Analysis of AI-Based Single-View 3D Reconstruction Methods for an Industrial Application.

Machine learning (ML) is a key technology in smart manufacturing as it provides insights into complex processes without requiring deep domain expertise. This work deals with deep learning algorithms to determine a 3D reconstruction from a single 2D grayscale image. The potential of 3D reconstruction can be used for quality control because the height values contain relevant information that is not visible in 2D data. Instead of 3D scans, estimated depth maps based on a 2D input image can be used with the advantage of a simple setup and a short recording time. Determining a 3D reconstruction from a single input image is a difficult task for which many algorithms and methods have been proposed in the past decades. In this work, three deep learning methods, namely stacked autoencoder (SAE), generative adversarial networks (GANs) and U-Nets are investigated, evaluated and compared for 3D reconstruction from a 2D grayscale image of laser-welded components. In this work, different variants of GANs are tested, with the conclusion that Wasserstein GANs (WGANs) are the most robust approach among them. To the best of our knowledge, the present paper considers for the first time the U-Net, which achieves outstanding results in semantic segmentation, in the context of 3D reconstruction tasks. Unlike the U-Net, which uses standard convolutions, the stacked dilated U-Net (SDU-Net) applies stacked dilated convolutions. Of all the 3D reconstruction approaches considered in this work, the SDU-Net shows the best performance, not only in terms of evaluation metrics but also in terms of computation time. Due to the comparably small number of trainable parameters and the suitability of the architecture for strong data augmentation, a robust model can be generated with only a few training data.

Time to adjust to changes in ventilation settings varies significantly between different T-piece resuscitators, self-inflating bags, and manometer equipped self-inflating bags.

OBJECTIVE: Resuscitation guidelines give no preference over use of self-inflating bags (SIBs) or T-piece resuscitators (TPR) for manual neonatal ventilation. We speculated that devices would differ significantly regarding time required to adjust to changed ventilation settings. STUDY DESIGN: This was a laboratory study. Time to adjust from baseline peak inflation pressure (PIP) (20 cmH2O) to target PIP (25 and 40 cmH2O), ability to adhere to predefined ventilation settings (PIP, PEEP, and inflation rate [IR]), and the variability within and between operators were assessed for a SIB without manometer, SIB with manometer (SIBM), and two TPRs. RESULTS: Adjustment time was significantly longer with TPRs, compared with SIB and SIBM. The SIBM and TPRs were < 5% (median) off target PIP, and the SIB was 14% off target PIP. Significant variability between operators (interquartile range [IQR]: 71%) was seen with SIBs. CONCLUSION: PIP adjustment takes longer with TPRs, compared with SIB/SIBM. TPRs and SIBM allow satisfactory adherence to ventilation parameters. SIBs should only be used with manometer attached.

Repeated thermo-sterilisation further affects the reliability of positive end-expiratory pressure valves.

AIM: Positive end-expiratory pressure (PEEP) valves are used together with self-inflating bags (SIB) to provide a preset PEEP during manual ventilation. It has recently been shown that these valves deliver highly variable levels of PEEP. We hypothesised that material fatigue due to repeated thermo-sterilisation (TS) may contribute to varying reliability of PEEP valves. METHODS: In a laboratory study 10 new PEEP valves were tested before and after 10, 20 and 30 cycles of routine TS (7 min at 134°C) by using a neonatal lung model (compliance 0.2 mL/kPa). Settings were positive inflation pressure = 20 and 40 cm H(2)O, PEEP = 5 and 10 cm H(2)O, respiratory rate = 40 and 60/min, flow = 8l/min. PEEP was recorded using a respiratory function monitor. RESULTS: Before TS, a mean (standard deviation) PEEP of 4.0 (0.9) and 7.7 (1.0) cm H(2)O was delivered by the 10 valves when the PEEP was set to 5 and 10 cm H(2)O, respectively. One new valve only delivered 2.0 (0.0) and 5.0 (0.0) cm H(2)O when the PEEP was adjusted to 5 and 10 cm H(2)O, respectively. Four of the 10 investigated valves showed significant variations in PEEP (coefficient of variation >10%) throughout the autoclaving process. One valve completely lost its function after the 20th TS. Common defects were tears in the softer materials or displacement of the rubber seal. Six of the 10 valves continued to provide PEEP in spite of repeated TS. CONCLUSION: The reliability of PEEP valves is affected by repeated TS. Multi-use PEEP valves should be tested for reliable PEEP provision following TS.

Manual neonatal ventilation training: a respiratory function monitor helps to reduce peak inspiratory pressures and tidal volumes during resuscitation.

BACKGROUND: Neonatal resuscitation training is considered to be multifarious and includes manual ventilation as an essential competence for any health-care provider. Usually, ventilation is applied with self-inflating bags (SIBs). These devices have been shown to produce highly variable, operator-dependent peak inspiratory pressures (PIPs) and tidal volumes (V(T)). Excessive PIP and V(T) contribute to lung injury. We studied a simple tool to improve resuscitation skills. OBJECTIVE: The objectives of this study were to train healthcare providers to avoid excessive PIP and V(T) by visualizing these values by using a respiratory function monitor (RFM) and to study the sustainability of such a training. MATERIAL AND METHODS: Previously untrained medical professionals were educated and trained to ventilate a neonatal preterm manikin. PIP and V(T) were measured with an RFM. Graphical representations of the measurements were displayed during training, but the RFM was blinded during subsequent recordings. Participants were reassessed directly after training and 1 month later. RESULTS: In total, 37 participants were trained and assessed three times during the study. Median PIPs (range) were 32.3 (4.1 ­ 44) cm H(2)O before training, 17.8 (9.6 ­ 23.6) cm H(2)O directly after training (P < 0.05), and 18.7 (7.5 ­ 41.6) cm H(2)O 1 month later, and the values remained low, compared with before training (P < 0.05). Median V(T)s were 6.7 (4.2 ­ 44) mL before training, 3.5 (1.8 ­ 7.3) mL directly after training (P < 0.05), and 4.1 (1.9 ­ 9.7 mL) 1 month after training (P < 0.05). CONCLUSION: Using a SIB, untrained staff produced excessive PIP and V(T). Training with a simple RFM significantly reduced the occurrence of excessive PIP and V(T). The effect was sustained for at least 1 month.

Reliability of Single-Use PEEP-Valves Attached to Self-Inflating Bags during Manual Ventilation of Neonates--An In Vitro Study.

INTRODUCTION: International resuscitation guidelines suggest to use positive end-expiratory pressure (PEEP) during manual ventilation of neonates. Aim of our study was to test the reliability of self-inflating bags (SIB) with single-use PEEP valves regarding PEEP delivery and the effect of different peak inflation pressures (PIP) and ventilation rates (VR) on the delivered PEEP. METHODS: Ten new single-use PEEP valves from 5 manufacturers were tested by ventilating an intubated 1 kg neonatal manikin containing a lung model with a SIB that was actuated by an electromechanical plunger device. Standard settings: PIP 20 cmH2O, VR 60/min, flow 8 L/min. PEEP settings of 5 and 10 cmH2O were studied. A second test was conducted with settings of PIP 40 cmH2O and VR 40/min. The delivered PEEP was measured by a respiratory function monitor (CO2SMO+). RESULTS: Valves from one manufacturer delivered no relevant PEEP and were excluded. The remaining valves showed a continuous decay of the delivered pressure during expiration. The median (25th and 75th percentile) delivered PEEP with standard settings was 3.4(2.7-3.8) cmH2O when set to 5 cmH2O and 6.1(4.9-7.1) cmH2O when set to 10 cmH2O. Increasing the PIP from 20 to 40 cmH2O led to a median (25th and 75th percentile) decrease in PEEP to 2.3(1.8-2.7) cmH2O and 4.3(3.2-4.8) cmH2O; changing VR from 60 to 40/min led to a PEEP decrease to 2.8(2.1-3.3) cmH2O and 5.0(3.5-6.2) cmH2O for both PEEP settings. CONCLUSION: Single-use PEEP valves do not reliably deliver the set PEEP. PIP and VR have an effect on the delivered PEEP. Operators should be aware of these limitations when manually ventilating neonates.

Physiological properties of milk ingredients released by fermentation.

The demand for health-promoting food ingredients rises within an increasing market worldwide. Different milks fermented with bacteria, yeasts, moulds or enzymes from animal, plant and microbial sources offer a broad range of possibilities to cover different health aspects with new bioactive components. By the fermentation process interesting ingredients are enriched and released from the matrix, like lactoferrin, micro-nutrients, CLA and sphingolipids or synthesized, such as exo-polysaccharides and bioactive peptides. In particular, milk derived bioactive peptides exert several important health-promoting activities, such as anti-hypertensive, anti-microbial, anti-oxidative, immune-modulatory, opioid and mineral-binding properties. Milk-fermentation processes with probiotic bacteria synergistically combine health supporting bacterial and milk ingredient aspects which include new therapeutic solutions concerning hypercholesterolemia, carcinogenic intoxications, treatment of diarrhea, reduction of intestine pathogens, and supporting natural immune defense. Especially, milk-proteins and associated bioactive peptides released during microbial or enzymatic fermentation of milk offer a broad spectrum of new functional properties, for instance anti-hypertensive, anti-microbial, anti-oxidative, immuno-modulatory, opioid and mineral-binding properties. This review aimed at discussing recent research activities on physiological purposes and technical process aspects of functional components from fermented milk with a specific focus on biofunctional peptides released from fermented milk proteins.