Camera fusion for real-time temperature monitoring of neonates using deep learning.

The continuous monitoring of vital signs is a crucial aspect of medical care in neonatal intensive care units. Since cable-based sensors pose a potential risk for the immature skin of preterm infants, unobtrusive monitoring techniques using camera systems are increasingly investigated. The combination of deep learning-based algorithms and camera modalities such as RGB and infrared thermography can improve the development of cable-free methods for the extraction of vital parameters. In this study, a real-time approach for local extraction of temperatures on the body surface of neonates using a multi-modal clinical dataset was implemented. Therefore, a trained deep learning-based keypoint detector was used for body landmark prediction in RGB. Image registration was conducted to transfer the RGB points to the corresponding thermographic recordings. These landmarks were used to extract the body surface temperature in various regions to determine the central-peripheral temperature difference. A validation of the keypoint detector showed a mean average precision of 0.82. The registration resulted in mean absolute errors of 16.4 px (8.2 mm) for x and 22.4 px (11.2 mm) for y. The evaluation of the temperature extraction revealed a mean absolute error of 0.55 [Formula: see text]C. A final performance of 31 fps was observed on the NVIDIA Jetson Xavier NX module, which proves real-time capability on an embedded GPU system. As a result, the approach can perform real-time temperature extraction on a low-cost GPU module.

Hypothermia for moderate or severe neonatal encephalopathy in low-income and middle-income countries (HELIX): a randomised controlled trial in India, Sri Lanka, and Bangladesh.

BACKGROUND: Although therapeutic hypothermia reduces death or disability after neonatal encephalopathy in high-income countries, its safety and efficacy in low-income and middle-income countries is unclear. We aimed to examine whether therapeutic hypothermia alongside optimal supportive intensive care reduces death or moderate or severe disability after neonatal encephalopathy in south Asia. METHODS: We did a multicountry open-label, randomised controlled trial in seven tertiary neonatal intensive care units in India, Sri Lanka, and Bangladesh. We enrolled infants born at or after 36 weeks of gestation with moderate or severe neonatal encephalopathy and a need for continued resuscitation at 5 min of age or an Apgar score of less than 6 at 5 min of age (for babies born in a hospital), or both, or an absence of crying by 5 min of age (for babies born at home). Using a web-based randomisation system, we allocated infants into a group receiving whole body hypothermia (33·5°C) for 72 h using a servo-controlled cooling device, or to usual care (control group), within 6 h of birth. All recruiting sites had facilities for invasive ventilation, cardiovascular support, and access to 3 Tesla MRI scanners and spectroscopy. Masking of the intervention was not possible, but those involved in the magnetic resonance biomarker analysis and neurodevelopmental outcome assessments were masked to the allocation. The primary outcome was a combined endpoint of death or moderate or severe disability at 18-22 months, assessed by the Bayley Scales of Infant and Toddler Development (third edition) and a detailed neurological examination. Analysis was by intention to treat. This trial is registered with ClinicalTrials.gov, NCT02387385. FINDINGS: We screened 2296 infants between Aug 15, 2015, and Feb 15, 2019, of whom 576 infants were eligible for inclusion. After exclusions, we recruited 408 eligible infants and we assigned 202 to the hypothermia group and 206 to the control group. Primary outcome data were available for 195 (97%) of the 202 infants in the hypothermia group and 199 (97%) of the 206 control group infants. 98 (50%) infants in the hypothermia group and 94 (47%) infants in the control group died or had a moderate or severe disability (risk ratio 1·06; 95% CI 0·87-1·30; p=0·55). 84 infants (42%) in the hypothermia group and 63 (31%; p=0·022) infants in the control group died, of whom 72 (36%) and 49 (24%; p=0·0087) died during neonatal hospitalisation. Five serious adverse events were reported: three in the hypothermia group (one hospital readmission relating to pneumonia, one septic arthritis, and one suspected venous thrombosis), and two in the control group (one related to desaturations during MRI and other because of endotracheal tube displacement during transport for MRI). No adverse events were considered causally related to the study intervention. INTERPRETATION: Therapeutic hypothermia did not reduce the combined outcome of death or disability at 18 months after neonatal encephalopathy in low-income and middle-income countries, but significantly increased death alone. Therapeutic hypothermia should not be offered as treatment for neonatal encephalopathy in low-income and middle-income countries, even when tertiary neonatal intensive care facilities are available. FUNDING: National Institute for Health Research, Garfield Weston Foundation, and Bill & Melinda Gates Foundation. TRANSLATIONS: For the Hindi, Malayalam, Telugu, Kannada, Singhalese, Tamil, Marathi and Bangla translations of the abstract see Supplementary Materials section.

Fast body part segmentation and tracking of neonatal video data using deep learning.

Photoplethysmography imaging (PPGI) for non-contact monitoring of preterm infants in the neonatal intensive care unit (NICU) is a promising technology, as it could reduce medical adhesive-related skin injuries and associated complications. For practical implementations of PPGI, a region of interest has to be detected automatically in real time. As the neonates' body proportions differ significantly from adults, existing approaches may not be used in a straightforward way, and color-based skin detection requires RGB data, thus prohibiting the use of less-intrusive near-infrared (NIR) acquisition. In this paper, we present a deep learning-based method for segmentation of neonatal video data. We augmented an existing encoder-decoder semantic segmentation method with a modified version of the ResNet-50 encoder. This reduced the computational time by a factor of 7.5, so that 30 frames per second can be processed at 960 × 576 pixels. The method was developed and optimized on publicly available databases with segmentation data from adults. For evaluation, a comprehensive dataset consisting of RGB and NIR video recordings from 29 neonates with various skin tones recorded in two NICUs in Germany and India was used. From all recordings, 643 frames were manually segmented. After pre-training the model on the public adult data, parts of the neonatal data were used for additional learning and left-out neonates are used for cross-validated evaluation. On the RGB data, the head is segmented well (82% intersection over union, 88% accuracy), and performance is comparable with those achieved on large, public, non-neonatal datasets. On the other hand, performance on the NIR data was inferior. By employing data augmentation to generate additional virtual NIR data for training, results could be improved and the head could be segmented with 62% intersection over union and 65% accuracy. The method is in theory capable of performing segmentation in real time and thus it may provide a useful tool for future PPGI applications. Graphical Abstract This work presents the development of a customized, real-time capable Deep Learning architecture for segmenting of neonatal videos recorded in the intensive care unit. In addition to hand-annotated data, transfer learning is exploited to improve performance.

Goat lung surfactant for treatment of respiratory distress syndrome among preterm neonates: a multi-site randomized non-inferiority trial.

OBJECTIVE: To investigate the safety and efficacy of goat lung surfactant extract (GLSE) compared with bovine surfactant extract (beractant; Survanta®, AbbVie, USA) for the treatment of neonatal respiratory distress syndrome (RDS). STUDY DESIGN: We conducted a double-blind, non-inferiority, randomized trial in seven Indian centers between June 22, 2016 and January 11, 2018. Preterm neonates of 26 to 32 weeks gestation with clinical diagnosis of RDS were randomized to receive either GLSE or beractant. Repeat dose, if required, was open-label beractant in both the groups. The primary outcome was a composite of death or bronchopulmonary dysplasia (BPD) at 36 weeks postmenstrual age (PMA). Interim analyses were done by an independent data and safety monitoring board (DSMB). RESULT: After the first interim analyses on 5% enrolment, the "need for repeat dose(s) of surfactant" was added as an additional primary outcome and enrolment restricted to intramural births at five of the seven participating centers. Following second interim analysis after 98 (10% of 900 planned) neonates were enroled, DSMB recommended closure of study in view of inferior efficacy of GLSE in comparison to beractant. There was no significant difference in the primary outcome of death or BPD between GLSE group (n = 52) and beractant group (n = 46) (50.0 vs. 39.1%; OR 1.5; 95% CI 0.7-3.5; p = 0.28). The need for repeat dose of surfactant was significantly higher in GLSE group (65.4 vs. 17.4%; OR 9.0; 95% CI 3.5-23.3; p < 0.001). CONCLUSIONS: Goat lung surfactant was less efficacious than beractant (Survanta®) for treatment of RDS in preterm infants. Reasons to ascertain inferior efficacy of goat lung surfactant requires investigation and possible mitigating strategies in order to develop a low-cost and effective surfactant.

Two Randomized Trials of the Effect of the Russian Strain of Bacillus Calmette-Guérin Alone or With Oral Polio Vaccine on Neonatal Mortality in Infants Weighing <2000 g in India.

BACKGROUND: In randomized trials in Guinea-Bissau, the Danish strain of Bacillus Calmette-Guérin (BCG) reduces neonatal mortality, primarily by reducing deaths from pneumonia and sepsis. Because World Health Organization-prequalified BCG-Denmark was not available in India, we conducted 2 randomized trials to test whether BCG-Russia alone or with oral polio vaccine (OPV) has similar effects to BCG-Denmark. METHODS: We randomized neonates weighing <2000 g to a control group that was not vaccinated before 28 days of age or to receive either BCG-Russia alone (first trial) or BCG-Russia with OPV (second trial) soon after birth. We performed intention-to-treat analysis using Cox hazards models with age as the underlying time and adjusted for weight, sex and inborn versus outborn status. RESULTS: Administration of BCG-Russia alone had no effect on neonatal mortality (to 28 days of age): 15.6% of 1537 infants died in the BCG-Russia group and 16.1% of 1535 died in the control group; the adjusted hazard ratio was 0.95 [95% confidence interval (CI): 0.80-1.13]. Administration of BCG-Russia with OPV also had no effect on neonatal mortality: 18.0% of 1103 infants died in the BCG-OPV group and 17.6% of 1104 died in the control group; the adjusted hazard ratio was 1.01 (95% CI: 0.83-1.23). The adjusted hazard ratio for the 2 trials combined was 0.98 (95% CI: 0.85-1.11). CONCLUSIONS: BCG-Russia with or without OPV had no effect on neonatal mortality. It is important to determine which strains of BCG have the greatest specific effects (on tuberculosis) and nonspecific effects (on infections other than tuberculosis) in high-mortality regions.