Comparison of Commercial AI Software Performance for Radiograph Lung Nodule Detection and Bone Age Prediction.

Background Multiple commercial artificial intelligence (AI) products exist for assessing radiographs; however, comparable performance data for these algorithms are limited. Purpose To perform an independent, stand-alone validation of commercially available AI products for bone age prediction based on hand radiographs and lung nodule detection on chest radiographs. Materials and Methods This retrospective study was carried out as part of Project AIR. Nine of 17 eligible AI products were validated on data from seven Dutch hospitals. For bone age prediction, the root mean square error (RMSE) and Pearson correlation coefficient were computed. The reference standard was set by three to five expert readers. For lung nodule detection, the area under the receiver operating characteristic curve (AUC) was computed. The reference standard was set by a chest radiologist based on CT. Randomized subsets of hand (n = 95) and chest (n = 140) radiographs were read by 14 and 17 human readers, respectively, with varying experience. Results Two bone age prediction algorithms were tested on hand radiographs (from January 2017 to January 2022) in 326 patients (mean age, 10 years ± 4 [SD]; 173 female patients) and correlated strongly with the reference standard (r = 0.99; P < .001 for both). No difference in RMSE was observed between algorithms (0.63 years [95% CI: 0.58, 0.69] and 0.57 years [95% CI: 0.52, 0.61]) and readers (0.68 years [95% CI: 0.64, 0.73]). Seven lung nodule detection algorithms were validated on chest radiographs (from January 2012 to May 2022) in 386 patients (mean age, 64 years ± 11; 223 male patients). Compared with readers (mean AUC, 0.81 [95% CI: 0.77, 0.85]), four algorithms performed better (AUC range, 0.86-0.93; P value range, <.001 to .04). Conclusions Compared with human readers, four AI algorithms for detecting lung nodules on chest radiographs showed improved performance, whereas the remaining algorithms tested showed no evidence of a difference in performance. © RSNA, 2024 Supplemental material is available for this article. See also the editorial by Omoumi and Richiardi in this issue.

Postoperative cerebral oxygenation was not associated with new brain injury in infants with congenital heart disease.

OBJECTIVE: The aim of this study was to evaluate postoperative indices of cerebral oxygenation and autoregulation in infants with critical congenital heart disease in relation to new postoperative ischemic brain injury. METHODS: This prospective, clinical cohort included 77 infants with transposition of the great arteries (N = 19), left ventricular outflow tract obstruction (N = 30), and single ventricle physiology (N = 28) undergoing surgery at 30 days or less of life. Postoperative near-infrared spectroscopy and physiologic monitoring were applied to extract mean arterial blood pressure, regional cerebral oxygen saturation, fractional tissue oxygen extraction, and regional cerebral oxygen saturation mean arterial blood pressure correlation coefficient (≥0.5 considered sign of impaired cerebral autoregulation). New postoperative ischemic injury was defined as moderate-severe white matter injury or focal infarction on magnetic resonance imaging. Low cardiac output syndrome was measured as lactate greater than 4 mmol/L with pH less than 7.30. RESULTS: After surgery, regional cerebral oxygen saturation was decreased in all congenital heart disease groups with a notable increase in regional cerebral oxygen saturation between 6 and 12 hours after surgery, on average with a factor of 1.4 (range, 1.1-2.4). Both single ventricle physiology and postoperative low cardiac output syndrome were associated with lower regional cerebral oxygen saturation and increased time with correlation coefficient of 0.5 or greater. New postoperative ischemic injury was seen in 39 patients (53%) and equally distributed across congenital heart disease groups. Postoperative regional cerebral oxygen saturation, fractional tissue oxygen extraction, and correlation coefficient were not independently associated with new postoperative white matter injury or focal infarction (mixed-model analysis, all F > 0.12). CONCLUSIONS: Postoperative indices of cerebral oxygenation and cerebral autoregulation are not independent predictors of new ischemic brain injury in infants with critical congenital heart disease. Further exploration of the complex interplay among low regional cerebral oxygen saturation, low cardiac output syndrome, and heart defect is required to identify potential biomarkers enabling early intervention for ischemic brain injury.

Amplitude-Integrated Electroencephalography for Early Recognition of Brain Injury in Neonates with Critical Congenital Heart Disease.

OBJECTIVE: To study perioperative amplitude-integrated electroencephalography (aEEG) as an early marker for new brain injury in neonates requiring cardiac surgery for critical congenital heart disease (CHD). STUDY DESIGN: This retrospective observational cohort study investigated 76 neonates with critical CHD who underwent neonatal surgery. Perioperative aEEG recordings were evaluated for background pattern (BGP), sleep-wake cycling (SWC), and ictal discharges. Spontaneous activity transient (SAT) rate, inter-SAT interval (ISI), and percentage of time with an amplitude <5 µV were calculated. Routinely obtained preoperative and postoperative magnetic resonance imaging of the brain were reviewed for brain injury (moderate-severe white matter injury, stroke, intraparenchymal hemorrhage, or cerebral sinovenous thrombosis). RESULTS: Preoperatively, none of the neonates showed an abnormal BGP (burst suppression or worse) or ictal discharges. Postoperatively, abnormal BGP was seen in 18 neonates (24%; 95% CI, 14%-33%) and ictal discharges was seen in 13 neonates (17%; 95% CI, 8%-26%). Abnormal BGP and ictal discharges were more frequent in neonates with new postoperative brain injury (P = .08 and .01, respectively). Abnormal brain activity (ie, abnormal BGP or ictal discharges) was the single risk factor associated with new postoperative brain injury in multivariable logistic regression analysis (OR, 4.0; 95% CI, 1.3-12.3; P = .02). Postoperative SAT rate, ISI, or time <5 µV were not associated with new brain injury. CONCLUSION: Abnormal brain activity is an early, bedside marker of new brain injury in neonates undergoing cardiac surgery. Not only ictal discharges, but also abnormal BGP, should be considered a clear sign of underlying brain pathology.

Perioperative neonatal brain injury is associated with worse school-age neurodevelopment in children with critical congenital heart disease.

AIM: To assess the impact of perioperative neonatal brain injury and brain volumes on neurodevelopment throughout school-age children with critical congenital heart disease (CHD). METHOD: Thirty-four survivors of neonatal cardiac surgery (seven females, 27 males) were included. Neonatal preoperative and postoperative cerebral magnetic resonance imaging was performed and neurodevelopment was assessed at 24 months (SD 0.7, n=32, using Bayley Score of Infant and Toddler Development, Child Behavior Checklist) and 6 years (mean age 5y 11mo; SD 0.3, n=30, using Movement Assessment Battery for Children, Wechsler Preschool and Primary Scale of Intelligence, Child Behavior Checklist, Teacher Report Form). Brain injury, brain volumes, and cortical measures were related to outcome with adjustment for maternal educational level. RESULTS: Two-year cognitive score and 6-year Full-scale IQ were poorer in children with neonatal white matter injury (n=21, all p<0.05), with higher teacher-reported attention problems (p=0.03). Five of six children with involvement of the posterior limb of the internal capsule showed motor problems (p=0.03). Children with a below-average Fulll-scale IQ (<85, n=9) showed smaller volumes of basal ganglia thalami (-8%, p=0.03) and brain stem (-7%, p=0.03). INTERPRETATION: Our findings provide evidence of unfavourable outcome in school-age children with critical CHD who acquire perioperative neonatal brain injury. WHAT THIS PAPER ADDS: This paper extends knowledge about neonatal brain injury and long-term outcome in congenital heart disease. Children with white matter injury show lower IQ and more attention problems at school age. Injury of the posterior limb of the internal capsule increases the risk of motor problems. This study provides evidence for worse outcomes in neonates acquiring brain injury around cardiac surgery.

Clinical and neuroimaging characteristics of cerebral sinovenous thrombosis in neonates undergoing cardiac surgery.

OBJECTIVES: Neonates with congenital heart disease may have an increased risk of cerebral sinovenous thrombosis, but incidence rates are lacking. This study describes the clinical and neuroimaging characteristics of cerebral sinovenous thrombosis in neonates undergoing cardiac surgery. METHODS: Forty neonates (78% male) requiring neonatal univentricular or biventricular cardiac repair using cardiopulmonary bypass were included. All underwent preoperative (median postnatal day 7) and postoperative (median postoperative day 7) magnetic resonance imaging of the brain, including venography, to detect cerebral sinovenous thrombosis. Clinical characteristics were compared between cerebral sinovenous thrombosis positive and cerebral sinovenous thrombosis negative neonates. RESULTS: Postoperatively, cerebral sinovenous thrombosis was diagnosed in 11 neonates (28%), with the transverse sinus affected in all, and involvement of multiple sinuses in 10 (91%). Preoperatively, signs of thrombosis were seen in 3 cases (8%). Focal infarction of the basal ganglia was significantly more common in cerebral sinovenous thrombosis positive than cerebral sinovenous thrombosis negative neonates (P = .025). Cerebral sinovenous thrombosis positive neonates spent more time in the intensive care unit preoperatively (P = .001), had lower weight (P = .024) and lower postmenstrual age (P = .030) at surgery, and had prolonged use of a central venous catheter (P = .023) and a catheter placed in the internal jugular vein more often (P = .039). Surgical and postoperative factors were not different between new postoperative cerebral sinovenous thrombosis positive and cerebral sinovenous thrombosis negative neonates. CONCLUSIONS: Cerebral sinovenous thrombosis might be more common than previously understood in neonates undergoing cardiac surgery. In our study, cerebral sinovenous thrombosis was associated with a higher risk of additional intra-parenchymal brain injury.

Minimizing the risk of preoperative brain injury in neonates with aortic arch obstruction.

OBJECTIVE: To determine whether prenatal diagnosis lowers the risk of preoperative brain injury by assessing differences in the incidence of preoperative brain injury across centers. STUDY DESIGN: From 2 prospective cohorts of newborns with complex congenital heart disease studied by preoperative cerebral magnetic resonance imaging, one cohort from the University Medical Center Utrecht (UMCU) and a combined cohort from the University of California San Francisco (UCSF) and University of British Columbia (UBC), patients with aortic arch obstruction were selected and their imaging and clinical course reviewed. RESULTS: Birth characteristics were comparable between UMCU (n = 33) and UCSF/UBC (n = 54). Patients had a hypoplastic aortic arch with either coarctation/interruption or hypoplastic left heart syndrome. In subjects with prenatal diagnosis, there was a significant difference in the prevalence of white matter injury (WMI) between centers (11 of 22 [50%] at UMCU vs 4 of 30 [13%] at UCSF/UBC; P < .01). Prenatal diagnosis was protective for WMI at UCSF/UBC (13% prenatal diagnoses vs 50% postnatal diagnoses; P < .01), but not at UMCU (50% vs 46%, respectively; P > .99). Differences in clinical practice between prenatally diagnosed subjects at UMCU vs UCSF/UBC included older age at surgery, less time spent in the intensive care unit, greater use of diuretics, less use of total parenteral nutrition (P < .01), and a greater incidence of infections (P = .01). In patients diagnosed postnatally, the prevalence of WMI was similar in the 2 centers (46% at UMCU vs 50% at UCSF/UBC; P > .99). Stroke prevalence was similar in the 2 centers regardless of prenatal diagnosis (prenatal diagnosis: 4.5% at Utrecht vs 6.7% at UCSF/UBC, P = .75; postnatal diagnosis: 9.1% vs 13%, respectively, P > .99). CONCLUSION: Prenatal diagnosis can be protective for WMI, but this protection may be dependent on specific clinical management practices that differ across centers.

Neurological injury after neonatal cardiac surgery: a randomized, controlled trial of 2 perfusion techniques.

BACKGROUND: Complex neonatal cardiac surgery is associated with cerebral injury. In particular, aortic arch repair, requiring either deep hypothermic circulatory arrest (DHCA) or antegrade cerebral perfusion (ACP), entails a high risk of perioperative injury. It is unknown whether ACP results in less cerebral injury than DHCA. METHODS AND RESULTS: Thirty-seven neonates with an aortic arch obstruction presenting for univentricular or biventricular repair were randomized to either DHCA or ACP. Preoperatively and 1 week after surgery, magnetic resonance imaging was performed in 36 patients (1 patient died during the hospital stay). The presence of new postoperative cerebral injury was scored, and results were entered into a sequential analysis, which allows for immediate data analysis. After the 36th patient, it was clear that there was no difference between DHCA and ACP in terms of new cerebral injury. Preoperatively, 50% of patients had evidence of cerebral injury. Postoperatively, 14 of 18 DHCA patients (78%) had new injury versus 13 of 18 ACP patients (72%) (P=0.66). White matter injury was the most common type of injury in both groups, but central infarctions occurred exclusively after ACP (0 vs. 6/18 [33%]; P=0.02). Early motor and cognitive outcomes at 24 months were assessed and were similar between groups (P=0.28 and P=0.25, respectively). Additional analysis revealed lower postoperative arterial Pco2 as a risk factor for new white matter injury (P=0.04). CONCLUSIONS: In this group of neonates undergoing complex cardiac surgery, we were unable to demonstrate a difference in the incidence of perioperative cerebral injury after ACP compared with DHCA. Both techniques resulted in a high incidence of new white matter injury, with central infarctions occurring exclusively after ACP. CLINICAL TRIAL REGISTRATION URL: http://www.clinicaltrials.gov. Unique identifier: NCT01032876.

Cerebral ischemia initiates an immediate innate immune response in neonates during cardiac surgery.

BACKGROUND: A robust inflammatory response occurs in the hours and days following cerebral ischemia. However, little is known about the immediate innate immune response in the first minutes after an ischemic insult in humans. We utilized the use of circulatory arrest during cardiac surgery to assess this. METHODS: Twelve neonates diagnosed with an aortic arch obstruction underwent cardiac surgery with cardiopulmonary bypass and approximately 30 minutes of deep hypothermic circulatory arrest (DHCA, representing cerebral ischemia). Blood samples were drawn from the vena cava superior immediately after DHCA and at various other time points from preoperatively to 24 hours after surgery. The innate immune response was assessed by neutrophil and monocyte count and phenotype using FACS, and concentrations of cytokines IL-1ß, IL-6, IL-8, IL-10, TNFα, sVCAM-1 and MCP-1 were assessed using multiplex immunoassay. Results were compared to a simultaneously drawn sample from the arterial cannula. Twelve other neonates were randomly allocated to undergo the same procedure but with continuous antegrade cerebral perfusion (ACP). RESULTS: Immediately after cerebral ischemia (DHCA), neutrophil and monocyte counts were higher in venous blood than arterial (P = 0.03 and P = 0.02 respectively). The phenotypes of these cells showed an activated state (both P <0.01). Most striking was the increase in the 'non-classical' monocyte subpopulations (CD16(intermediate); arterial 6.6% vs. venous 14%; CD16+ 13% vs. 22%, both P <0.01). Also, higher IL-6 and lower sVCAM-1 concentrations were found in venous blood (both P = 0.03). In contrast, in the ACP group, all inflammatory parameters remained stable. CONCLUSIONS: In neonates, approximately 30 minutes of cerebral ischemia during deep hypothermia elicits an immediate innate immune response, especially of the monocyte compartment. This phenomenon may hold important clues for the understanding of the inflammatory response to stroke and its potentially detrimental consequences. TRIAL REGISTRATION: ClinicalTrial.gov: NCT01032876.

Low-flow antegrade cerebral perfusion attenuates early renal and intestinal injury during neonatal aortic arch reconstruction.

OBJECTIVE: Deep hypothermic circulatory arrest (DHCA) and antegrade cerebral perfusion (ACP) are 2 cardiopulmonary bypass strategies mainly used in aortic arch reconstructions. It has been suggested that during ACP, abdominal organs are better protected than during DHCA owing to partial perfusion via collaterals. We tested this hypothesis using intraoperative near-infrared spectroscopy (NIRS), lactate measurements, and biomarkers for early abdominal injury in neonates undergoing complex aortic arch repair. METHODS: Neonates scheduled for aortic arch reconstruction via median sternotomy between 2009 and 2011 were randomized to either DHCA or ACP. During surgery, regional oxygen saturations of the abdomen were monitored using NIRS. Immediately aafter DHCA or ACP, lactate concentrations from the inferior vena cava were compared with those from the arterial cannula. Postoperatively, biomarkers for early abdominal organ injury were measured in urine. RESULTS: Twenty-five neonates were analyzed (DHCA, n = 12; ACP, n = 13). Procedures were performed at 18°C, and ACP flow was set at 35 to 50 mL · kg(-1) · min(-1). Median abdominal NIRS value during DHCA was 31% (IQR, 28%-41%) whereas during ACP it was 56% (IQR, 34%-64%; P < .01 between groups). Immediately after DHCA, median lactate from the inferior vena cava was 4.2 mmol/L (IQR, 3.3-5.3 mmol/L) compared with 3.1 mmol/L (IQR, 2.9-4.4 mmol/L) after ACP (P = .03). Postoperatively, biomarkers for renal and intestinal damage (gluthatione s-transferase and intestinal fatty acid binding protein, respectively) were higher in the DHCA group than for the ACP group (P = .03, P = .04, respectively). CONCLUSIONS: These results substantiate earlier suggestions that ACP provides more abdominal organ protection than DHCA in neonates undergoing aortic arch reconstruction.

Bedside prediction rule for infections after pediatric cardiac surgery.

PURPOSE: Infections after pediatric cardiac surgery are a common complication, occurring in up to 30% of cases. The purpose of this study was to develop a bedside prediction rule to estimate the risk of a postoperative infection. METHODS: All consecutive pediatric cardiac surgery procedures between April 2006 and May 2009 were retrospectively analyzed. The primary outcome variable was any postoperative infection, as defined by the Center of Disease Control (2008). All variables known to the clinician at the bedside at 48 h post cardiac surgery were included in the primary analysis, and multivariable logistic regression was used to construct a prediction rule. RESULTS: A total of 412 procedures were included, of which 102 (25%) were followed by an infection. Most infections were surgical site infections (26% of all infections) and bloodstream infections (25%). Three variables proved to be most predictive of an infection: age less than 6 months, postoperative pediatric intensive care unit (PICU) stay longer than 48 h, and open sternum for longer than 48 h. Translation into prediction rule points yielded 1, 4, and 1 point for each variable, respectively. Patients with a score of 0 had 6.6% risk of an infection, whereas those with a maximal score of 6 had a risk of 57%. The area under the receiver operating characteristic curve was 0.78 (95% confidence interval 0.72-0.83). CONCLUSIONS: A simple bedside prediction rule designed for use at 48 h post cardiac surgery can discriminate between children at high and low risk for a subsequent infection.

Increasing duration of circulatory arrest, but not antegrade cerebral perfusion, prolongs postoperative recovery after neonatal cardiac surgery.

OBJECTIVE: Deep hypothermic circulatory arrest (DHCA) and antegrade cerebral perfusion (ACP) are 2 cardiopulmonary bypass techniques applied in aortic arch repair. In recent literature, cerebral effects of both techniques have received most attention, whereas the consequences for other organs have not been thoroughly investigated. Therefore, in this study, the impact of duration of DHCA and ACP on postoperative recovery was analyzed in a cohort of neonates undergoing aortic arch reconstruction. METHODS: All consecutive neonates who underwent aortic arch reconstruction from 2004 to 2009 were included in this retrospective study. Length of stay on the intensive care unit (ICU-LOS), duration of mechanical ventilation, inotrope score, and areas under the curve (AUC) for lactate and creatinine were compared with respect to durations of DHCA and ACP, respectively. Correction for confounders was performed using multivariable linear regression. RESULTS: Eighty-three neonates were included, with a 30-day mortality of 4.8%. Longer duration of DHCA was associated with longer ICU-LOS both in univariable and multivariable analyses. Similarly, duration of mechanical ventilation and lactate and creatinine AUCs increased with duration of DHCA. Inotrope score was only associated with DHCA duration in univariable analysis. Duration of ACP did not affect any of the outcome parameters. CONCLUSIONS: Increasing duration of DHCA, but not ACP, during neonatal aortic arch reconstruction prolongs short-term postoperative recovery. This suggests all efforts should be made to reduce the duration of DHCA to the shortest period possible, which may be achieved by exclusive use of ACP or a combination of the 2 perfusion techniques.

Characterization of expression in mice of a transgene containing 3.3 kb of the human lactase-phlorizin hydrolase (LPH) 5' flanking sequence.

BACKGROUND AND AIM: The regulation of human intestinal lactase-phlorizin hydrolase remains incompletely understood. One kb of pig and 2 kb of rat 5'-flanking sequence controls correct tissue, cell, topographic, and villus LCT expression. To gain insight into human LCT expression, transgenic mouse lines were generated from 3.3 kb of human LPH 5' flanking sequence from a lactase persistent individual fused to a human growth hormone (hGH) reporter bounded by an insulator. METHODS: Four lines were identified in which reporter expression was specifically detectable in the intestine and no other organ, two of which demonstrated hGH expression specific to small and large intestine. Quantitative RT-PCR was carried out on proximal to distal segments of small intestine at fetal days 16.5 and 18.5 and at birth, postnatal days 7 and 28 in line 22. RESULTS: In fetal intestine, hGH expression demonstrated a proximal to distal gradient similar to that in native intestine. There was no significant difference between hGH expression levels at 7 and 28 days in segment 3, the midpoint of the small intestine, where expression of endogenous lactase is maximal at 7 days and declines significantly by 28 days. Distal small intestine displayed high levels of hGH expression in enteroendocrine cells, which were shown to be a subset of the PYY cells. CONCLUSIONS: Thus, a 3.3-kb LPH 5' flanking sequence construct from a lactase persistent individual is able to maintain postnatal expression in transgenic mice post weaning.

Mouse telomerase reverse transcriptase (mTert) expression marks slowly cycling intestinal stem cells.

The intestinal epithelium is maintained by a population of rapidly cycling (Lgr5(+)) intestinal stem cells (ISCs). It has been postulated, however, that slowly cycling ISCs must also be present in the intestine to protect the genome from accumulating deleterious mutations and to allow for a response to tissue injury. Here, we identify a subpopulation of slowly cycling ISCs marked by mouse telomerase reverse transcriptase (mTert) expression that can give rise to Lgr5(+) cells. mTert-expressing cells distribute in a pattern along the crypt-villus axis similar to long-term label-retaining cells (LRCs) and are resistant to tissue injury. Lineage-tracing studies demonstrate that mTert(+) cells give rise to all differentiated intestinal cell types, persist long term, and contribute to the regenerative response following injury. Consistent with other highly regenerative tissues, our results demonstrate that a slowly cycling stem cell population exists within the intestine.

Generation of mTert-GFP mice as a model to identify and study tissue progenitor cells.

Stem cells hold great promise for regenerative medicine, but remain elusive in many tissues in part because universal markers of "stemness" have not been identified. The ribonucleoprotein complex telomerase catalyzes the extension of chromosome ends, and its expression is associated with failure of cells to undergo cellular senescence. Because such resistance to senescence is a common characteristic of many stem cells, we hypothesized that telomerase expression may provide a selective biomarker for stem cells in multiple tissues. In fact, telomerase expression has been demonstrated within hematopoietic stem cells. We therefore generated mouse telomerase reverse transcriptase (mTert)-GFP-transgenic mice and assayed the ability of mTert-driven GFP to mark tissue stem cells in testis, bone marrow (BM), and intestine. mTert-GFP mice were generated by using a two-step embryonic stem cell-based strategy, which enabled primary and secondary screening of stably transfected clones before blastocyst injection, greatly increasing the probability of obtaining mTert reporter mice with physiologically appropriate regulation of GFP expression. Analysis of adult mice showed that GFP is expressed in differentiating male germ cells, is enriched among BM-derived hematopoietic stem cells, and specifically marks long-term BrdU-retaining intestinal crypt cells. In addition, telomerase-expressing GFP(+) BM cells showed long-term, serial, multilineage BM reconstitution, fulfilling the functional definition of hematopoietic stem cells. Together, these data provide direct evidence that mTert-GFP expression marks progenitor cells in blood and small intestine, validating these mice as a useful tool for the prospective identification, isolation, and functional characterization of progenitor/stem cells from multiple tissues.