BMC biomedical engineering最新文献

Background: The ever-growing need for cheap, simple, fast, and accurate healthcare solutions spurred a lot of research activities which are aimed at the reliable deployment of artificial intelligence in the medical fields. However, this has proved to be a daunting task especially when looking to make automated diagnoses using biomedical image data. Biomedical image data have complex patterns which human experts find very hard to comprehend. Against this backdrop, we applied a representation or feature learning algorithm: Invariant Scattering Convolution Network or Wavelet scattering Network to retinal fundus images and studied the the efficacy of the automatically extracted features therefrom for glaucoma diagnosis/detection. The influence of wavelet scattering network parameter settings as well as 2-D channel image type on the detection correctness is also examined. Our work is a distinct departure from the usual method where wavelet transform is applied to pre-processed retinal fundus images and handcrafted features are extracted from the decomposition results. Here, the RIM-ONE DL image dataset was fed into a wavelet scattering network developed in the Matlab environment to achieve a stage-wise decomposition process called wavelet scattering of the retinal fundus images thereby, automatically learning features from the images. These features were then used to build simple and computationally cheap classification algorithms.

Results: Maximum detection correctness of 98% was achieved on the held-out test set. Detection correctness is highly sensitive to scattering network parameter setting and 2-D channel image type.

Conclusion: A superficial comparison of the classification results obtained from our work and those obtained using a convolutional neural network underscores the potentiality of the proposed method for glaucoma detection.

Introduction: Short-term emergency ventilation is most typically accomplished through bag valve mask (BVM) techniques. BVMs like the AMBU^® bag are cost-effective and highly portable but are also highly prone to user error, especially in high-stress emergent situations. Inaccurate and inappropriate ventilation has the potential to inflict great injury to patients through hyper- and hypoventilation. Here, we present the BVM Emergency Narration-Guided Instrument (BENGI) - a tidal volume feedback monitoring device that provides instantaneous visual and audio feedback on delivered tidal volumes, respiratory rates, and inspiratory/expiratory times. Providing feedback on the depth and regularity of respirations enables providers to deliver more consistent and accurate tidal volumes and rates. We describe the design, assembly, and validation of the BENGI as a practical tool to reduce manual ventilation-induced lung injury.

Methods: The prototype BENGI was assembled with custom 3D-printed housing and commercially available electronic components. A mass flow sensor in the central channel of the device measures air flow, which is used to calculate tidal volume. Tidal volumes are displayed via an LED ring affixed to the top of the BENGI. Additional feedback is provided through a speaker in the device. Central processing is accomplished through an Arduino microcontroller. Validation of the BENGI was accomplished using benchtop simulation with a clinical ventilator, BVM, and manikin test lung. Known respiratory quantities were delivered by the ventilator which were then compared to measurements from the BENGI to validate the accuracy of flow measurements, tidal volume calculations, and audio cue triggers.

Results: BENGI tidal volume measurements were found to lie within 4% of true delivered tidal volume values (95% CI of 0.53 to 3.7%) when breaths were delivered with 1-s inspiratory times, with similar performance for breaths delivered with 0.5-s inspiratory times (95% CI of 1.1 to 6.7%) and 2-s inspiratory times (95% CI of -1.1 to 2.3%). Audio cues "Bag faster" (1.84 to 2.03 s), "Bag slower" (0.35 to 0.41 s), and "Leak detected" (43 to 50%) were triggered close to target trigger values (2.00 s, 0.50 s, and 50%, respectively) across varying tidal volumes.

Conclusions: The BENGI achieved its proposed goals of accurately measuring and reporting tidal volumes delivered through BVM systems, providing immediate feedback on the quality of respiratory performance through audio and visual cues. The BENGI has the potential to reduce manual ventilation-induced lung injury and improve patient outcomes by providing accurate feedback on ventilatory parameters.

Background: Finite element modelling the material behavior of bone in-silico is a powerful tool to predict the best suited surgical treatment for individual patients.

Results: We demonstrate the development and use of a pre-processing plug-in program with a 3D modelling image processing software suite (Synopsys Simpleware, ScanIP) to assist with identifying, isolating, and defining cortical and trabecular bone material properties from patient specific computed tomography scans. The workflow starts by calibrating grayscale values of each constituent element with a phantom - a standardized object with defined densities. Using an established power law equation, we convert the apparent density value per voxel to a Young's Modulus. The resulting "calibrated" scan can be used for modeling and in-silico experimentation with Finite Element Analysis.

Conclusions: This process allows for the creation of realistic and personalized simulations to inform a surgeon's decision-making. We have made this plug-in program open and accessible as a supplemental file.

Neural interface devices interact with the central nervous system (CNS) to substitute for some sort of functional deficit and improve quality of life for persons with disabilities. Design of safe, biocompatible neural interface devices is a fast-emerging field of neuroscience research. Development of invasive implant materials designed to directly interface with brain or spinal cord tissue has focussed on mitigation of glial scar reactivity toward the implant itself, but little exists in the literature that directly documents the effects of electrical stimulation on glial cells. In this review, a survey of studies documenting such effects has been compiled and categorized based on the various types of stimulation paradigms used and their observed effects on glia. A hybrid neuroscience cell biology-engineering perspective is offered to highlight considerations that must be made in both disciplines in the development of a safe implant. To advance knowledge on how electrical stimulation affects glia, we also suggest experiments elucidating electrochemical reactions that may occur as a result of electrical stimulation and how such reactions may affect glia. Designing a biocompatible stimulation paradigm should be a forefront consideration in the development of a device with improved safety and longevity.

Replicating the mechanical behavior of human bones, especially cancellous bone tissue, is challenging. Typically, conventional bone models primarily consist of polyurethane foam surrounded by a solid shell. Although nearly isotropic foam components have mechanical properties similar to cancellous bone, they do not represent the anisotropy and inhomogeneity of bone architecture. To consider the architecture of bone, models were developed whose core was additively manufactured based on CT data. This core was subsequently coated with glass fiber composite. Specimens consisting of a gyroid-structure were fabricated using fused filament fabrication (FFF) techniques from different materials and various filler levels. Subsequent compression tests showed good accordance between the mechanical behavior of the printed specimens and human bone. The unidirectional fiberglass composite showed higher strength and stiffness than human cortical bone in 3-point bending tests, with comparable material behaviors being observed. During biomechanical investigation of the entire assembly, femoral prosthetic stems were inserted into both artificial and human bones under controlled conditions, while recording occurring forces and strains. All of the artificial prototypes, made of different materials, showed analogous behavior to human bone. In conclusion, it was shown that low-cost FFF technique can be used to generate valid bone models and selectively modify their properties by changing the infill.

Background: Electrocorticography (ECoG) arrays are commonly used to record the brain activity both in animal and human subjects. There is a lack of guidelines in the literature as to how the array geometry, particularly the via holes in the substrate, affects the recorded signals. A finite element (FE) model was developed to simulate the electric field generated by neurons located at different depths in the rat brain cortex and a micro ECoG array (μECoG) was placed on the pia surface for recording the neural signal. The array design chosen was a typical array of 8 × 8 circular (100 μm in diam.) contacts with 500 μm pitch. The size of the via holes between the recording contacts was varied to see the effect.

Results: The results showed that recorded signal amplitudes were reduced if the substrate was smaller than about four times the depth of the neuron in the gray matter. The signal amplitude profiles had dips around the via holes and the amplitudes were also lower at the contact sites as compared to the design without the holes; an effect that increased with the hole size. Another noteworthy result is that the spatial selectivity of the multi-contact recordings could be improved or reduced by the selection of the via hole sizes, and the effect depended on the distance between the neuron pair targeted for selective recording and its depth.

Conclusions: The results suggest that the via-hole size clearly affects the recorded neural signal amplitudes and it can be leveraged as a parameter to reduce the inter-channel correlation and thus maximize the information content of neural signals with μECoG arrays.

Background: Detection of locally increased blood concentration and perfusion is critical for assessment of functional cortical activity as well as diagnosis of conditions such as intracerebral hemorrhage (ICH). Current paradigms for assessment of regional blood concentration in the brain rely on computed tomography (CT), magnetic resonance imaging (MRI), and perfusion blood oxygen level dependent functional magnetic resonance imaging (BOLD-fMRI).

Results: In this study, we developed computational models to test the feasibility of novel magnetic sensors capable of detecting hemodynamic changes within the brain on a microtesla-level. We show that low-field magnetic sensors can accurately detect changes in magnetic flux density and eddy current damping signals resulting from increases in local blood concentration. These models predicted that blood volume changes as small as 1.26 mL may be resolved by the sensors, implying potential use for diagnosis of ICH and assessment of regional blood flow as a proxy for cerebral metabolism and neuronal activity. We then translated findings from our computational model to demonstrate feasibility of accurate detection of modeled ICH in a simulated human cadaver setting.

Conclusions: Overall, microtesla-level magnetic scanning is feasible, safe, and has distinct advantages compared to current standards of care. Computational modeling may facilitate rapid prototype development and testing of novel medical devices with minimal risk to human participants prior to device construction and clinical trials.