Deep learning aided preoperative diagnosis of primary central nervous system lymphoma.

The preoperative distinction between glioblastoma (GBM) and primary central nervous system lymphoma (PCNSL) can be difficult, even for experts, but is highly relevant. We aimed to develop an easy-to-use algorithm, based on a convolutional neural network (CNN) to preoperatively discern PCNSL from GBM and systematically compare its performance to experienced neurosurgeons and radiologists. To this end, a CNN-based on DenseNet169 was trained with the magnetic resonance (MR)-imaging data of 68 PCNSL and 69 GBM patients and its performance compared to six trained experts on an external test set of 10 PCNSL and 10 GBM. Our neural network predicted PCNSL with an accuracy of 80% and a negative predictive value (NPV) of 0.8, exceeding the accuracy achieved by clinicians (73%, NPV 0.77). Combining expert rating with automated diagnosis in those cases where experts dissented yielded an accuracy of 95%. Our approach has the potential to significantly augment the preoperative radiological diagnosis of PCNSL.

Different Techniques of Respiratory Support Do Not Significantly Affect Gas Exchange during Cardiopulmonary Resuscitation in a Newborn Piglet Model.

BACKGROUND: There are no evidence-based recommendations on the use of different techniques of respiratory support and chest compressions (CC) during neonatal cardiopulmonary resuscitation (CPR). OBJECTIVES: We studied the short-term effects of different ventilatory support strategies along with CC representing clinical practice on gas exchange [arterial oxygen saturation (SaO2), arterial partial pressure of oxygen (PaO2) and arterial partial pressure of carbon dioxide (PaCO2)], hemodynamics and cerebral oxygenation. We hypothesized that in newborn piglets with cardiac arrest, use of a T-piece resuscitator (TPR) providing positive end-expiratory pressure (PEEP) improves gas exchange as measured by SaO2 during CPR as compared to using a self-inflating bag (SIB) without PEEP. Furthermore, we explored the effects of a mechanical ventilator without synchrony to CC. METHODS: Thirty newborn piglets with asystole were randomized into three groups and resuscitated for 20 min [fraction of inspired oxygen (FiO2) = 0.21 for 10 min and 1.0 thereafter]. Group 1 received ventilation using a TPR [peak inspiratory pressure (PIP)/PEEP of 20/5 cm H2O, rate 30/min] with inflations interposed between CC (3:1 ratio). Group 2 received ventilation using a SIB (PIP of 20 cm H2O without PEEP, rate 30/min) with inflations interposed between CC (3:1 ratio). Group 3 received ventilation using a mechanical ventilator (PIP/PEEP of 20/5 cm H2O, rate 30/min). CC were applied with a rate of 120/min without synchrony to inflations. RESULTS: We found no significant differences in SaO2 between the three groups. However, there was a trend toward a higher SaO2 [TPR: 28.0% (22.3-40.0); SIB: 23.7% (13.4-52.3); ventilator: 44.1% (39.2-54.3); median (interquartile range)] and a lower PaCO2 [TPR: 95.6 mm Hg (82.1-113.6); SIB: 100.8 mm Hg (83.0-108.0); ventilator: 74.1 mm Hg (68.5-83.1); median (interquartile range)] in the mechanical ventilator group. CONCLUSIONS: We found no significant effect on gas exchange using different respiratory support strategies during CPR.

Reliability of pulse oximetry during cardiopulmonary resuscitation in a piglet model of neonatal cardiac arrest.

BACKGROUND: Pulse oximetry is widely used in intensive care and emergency conditions to monitor arterial oxygenation and to guide oxygen therapy. OBJECTIVE: To study the reliability of pulse oximetry in comparison with CO-oximetry in newborn piglets during cardiopulmonary resuscitation (CPR). METHODOLOGY: In a prospective cohort study in 30 healthy newborn piglets, cardiac arrest was induced, and thereafter each piglet received CPR for 20 min. Arterial oxygen saturation was monitored continuously by pulse oximetry (SpO2). Arterial blood was analyzed for functional oxygenation (SaO2) every 2 min. SpO2 was compared with coinciding SaO2 values and bias considered whenever the difference (SpO2 - SaO2) was beyond ±5%. RESULTS: Bias values were decreased at the baseline measurements (mean: 2.5 ± 4.6%) with higher precision and accuracy compared with values across the experiment. Two minutes after cardiac arrest, there was a marked decrease in precision and accuracy as well as an increase in bias up to 13 ± 34%, reaching a maximum of 45.6 ± 28.3% after 10 min over a mean SaO2 range of 29-58%. CONCLUSION: Pulse oximetry showed increased bias and decreased accuracy and precision during CPR in a model of neonatal cardiac arrest. We recommend further studies to clarify the exact mechanisms of these false readings to improve reliability of pulse oximetry during the marked desaturation and hypoperfusion found during CPR.