional trainable parameters or standalone networks as in other literature.

Rapid likelihood free inference of compact binary coalescences using accelerated hardware

Abstract We report a gravitational-wave parameter estimation algorithm, AMPLFI, based on likelihood-free inference using normalizing flows. The focus of AMPLFI is to perform real-time parameter estimation for candidates detected by machine-learning based compact binary coalescence search, Aframe. We present details of our algorithm and optimizations done related to data-loading and pre-processing on accelerated hardware. We train our model using binary black-hole (BBH) simulations on real LIGO-Virgo detector noise. Our model has ∼ 6 million trainable parameters with training times ≲ 24 h. Based on online deployment on a mock data stream of LIGO-Virgo data, Aframe + AMPLFI is able to pick up BBH candidates and infer parameters for real-time alerts from data acquisition with a net latency of ∼ 6 s.

Solving transport equations on quantum computers—potential and limitations of physics-informed quantum circuits

AbstractQuantum circuits with trainable parameters, paired with classical optimization routines can be used as machine learning models. The recently popularized physics-informed neural network (PINN) approach is a machine learning algorithm that solves differential equations by incorporating them into a loss function. Being a mesh-free method, it is a promising approach for computational fluid dynamics. The question arises whether the properties of quantum circuits can be leveraged for a quantum physics-informed machine learning model. In this study, we compare the classical PINN-ansatz and its quantum analog, which we name the physics-informed quantum circuit (PIQC). The PIQC simulations are performed on a noise-free quantum computing simulator. Studying various differential equations, we compare expressivity, accuracy and convergence properties. We find that one-dimensional problems, such as the linear transport of a Gaussian-pulse or Burgers’ equation, allow a successful approximation with the classical and the quantum ansatz. For these examples, the PIQC overall performs similarly to PINN and converges more consistently and for Burgers’ equations even faster. While this is promising, the chosen quantum circuit approach struggles to approximate discontinuous solutions which the classical PINN-ansatz can represent. Based on this comparison, we extrapolate that the required number of qubits for solving two-dimensional problems in aerodynamics may already be available in the next few years. However, the acceleration potential is currently unclear and challenges like noisy circuits and approximations of discontinuous solutions have to be overcome.

Left atrial function in uraemic patients: Four-dimensional automatic left atrial quantitative technology study.

This study aimed to evaluate the utility of left atrial volume and function in uraemic patients using four-dimensional automatic left atrial quantification (4D auto LAQ) technology. Thirty-four undialysed uraemic patients (U-ND group), 60 dialysed uraemic patients (U-D group), and 32 healthy volunteers (N group) were enrolled in our current study. Conventional echocardiographic parameters were recorded, and left atrial volume and strain parameters were analysed to determine statistical differences among the three groups. The Pearson correlation coefficient was employed to assess the relationships between left atrial ejection fraction and left atrial strain parameters. Compared to the N group, uraemic patients often displayed left atrial enlargement and left ventricular hypertrophy. Significant increases were noted in left atrial diameter, interventricular septum thickness, left ventricular posterior wall thickness, E/e', diastolic blood pressure, systolic blood pressure, left atrial minimum volume, left atrial maximum volume, left atrial pre-atrial contraction volume, left atrial emptying volume and left atrial maximum volume index (P<0.05). Conversely, the e', E/A ratio and left atrial reservoir longitudinal strain were significantly decreased (P<0.05). However, no statistically significant differences were observed in the aforementioned parameters between the U-ND and U-D groups. The absolute values of left atrial conduit longitudinal strain and left atrial conduit circumferential strain, as well as left atrial passive ejection fraction, were notably lower in the U-D group compared to the N and U-ND groups, with statistically significant differences identified among the three groups (P<0.05). Uraemic patients exhibit marked left atrial enlargement and left ventricular hypertrophy, coupled with altered atrial function, particularly ductal dysfunction in the U-D group. The 4D auto LAQ technology proves advantageous in detecting these alterations, offering a promising tool for thorough cardiac assessment in this patient cohort.

LostNet: A smart way for lost and find.

The rapid population growth in urban areas has led to an increased frequency of lost and unclaimed items in public spaces such as public transportation, restaurants, and other venues. Services like Find My iPhone efficiently track lost electronic devices, but many valuable items remain unmonitored, resulting in delays in reclaiming lost and found items. This research presents a method to streamline the search process by comparing images of lost and recovered items provided by owners with photos taken when items are registered as lost and found. A photo matching network is proposed, integrating the transfer learning capabilities of MobileNetV2 with the Convolutional Block Attention Module (CBAM) and utilizing perceptual hashing algorithms for their simplicity and speed. An Internet framework based on the Spring Boot system supports the development of an online lost and found image identification system. The implementation achieves a testing accuracy of 96.8%, utilizing only 0.67 GFLOPs and 3.5M training parameters, thus enabling the recognition of images in real-world scenarios and operable on standard laptops.

Continuous and discrete decoding of overt speech with scalp electroencephalography (EEG).

Neurological disorders affecting speech production adversely impact quality of&#xD;life for over 7 million individuals in the US. Traditional speech interfaces like eyetracking&#xD;devices and P300 spellers are slow and unnatural for these patients. An&#xD;alternative solution, speech Brain-Computer Interfaces (BCIs), directly decodes speech&#xD;characteristics, offering a more natural communication mechanism. This research&#xD;explores the feasibility of decoding speech features using non-invasive EEG. Nine&#xD;neurologically intact participants were equipped with a 63-channel EEG system&#xD;with additional sensors to eliminate eye artifacts. Participants read aloud sentences&#xD;displayed on a screen selected for phonetic similarity to the English language. Deep&#xD;learning models, including Convolutional Neural Networks and Recurrent Neural&#xD;Networks with and without attention modules, were optimized with a focus on&#xD;minimizing trainable parameters and utilizing small input window sizes for real-time&#xD;application. These models were employed for discrete and continuous speech decoding&#xD;tasks, achieving statistically significant participant-independent decoding performance&#xD;for discrete classes and continuous characteristics of the produced audio signal. A&#xD;frequency sub-band analysis highlighted the significance of certain frequency bands&#xD;(delta, theta, and gamma) for decoding performance, and a perturbation analysis&#xD;was used to identify crucial channels. Assessed channel selection methods did not&#xD;significantly improve performance, suggesting a distributed representation of speech&#xD;information encoded in the EEG signals. Leave-One-Out training demonstrated&#xD;the feasibility of utilizing common speech neural correlates, reducing data collection&#xD;requirements from individual participants.

BASE: a web service for providing compound-protein binding affinity prediction datasets with reduced similarity bias.

Deep learning-based drug-target affinity (DTA) prediction methods have shown impressive performance, despite a high number of training parameters relative to the available data. Previous studies have highlighted the presence of dataset bias by suggesting that models trained solely on protein or ligand structures may perform similarly to those trained on complex structures. However, these studies did not propose solutions and focused solely on analyzing complex structure-based models. Even when ligands are excluded, protein-only models trained on complex structures still incorporate some ligand information at the binding sites. Therefore, it is unclear whether binding affinity can be accurately predicted using only compound or protein features due to potential dataset bias. In this study, we expanded our analysis to comprehensive databases and investigated dataset bias through compound and protein feature-based methods using multilayer perceptron models. We assessed the impact of this bias on current prediction models and proposed the binding affinity similarity explorer (BASE) web service, which provides bias-reduced datasets. By analyzing eight binding affinity databases using multilayer perceptron models, we confirmed a bias where the compound-protein binding affinity can be accurately predicted using compound features alone. This bias arises because most compounds show consistent binding affinities due to high sequence or functional similarity among their target proteins. Our Uniform Manifold Approximation and Projection analysis based on compound fingerprints further revealed that low and high variation compounds do not exhibit significant structural differences. This suggests that the primary factor driving the consistent binding affinities is protein similarity rather than compound structure. We addressed this bias by creating datasets with progressively reduced protein similarity between the training and test sets, observing significant changes in model performance. We developed the BASE web service to allow researchers to download and utilize these datasets. Feature importance analysis revealed that previous models heavily relied on protein features. However, using bias-reduced datasets increased the importance of compound and interaction features, enabling a more balanced extraction of key features. We propose the BASE web service, providing both the affinity prediction results of existing models and bias-reduced datasets. These resources contribute to the development of generalized and robust predictive models, enhancing the accuracy and reliability of DTA predictions in the drug discovery process. BASE is freely available online at https://synbi2024.kaist.ac.kr/base .

Enhanced PINNs with augmented Lagrangian method and transfer learning for hydrodynamic lubrication analysis

Purpose By seamlessly integrating physical laws, physics-informed neural networks (PINNs) have flexibly solved a wide variety of partial differential equations (PDEs). However, encoding PDEs and constraints as soft penalties in the loss function can cause gradient imbalances, leading to training and accuracy issues. This study aims to introduce the augmented Lagrangian method (ALM) and transfer learning to address these challenges and enhance the effectiveness of PINNs for hydrodynamic lubrication analysis. Design/methodology/approach The loss function was reformatted by ALM, adaptively adjusting the loss weights during training. Transfer learning was used to accelerate the convergence of PINNs under similar conditions. Additionally, the iterative process for load balancing was reframed as an inverse problem by extending film thickness as a trainable variable. Findings ALM-PINNs significantly reduced the maximum absolute boundary error by almost 80%. Transfer learning accelerated PINNs for solving the Reynolds equation, reducing training epochs by an order of magnitude. The iterative process for load balancing was effectively eliminated by extending the thickness as a trainable parameter, achieving a maximum percentage error of 2.31%. These outcomes demonstrated strong agreement with FDM results, analytical solutions and experimental data. Originality/value This study proposes a PINN-based approach for hydrodynamic lubrication analysis that significantly improves boundary accuracy and the training process. Additionally, it effectively replaces the load balancing procedure. This methodology demonstrates considerable potential for broader applications across various boundary value problems and iterative processes. Peer review The peer review history for this article is available at: https://publons.com/publon/10.1108/ILT-07-2024-0277/

Embedded System‐Based Malaria Detection From Blood Smear Images Using Lightweight Deep Learning Model

ABSTRACTThe disease of malaria, transmitted by female Anopheles mosquitoes, is highly contagious, resulting in numerous deaths across various regions. Microscopic examination of blood cells remains one of the most accurate methods for malaria diagnosis, but it is time‐consuming and can produce inaccurate results occasionally. Due to machine learning and deep learning advances in medical diagnosis, improved diagnostic accuracy can now be achieved while costs can be reduced compared to conventional microscopy methods. This work utilizes an open‐source dataset with 26 161 blood smear images in RGB for malaria detection. Our preprocessing resized the original dimensions of the images into 64 × 64 due to the limitations in computational complexity in developing embedded systems‐based malaria detection. We present a novel embedded system approach using 119 154 trainable parameters in a lightweight 17‐layer SqueezeNet model for the automatic detection of malaria. Incredibly, the model is only 1.72 MB in size. An evaluation of the model's performance on the original NIH malaria dataset shows that it has exceptional accuracy, precision, recall, and F1 scores of 96.37%, 95.67%, 97.21%, and 96.44%, respectively. Based on a modified dataset, the results improved further to 99.71% across all metrics. Compared to current deep learning models, our model significantly outperforms them for malaria detection, making it ideal for embedded systems. This model has also been rigorously tested on the Jetson Nano B01 edge device, demonstrating a rapid single image prediction time of only 0.24 s. The fusion of deep learning with embedded systems makes this research a crucial step toward improving malaria diagnosis. In resource‐constrained settings, the model's lightweight architecture and accuracy enhancements hold great promise for addressing the critical challenge of malaria detection.

Exploring the Impact of Architectural Variations in ResNet on CIFAR-100 Performance: An Investigation on Fully Connected Layers, Residual Blocks, and Kernel Size

ResNet is one of the leading neutral networks that has been widely applied in image classification. This study built a simple baseline network based on the concept of ResNet and then examines how variations in ResNet’s architecture affect its model performance to provide insights for optimizing network design. Firstly, this study investigates the number of fully connected layers, the results show that by reducing the number of fully connected layers significantly decreases the total number of trainable parameters, which in turn reduces the training time. However, this reduction does not lead to a noticeable improvement in the accuracy after convergence. In addition, increasing the number of fully connected layers not only greatly increases the training time but also leads to overfitting on the CIFAR-100 dataset, slightly reducing the training performance. Secondly, this study also analyzes the influence of reducing number of residual basic block. Analysis suggests that reducing the use of Residual Blocks has significantly negatively impacted both accuracy and training time. This may be because the use of Residual Blocks positively affects the network’s ability to learn features from the CIFAR-100 dataset. Finally, this study explores the effect of bigger kernel size of convolution layer in Residual Basic Block. The outcome demonstrates increasing the kernel size in the Residual Blocks significantly improves both training time and accuracy. Additionally, it was observed that in this variant experiment, a definitive convergence has not yet been clearly established, leaving the possibility that accuracy might continue to improve with more training epochs.

Review of Convolutional Neural Network

Over the past decade, the domain of deep learning has emerged as a swiftly expanding area of interest among researchers globally. It offers substantial benefits over traditional shallow networks, particularly in the realms of feature extraction and model fitting. Deep learning excels at uncovering intricate, distributed features from raw data, which exhibit robust generalization capabilities. It has triumphantly addressed challenges that were once deemed intractable within the field of artificial intelligence. With the exponential growth in the volume of training data and a marked increase in computational power, deep learning has achieved remarkable strides and found applications across a spectrum of domains, including but not limited to object detection, computer vision, natural language processing, speech recognition, and semantic parsing, thereby accelerating the evolution of artificial intelligence. Deep learning encompasses a series of non-linear transformations organized hierarchically, with deep neural networks representing the predominant form of contemporary deep learning methodologies. Inspired by the neural connections found in the animal visual cortex, these networks are characterized by local connectivity, shared weights, and pooling operations. Such attributes not only simplify the network model and curtail the number of trainable parameters but also engender invariance to shifts, distortions, and scaling, thereby endowing the model with enhanced robustness and fault tolerance. This makes the training and optimization of the network structure more manageable. This paper commences with a retrospective on the evolution of convolutional neural networks (CNNs). Subsequently, it delineates the architectures of neuron models and multilayer perceptron. It proceeds to dissect the CNN architecture, which typically encompasses a sequence of convolutional and pooling layers, culminating in fully connected layers, each serving distinct functions. The paper also elaborates on various enhanced algorithms for CNNs, such as the “Network in Network” and “Spatial Transformer Networks,” while also introducing supervised and unsupervised learning methodologies pertinent to CNNs and highlighting several prevalent open-source tools. Further, the paper scrutinizes the application of CNNs in various fields, including image classification, facial recognition, audio retrieval, electrocardiogram analysis, and object detection. It posits the amalgamation of CNNs with recurrent neural networks as a potential alternative for training datasets. The research culminates in the design of CNN structures with varying parameters and depths, supported by experimental analysis that elucidates the interplay among these parameters and the ramifications of their configuration. Ultimately, the paper synthesizes the merits and lingering challenges associated with CNNs and their practical applications.