Learning manufacturing computer vision systems using tiny YOLOv4.

Implementing and deploying advanced technologies are principal in improving manufacturing processes, signifying a transformative stride in the industrial sector. Computer vision plays a crucial innovation role during this technological advancement, demonstrating broad applicability and profound impact across various industrial operations. This pivotal technology is not merely an additive enhancement but a revolutionary approach that redefines quality control, automation, and operational efficiency parameters in manufacturing landscapes. By integrating computer vision, industries are positioned to optimize their current processes significantly and spearhead innovations that could set new standards for future industrial endeavors. However, the integration of computer vision in these contexts necessitates comprehensive training programs for operators, given this advanced system's complexity and abstract nature. Historically, training modalities have grappled with the complexities of understanding concepts as advanced as computer vision. Despite these challenges, computer vision has recently surged to the forefront across various disciplines, attributed to its versatility and superior performance, often matching or exceeding the capabilities of other established technologies. Nonetheless, there is a noticeable knowledge gap among students, particularly in comprehending the application of Artificial Intelligence (AI) within Computer Vision. This disconnect underscores the need for an educational paradigm transcending traditional theoretical instruction. Cultivating a more practical understanding of the symbiotic relationship between AI and computer vision is essential. To address this, the current work proposes a project-based instructional approach to bridge the educational divide. This methodology will enable students to engage directly with the practical aspects of computer vision applications within AI. By guiding students through a hands-on project, they will learn how to effectively utilize a dataset, train an object detection model, and implement it within a microcomputer infrastructure. This immersive experience is intended to bolster theoretical knowledge and provide a practical understanding of deploying AI techniques within computer vision. The main goal is to equip students with a robust skill set that translates into practical acumen, preparing a competent workforce to navigate and innovate in the complex landscape of Industry 4.0. This approach emphasizes the criticality of adapting educational strategies to meet the evolving demands of advanced technological infrastructures. It ensures that emerging professionals are adept at harnessing the potential of transformative tools like computer vision in industrial settings.

Harnessing the potential of shared data in a secure, inclusive, and resilient manner via multi-key homomorphic encryption.

In this manuscript, we develop a multi-party framework tailored for multiple data contributors seeking machine learning insights from combined data sources. Grounded in statistical learning principles, we introduce the Multi-Key Homomorphic Encryption Logistic Regression (MK-HELR) algorithm, designed to execute logistic regression on encrypted multi-party data. Given that models built on aggregated datasets often demonstrate superior generalization capabilities, our approach offers data contributors the collective strength of shared data while ensuring their original data remains private due to encryption. Apart from facilitating logistic regression on combined encrypted data from diverse sources, this algorithm creates a collaborative learning environment with dynamic membership. Notably, it can seamlessly incorporate new participants during the learning process, addressing the key limitation of prior methods that demanded a predetermined number of contributors to be set before the learning process begins. This flexibility is crucial in real-world scenarios, accommodating varying data contribution timelines and unanticipated fluctuations in participant numbers, due to additions and departures. Using the AI4I public predictive maintenance dataset, we demonstrate the MK-HELR algorithm, setting the stage for further research in secure, dynamic, and collaborative multi-party learning scenarios.

3D Ultrasound Shear Wave Elastography for Musculoskeletal Tissue Assessment Under Compressive Load: A Feasibility Study.

Given its real-time capability to quantify mechanical tissue properties, ultrasound shear wave elastography holds significant promise in clinical musculoskeletal imaging. However, existing shear wave elastography methods fall short in enabling full-limb analysis of 3D anatomical structures under diverse loading conditions, and may introduce measurement bias due to sonographer-applied force on the transducer. These limitations pose numerous challenges, particularly for 3D computational biomechanical tissue modeling in areas like prosthetic socket design. In this feasibility study, a clinical linear ultrasound transducer system with integrated shear wave elastography capabilities was utilized to scan both a calibrated phantom and human limbs in a water tank imaging setup. By conducting 2D and 3D scans under varying compressive loads, this study demonstrates the feasibility of volumetric ultrasound shear wave elastography of human limbs. Our preliminary results showcase a potential method for evaluating 3D spatially varying tissue properties, offering future extensions to computational biomechanical modeling of tissue for various clinical scenarios.

Assessment of respiratory rate monitoring in the emergency department.

Objectives: This study aimed to compare the different respiratory rate (RR) monitoring methods used in the emergency department (ED): manual documentation, telemetry, and capnography. Methods: This is a retrospective study using recorded patient monitoring data. The study population includes patients who presented to a tertiary care ED between January 2020 and December 2022. Inclusion and exclusion criteria were patients with simultaneous recorded RR data from all three methods and less than 10 min of recording, respectively. Linear regression and Bland-Altman analysis were performed between different methods. Results: A total of 351 patient encounters met study criteria. Linear regression yielded an R-value of 0.06 (95% confidence interval [CI] 0.00-0.12) between manual documentation and telemetry, 0.07 (95% CI 0.01-0.13) between manual documentation and capnography, and 0.82 (95% CI 0.79-0.85) between telemetry and capnography. The Bland-Altman analysis yielded a bias of -0.8 (95% limits of agreement [LOA] -12.2 to 10.6) between manual documentation and telemetry, bias of -0.6 (95% LOA -13.5 to 12.3) between manual documentation and capnography, and bias of 0.2 (95% LOA -6.2 to 6.6) between telemetry and capnography. Conclusion: There is a poor correlation between manual documentation and both automated methods, while there is relatively good agreement between the automated methods. This finding highlights the need to further investigate the methodology used by the ED staff in monitoring and documenting RR and ways to improve its reliability given that many important clinical decisions are made based on these assessments.

Developing a virtual reality and AI-based framework for advanced digital manufacturing and nearshoring opportunities in Mexico.

The growing expansion of the manufacturing sector, particularly in Mexico, has revealed a spectrum of nearshoring opportunities yet is paralleled by a discernible void in educational tools for various stakeholders, such as engineers, students, and decision-makers. This paper introduces a state-of-the-art framework, incorporating virtual reality (VR) and artificial intelligence (AI) to metamorphose the pedagogy of advanced manufacturing systems. Through a case study focused on the design, production, and evaluation of a robotic platform, the framework endeavors to offer an exhaustive educational experience via an interactive VR environment, encapsulating (1) Robotic platform system design and modeling, enabling users to immerse themselves in the design and simulation of robotic platforms under varied conditions; (2) Virtual manufacturing company, presenting a detailed virtual manufacturing setup to enhance users' comprehension of manufacturing processes and systems, and problem-solving in realistic settings; and (3) Product evaluation, wherein users employ VR to meticulously assess the robotic platform, ensuring optimal functionality and customer satisfaction. This innovative framework melds theoretical acumen with practical application in advanced manufacturing, preparing entities to navigate Mexico's manufacturing sector's vibrant and competitive nearshoring landscape. It creates an immersive environment for understanding modern manufacturing challenges, fostering Mexico's manufacturing sector growth, and maximizing nearshoring opportunities for stakeholders.

Can small bowel obstruction during pregnancy be treated with conservative management? A review.

BACKGROUND: Small bowel obstruction can occur during pregnancy, which, if missed, can lead to dire consequences for both the mother and foetus. Management of this condition usually requires surgical intervention. However, only a small number of patients are treated conservatively. OBJECTIVE: The objective was to review the literature to determine the feasibility of conservative management for small bowel obstruction. METHODS: A systematic search of the PubMed and Embase databases was performed using the keywords [small bowel obstruction AND pregnancy]. All original articles were then reviewed and included in this review if deemed suitable. CONCLUSION: Conservative management of small bowel obstruction in pregnant women is feasible if the patient is clinically stable and after ruling out bowel ischaemia and closed-loop obstruction.

Computing Speed-of-Sound From Ultrasound: User-Agnostic Recovery and a New Benchmark.

OBJECTIVE: Medical ultrasound is one of the most accessible imaging modalities, but is a challenging modality for quantitative parameters comparison across vendors and sonographers. B-Mode imaging, with limited exceptions, provides a map of tissue boundaries; crucially, it does not provide diagnostically relevant physical quantities of the interior of organ domains.This can be remedied: the raw ultrasound signal carries significantly more information than is present in the B-Mode image. Specifically, the ability to recover speed-of-sound and attenuation maps from the raw ultrasound signal transforms the modality into a tissue-property modality. Deep learning was shown to be a viable tool for recovering speed-of-sound maps. A major hold-back towards deployment is the domain transfer problem, i.e., generalizing from simulations to real data. This is due in part to dependence on the (hard-to-calibrate) system response. METHODS: We explore a remedy to the problem of operator-dependent effects on the system response by introducing a novel approach utilizing the phase information of the IQ demodulated signal. RESULTS: We show that the IQ-phase information effectively decouples the operator-dependent system response from the data, significantly improving the stability of speed-of-sound recovery. We also introduce an improvement to the network topology providing faster and improved results to the state-of-the-art. We present the first publicly available benchmark for this problem: a simulated dataset for raw ultrasound plane wave processing. CONCLUSION: The consideration of the phase of the IQ-signals presents a promising appeal to traversing the transfer learning problem, advancing the goal of real-time speed-of-sound imaging.

Visual inertial odometry enabled 3D ultrasound and photoacoustic imaging.

There is an increasing need for 3D ultrasound and photoacoustic (USPA) imaging technology for real-time monitoring of dynamic changes in vasculature or molecular markers in various malignancies. Current 3D USPA systems utilize expensive 3D transducer arrays, mechanical arms or limited-range linear stages to reconstruct the 3D volume of the object being imaged. In this study, we developed, characterized, and demonstrated an economical, portable, and clinically translatable handheld device for 3D USPA imaging. An off-the-shelf, low-cost visual odometry system (the Intel RealSense T265 camera equipped with simultaneous localization and mapping technology) to track free hand movements during imaging was attached to the USPA transducer. Specifically, we integrated the T265 camera into a commercially available USPA imaging probe to acquire 3D images and compared it to the reconstructed 3D volume acquired using a linear stage (ground truth). We were able to reliably detect 500 µm step sizes with 90.46% accuracy. Various users evaluated the potential of handheld scanning, and the volume calculated from the motion-compensated image was not significantly different from the ground truth. Overall, our results, for the first time, established the use of an off-the-shelf and low-cost visual odometry system for freehand 3D USPA imaging that can be seamlessly integrated into several photoacoustic imaging systems for various clinical applications.

A comparison of point-tracking algorithms in ultrasound videos from the upper limb.

Tracking points in ultrasound (US) videos can be especially useful to characterize tissues in motion. Tracking algorithms that analyze successive video frames, such as variations of Optical Flow and Lucas-Kanade (LK), exploit frame-to-frame temporal information to track regions of interest. In contrast, convolutional neural-network (CNN) models process each video frame independently of neighboring frames. In this paper, we show that frame-to-frame trackers accumulate error over time. We propose three interpolation-like methods to combat error accumulation and show that all three methods reduce tracking errors in frame-to-frame trackers. On the neural-network end, we show that a CNN-based tracker, DeepLabCut (DLC), outperforms all four frame-to-frame trackers when tracking tissues in motion. DLC is more accurate than the frame-to-frame trackers and less sensitive to variations in types of tissue movement. The only caveat found with DLC comes from its non-temporal tracking strategy, leading to jitter between consecutive frames. Overall, when tracking points in videos of moving tissue, we recommend using DLC when prioritizing accuracy and robustness across movements in videos, and using LK with the proposed error-correction methods for small movements when tracking jitter is unacceptable.

Central venous pressure estimation with force-coupled ultrasound of the internal jugular vein.

We estimate central venous pressure (CVP) with force-coupled ultrasound imaging of the internal jugular vein (IJV). We acquire ultrasound images while measuring force applied over the IJV by the ultrasound probe imaging surface. We record collapse force, the force required to completely occlude the vein, in 27 healthy subjects. We find supine collapse force and jugular venous pulsation height (JVP), the clinical noninvasive standard, have a linear correlation coefficient of r2 = 0.89 and an average absolute difference of 0.23 mmHg when estimating CVP. We perturb our estimate negatively by tilting 16 degrees above supine and observe decreases in collapse force for every subject which are predictable from our CVP estimates. We perturb venous pressure positively to values experienced in decompensated heart failure by having subjects perform the Valsalva maneuver while the IJV is being collapsed and observe an increase in collapse force for every subject. Finally, we derive a CVP waveform with an inverse three-dimensional finite element optimization that uses supine collapse force and segmented force-coupled ultrasound data at approximately constant force.

Constitutive parameter identification of transtibial residual limb soft tissue using ultrasound indentation and shear wave elastography.

Finite element analysis (FEA) can be used to evaluate applied interface pressures and internal tissue strains for computational prosthetic socket design. This type of framework requires realistic patient-specific limb geometry and constitutive properties. In recent studies, indentations and inverse FEA with MRI-derived 3D patient geometries were used for constitutive parameter identification. However, long computational times and use of specialized equipment presents challenges for clinical, deployment. In this study, we present a novel approach for constitutive parameter identification using a combination of FEA, ultrasound indentation, and shear wave elastography. Local shear modulus measurement using elastography during an ultrasound indentation experiment has particular significance for biomechanical modeling of the residual limb since there are known regional dependencies of soft tissue properties such as varying levels of scarring and atrophy. Beyond prosthesis design, this work has broader implications to the fields of muscle health and monitoring of disease progression.

Comparison of Quantitative Ultrasound Methods to Classify Dystrophic and Obese Models of Skeletal Muscle.

In this study, we compared multiple quantitative ultrasound metrics for the purpose of differentiating muscle in 20 healthy, 10 dystrophic and 10 obese mice. High-frequency ultrasound scans were acquired on dystrophic (D2-mdx), obese (db/db) and control mouse hindlimbs. A total of 248 image features were extracted from each scan, using brightness-mode statistics, Canny edge detection metrics, Haralick features, envelope statistics and radiofrequency statistics. Naïve Bayes and other classifiers were trained on single and pairs of features. The a parameter from the Homodyned K distribution at 40 MHz achieved the best univariate classification (accuracy = 85.3%). Maximum classification accuracy of 97.7% was achieved using a logistic regression classifier on the feature pair of a2 (K distribution) at 30 MHz and brightness-mode variance at 40MHz. Dystrophic and obese mice have muscle with distinct acoustic properties and can be classified to a high level of accuracy using a combination of multiple features.

Automated Force-Coupled Ultrasound Method for Calibration-Free Carotid Artery Blood Pressure Estimation.

We develop, automate and evaluate a calibration-free technique to estimate human carotid artery blood pressure from force-coupled ultrasound images. After acquiring images and force, we use peak detection to align the raw force signal with an optical flow signal derived from the images. A trained convolutional neural network selects a seed point within the carotid in a single image. We then employ a region-growing algorithm to segment and track the carotid in subsequent images. A finite-element deformation model is fit to the observed segmentation and force via a two-stage iterative non-linear optimization. The first-stage optimization estimates carotid artery wall stiffness parameters along with systolic and diastolic carotid pressures. The second-stage optimization takes the output parameters from the first optimization and estimates the carotid blood pressure waveform. Diastolic and systolic measurements are compared with those of an oscillometric brachial blood pressure cuff. In 20 participants, average absolute diastolic and systolic errors are 6.2 and 5.6 mm Hg, respectively, and correlation coefficients are r = 0.7 and r = 0.8, respectively. Force-coupled ultrasound imaging represents an automated, standalone ultrasound-based technique for carotid blood pressure estimation, which motivates its further development and expansion of its applications.

Respiratory Inductance Plethysmography to Assess Fatigability during Repetitive Work.

Cumulative fatigue during repetitive work is associated with occupational risk and productivity reduction. Usually, subjective measures or muscle activity are used for a cumulative evaluation; however, Industry 4.0 wearables allow overcoming the challenges observed in those methods. Thus, the aim of this study is to analyze alterations in respiratory inductance plethysmography (RIP) to measure the asynchrony between thorax and abdomen walls during repetitive work and its relationship with local fatigue. A total of 22 healthy participants (age: 27.0 ± 8.3 yrs; height: 1.72 ± 0.09 m; mass: 63.4 ± 12.9 kg) were recruited to perform a task that includes grabbing, moving, and placing a box in an upper and lower shelf. This task was repeated for 10 min in three trials with a fatigue protocol between them. Significant main effects were found from Baseline trial to the Fatigue trials (p < 0.001) for both RIP correlation and phase synchrony. Similar results were found for the activation amplitude of agonist muscle (p < 0.001), and to the muscle acting mainly as a joint stabilizer (p < 0.001). The latter showed a significant effect in predicting both RIP correlation and phase synchronization. Both RIP correlation and phase synchronization can be used for an overall fatigue assessment during repetitive work.

Quantitative Sound Speed Imaging of Cortical Bone and Soft Tissue: Results From Observational Data Sets.

This work presents the first quantitative ultrasonic sound speed images of ex vivo limb cross-sections containing both soft tissue and bone using Full Waveform Inversion (FWI) with level set (LS) and travel time regularization. The estimated bulk sound speed of bone and soft tissue are within 10% and 1%, respectively, of ground truth estimates. The sound speed imagery shows muscle, connective tissue and bone features. Typically, ultrasound tomography (UST) using FWI is applied to imaging breast tissue properties (e.g. sound speed and density) that correlate with cancer. With further development, UST systems have the potential to deliver volumetric operator independent tissue property images of limbs with non-ionizing and portable hardware platforms. This work addresses the algorithmic challenges of imaging the sound speed of bone and soft tissue by combining FWI with LS regularization and travel time methods to recover soft tissue and bone sound speed with improved accuracy and reduced soft tissue artifacts when compared to conventional FWI. The value of leveraging LS and travel time methods is realized by evidence of improved bone geometry estimates as well as promising convergence properties and reduced risk of final model errors due to un-modeled shear wave propagation. Ex vivo bulk measurements of sound speed and MRI cross-sections validates the final inversion results.

Indoor Human Localization and Gait Analysis using Machine Learning for In-home Health Monitoring.

Homes equipped with ambient sensors can measure physiological signals correlated with the resident's health without requiring a wearable device. Gait characteristics may reveal physical imbalances or recognize changes in cognitive health. In this paper, we use the physical interactions with floor to both localize the resident and monitor their gait. Accelerometers are placed at the corners of the room for sensing. Gradient boosting regression was used to perform localization with an accuracy of 82%, reasonably accounting for inhomogeneity in the floor with just 3 sensors. A method using step time variance is proposed to detect gait imbalances; results on induced limps are presented.

Inverse design of photonic meta-structure for beam collimation in on-chip sensing.

Designed or patterned structured surfaces, metasurfaces, enable the miniaturization of complex arrangements of optical elements on a plane. Most of the existing literature focuses on miniaturizing the optical detection; little attention is directed to on-chip optical excitation. In this work, we design a metasurface to create a planar integrated photonic source beam collimator for use in on-chip optofluidic sensing applications. We use an iterative inverse design approach in order to optimize the metasurface to achieve a target performance using gradient descent method. We then fabricate beam collimators and experimentally compare performance characteristics with conventional uniform binary grating-based photonic beam diffractors. The optimal design enhances the illumination power by a factor of 5. The reinforced beam is more uniform with 3 dB beam spot increased almost ~ 3 times for the same device footprint area. The design approach will be useful in on-chip applications of fluorescence imaging, Raman, and IR spectroscopy and will enable better multiplexing of light sources for high throughput biosensing.

2D speed of sound mapping using a multilook reflection ultrasound tomography framework.

Quantitative ultrasound (QUS) has emerged as a viable tool in diagnosing and staging the onset and progression of various diseases. Within the field of QUS, shear wave elastography (SWE) has emerged as the clinical standard for quantifying and correlating the stiffness of tissue to its underlying pathology. Despite its widespread use, SWE suffers from drawbacks that limit its widespread clinical use; among these are low-frame rates, long settling times, and high sensitivity to operating conditions. Longitudinal speed of sound (SOS) has emerged as a viable alternative to SWE. We propose a framework to obtain 2D sound speed maps using a commercial ultrasound probe. A commercial ultrasound probe is localized in space and used to scan a domain of interest from multiple vantage points; the use of a reflector at the far end of the domain allows us to measure the round trip travel times to and from it. The known locations of the probe and the measured travel times are used to estimate the depth and inclination of the reflector as well as the unknown sound speed map. The use of multiple looks increases the effective aperture of the ultrasound probe and allows for a higher fidelity reconstruction of sound speed maps. We validate the framework using simulated and experimental data and propose a rigorous framework to quantify the uncertainty of the estimated sound speed maps.

Variation of Shear Wave Elastography With Preload in the Thyroid: Quantitative Validation.

OBJECTIVES: Thyroid shear wave elastography (SWE) has been shown to have advantages compared to biopsy or other imaging modalities in the evaluation of thyroid nodules. However, studies show variability in its assessment. The objective of this study was to evaluate whether stiffness measurements of the normal thyroid, as estimated by SWE, varied due to preload force or the pressure applied between the transducer and the patient. METHODS: In this study, a measurement system was attached to the ultrasound transducer to measure the applied load. Shear wave elastographic measurements were obtained from the left lobe of the thyroid at applied transducer forces between 2 and 10 N. A linear mixed-effects model was constructed to quantify the association between the preload force and stiffness while accounting for correlations between repeated measurements within each participant. The preload force effect on elasticity was modeled by both linear and quadratic terms to account for a possible nonlinear association between these variables. RESULTS: Nineteen healthy volunteers without known thyroid disease participated in the study. The participants had a mean age ± SD of 36 ± 8 years; 74% were female; 74% had a normal body mass index; and 95% were white non-Hispanic/Latino. The estimated elastographic value at a 2-N preload force was 16.7 kPa (95% confidence interval, 14.1-19.3 kPa), whereas the value at 10 N was 29.9 kPa (95% confidence interval, 24.9-34.9 kPa). CONCLUSIONS: The preload force was significantly and nonlinearly associated with SWE estimates of thyroid stiffness. Quantitative standardization of preload forces in the assessment of thyroid nodules using elastography is an integral factor for improving the accuracy of thyroid nodule evaluation.

Design of optical meta-structures with applications to beam engineering using deep learning.

Nanophotonics is a rapidly emerging field in which complex on-chip components are required to manipulate light waves. The design space of on-chip nanophotonic components, such as an optical meta surface which uses sub-wavelength meta-atoms, is often a high dimensional one. As such conventional optimization methods fail to capture the global optimum within the feasible search space. In this manuscript, we explore a Machine Learning (ML)-based method for the inverse design of the meta-optical structure. We present a data-driven approach for modeling a grating meta-structure which performs photonic beam engineering. On-chip planar photonic waveguide-based beam engineering offers the potential to efficiently manipulate photons to create excitation beams (Gaussian, focused and collimated) for lab-on-chip applications of Infrared, Raman and fluorescence spectroscopic analysis. Inverse modeling predicts meta surface design parameters based on a desired electromagnetic field outcome. Starting with the desired diffraction beam profile, we apply an inverse model to evaluate the optimal design parameters of the meta surface. Parameters such as the repetition period (in 2D axis), height and size of scatterers are calculated using a feedforward deep neural network (DNN) and convolutional neural network (CNN) architecture. A qualitative analysis of the trained neural network, working in tandem with the forward model, predicts the diffraction profile with a correlation coefficient as high as 0.996. The developed model allows us to rapidly estimate the desired design parameters, in contrast to conventional (gradient descent based or genetic optimization) time-intensive optimization approaches.