Development and Validation of a Deep Learning Method to Predict Cerebral Palsy From Spontaneous Movements in Infants at High Risk.

Importance: Early identification of cerebral palsy (CP) is important for early intervention, yet expert-based assessments do not permit widespread use, and conventional machine learning alternatives lack validity. Objective: To develop and assess the external validity of a novel deep learning-based method to predict CP based on videos of infants' spontaneous movements at 9 to 18 weeks' corrected age. Design, Setting, and Participants: This prognostic study of a deep learning-based method to predict CP at a corrected age of 12 to 89 months involved 557 infants with a high risk of perinatal brain injury who were enrolled in previous studies conducted at 13 hospitals in Belgium, India, Norway, and the US between September 10, 2001, and October 25, 2018. Analysis was performed between February 11, 2020, and September 23, 2021. Included infants had available video recorded during the fidgety movement period from 9 to 18 weeks' corrected age, available classifications of fidgety movements ascertained by the general movement assessment (GMA) tool, and available data on CP status at 12 months' corrected age or older. A total of 418 infants (75.0%) were randomly assigned to the model development (training and internal validation) sample, and 139 (25.0%) were randomly assigned to the external validation sample (1 test set). Exposure: Video recording of spontaneous movements. Main Outcomes and Measures: The primary outcome was prediction of CP. Deep learning-based prediction of CP was performed automatically from a single video. Secondary outcomes included prediction of associated functional level and CP subtype. Sensitivity, specificity, positive and negative predictive values, and accuracy were assessed. Results: Among 557 infants (310 [55.7%] male), the median (IQR) corrected age was 12 (11-13) weeks at assessment, and 84 infants (15.1%) were diagnosed with CP at a mean (SD) age of 3.4 (1.7) years. Data on race and ethnicity were not reported because previous studies (from which the infant samples were derived) used different study protocols with inconsistent collection of these data. On external validation, the deep learning-based CP prediction method had sensitivity of 71.4% (95% CI, 47.8%-88.7%), specificity of 94.1% (95% CI, 88.2%-97.6%), positive predictive value of 68.2% (95% CI, 45.1%-86.1%), and negative predictive value of 94.9% (95% CI, 89.2%-98.1%). In comparison, the GMA tool had sensitivity of 70.0% (95% CI, 45.7%-88.1%), specificity of 88.7% (95% CI, 81.5%-93.8%), positive predictive value of 51.9% (95% CI, 32.0%-71.3%), and negative predictive value of 94.4% (95% CI, 88.3%-97.9%). The deep learning method achieved higher accuracy than the conventional machine learning method (90.6% [95% CI, 84.5%-94.9%] vs 72.7% [95% CI, 64.5%-79.9%]; P < .001), but no significant improvement in accuracy was observed compared with the GMA tool (85.9%; 95% CI, 78.9%-91.3%; P = .11). The deep learning prediction model had higher sensitivity among infants with nonambulatory CP (100%; 95% CI, 63.1%-100%) vs ambulatory CP (58.3%; 95% CI, 27.7%-84.8%; P = .02) and spastic bilateral CP (92.3%; 95% CI, 64.0%-99.8%) vs spastic unilateral CP (42.9%; 95% CI, 9.9%-81.6%; P < .001). Conclusions and Relevance: In this prognostic study, a deep learning-based method for predicting CP at 9 to 18 weeks' corrected age had predictive accuracy on external validation, which suggests possible avenues for using deep learning-based software to provide objective early detection of CP in clinical settings.

The Predictive Accuracy of the General Movement Assessment for Cerebral Palsy: A Prospective, Observational Study of High-Risk Infants in a Clinical Follow-Up Setting.

BACKGROUND: Early prediction of cerebral palsy (CP) using the General Movement Assessment (GMA) during the fidgety movements (FM) period has been recommended as standard of care in high-risk infants. The aim of this study was to determine the accuracy of GMA, alone or in combination with neonatal imaging, in predicting cerebral palsy (CP). METHODS: Infants with increased risk of perinatal brain injury were prospectively enrolled from 2009-2014 in this multi-center, observational study. FM were classified by two certified GMA observers blinded to the clinical history. Abnormal GMA was defined as absent or sporadic FM. CP-status was determined by clinicians unaware of GMA results. RESULTS: Of 450 infants enrolled, 405 had scorable video and follow-up data until at least 18-24 months. CP was confirmed in 42 (10.4%) children at mean age 3 years 1 month. Sensitivity, specificity, positive and negative predictive values, and accuracy of absent/sporadic FM for CP were 76.2, 82.4, 33.3, 96.8, and 81.7%, respectively. Only three (8.1%) of 37 infants with sporadic FM developed CP. The highest accuracy (95.3%) was achieved by a combination of absent FM and abnormal neonatal imaging. CONCLUSION: In infants with a broad range of neonatal risk factors, accuracy of early CP prediction was lower for GMA than previously reported but increased when combined with neonatal imaging. Sporadic FM did not predict CP in this study.

Computer-based video analysis identifies infants with absence of fidgety movements.

BackgroundAbsence of fidgety movements (FMs) at 3 months' corrected age is a strong predictor of cerebral palsy (CP) in high-risk infants. This study evaluates the association between computer-based video analysis and the temporal organization of FMs assessed with the General Movement Assessment (GMA).MethodsInfants were eligible for this prospective cohort study if referred to a high-risk follow-up program in a participating hospital. Video recordings taken at 10-15 weeks post term age were used for GMA and computer-based analysis. The variation of the spatial center of motion, derived from differences between subsequent video frames, was used for quantitative analysis.ResultsOf 241 recordings from 150 infants, 48 (24.1%) were classified with absence of FMs or sporadic FMs using the GMA. The variation of the spatial center of motion (CSD) during a recording was significantly lower in infants with normal (0.320; 95% confidence interval (CI) 0.309, 0.330) vs. absence of or sporadic (0.380; 95% CI 0.361, 0.398) FMs (P<0.001). A triage model with CSD thresholds chosen for sensitivity of 90% and specificity of 80% gave a 40% referral rate for GMA.ConclusionQuantitative video analysis during the FMs' period can be used to triage infants at high risk of CP to early intervention or observational GMA.

A change in temporal organization of fidgety movements during the fidgety movement period is common among high risk infants.

AIM: General movement assessment (GMA) at 9-20 weeks post-term, can effectively predict cerebral palsy. Our aim was to evaluate intra-individual variability of the temporal organization of fidgety movements (FMs) in high risk infants. MATERIAL AND METHODS: 104 High risk infants (66 males) with at least two video recordings from the FMs period participated. 45 of the infants had GA <28 weeks and/or BW ≤800 g. Mean post-term age at first and second assessments was 11.0 (8-16) and 14.0 (11-17) weeks, respectively, and median time-difference between the assessments was 2.0 (range: three days to six weeks) weeks. Video recordings were analyzed according to Prechtl's GMA. RESULTS: 33 (32%) Infants were classified differently at first and second assessments. Six infants (6%) changed from normal to abnormal, and 10 (10%) changed from abnormal to normal FMs. Seven of the ten who changed classification from abnormal to normal were born before GA 26 weeks. A change between intermittent and continual, which are both considered normal, was observed in 17 (16%) infants. CONCLUSION: A change in temporal organization of FMs is common in high risk infants. Especially in extremely preterm infants with abnormal FMs, more than one assessment should be performed before long-term prognosis is considered.

High prevalence of abnormal motor repertoire at 3 months corrected age in extremely preterm infants.

AIMS: To compare early motor repertoire between extremely preterm and term-born infants. An association between the motor repertoire and gestational age and birth weight was explored in extremely preterm infants without severe ultrasound abnormalities. METHODS: In a multicentre study, the early motor repertoire of 82 infants born extremely preterm (ELGAN:<28 weeks) and/or with extremely low birth weight (ELBW:<1000 g) and 87 term-born infants were assessed by the "Assessment of Motor Repertoire - 2 to 5 Months" (AMR) which is part of Prechtl's "General Movement Assessment", at 12 weeks post-term age. Fidgety movements were classified as normal if present and abnormal if absent, sporadic or exaggerated. Concurrent motor repertoire was classified as normal if smooth and fluent and abnormal if monotonous, stiff, jerky and/or predominantly fast or slow. RESULTS: Eight-teen ELBW/ELGAN infants had abnormal fidgety movements (8 absent, 7 sporadic and 3 exaggerated fidgety movements) compared with 2 control infants (OR:12.0; 95%CI:2.7-53.4) and 46 ELBW/ELGAN infants had abnormal concurrent motor repertoire compared with 17 control infants (OR:5.3; 95%CI:2.6-10.5). Almost all detailed aspects of the AMR differed between the groups. Results were the same when three infants with severe ultrasound abnormalities were excluded. In the remaining ELBW/ELGAN infants, there was no association between motor repertoire and gestational age or birth weight. CONCLUSION: ELBW/ELGAN infants had poorer quality of early motor repertoire than term-born infants.The findings were not explained by severe abnormalities on neonatal ultrasound scans and were not correlated to the degree of prematurity. The consequences of these abnormal movement patterns remain to be seen in future follow-up studies.

Pulse oximetry screening as a complementary strategy to detect critical congenital heart defects.

OBJECTIVE: To compare strategies with and without first-day of life pulse oximetry screening to detect critical congenital heart defects (CCHDs). STUDY DESIGN: Population based study including all live born infants in Norway in 2005 and 2006 (n = 116 057). Postductal (foot) arterial oxygen saturation (SpO(2)) was measured in apparently healthy newborns after transferral to the nursery, with SpO(2) < 95% as cut-off point. Out of 57 959 live births in the hospitals performing pulse oximetry screening, 50 008 (86%) were screened. RESULTS: A total of 136 CCHDs (1.2 per 1000) were diagnosed, 38 (28%) of these prenatally. Of the CCHDs detected after birth, 44/50 (88%) were detected before discharge in the population offered pulse oximetry screening (25 by pulse oximetry), compared to 37/48 (77%) in the non-screened population (p = 0.15). Median times for diagnosing CCHDs in-hospital before discharge were 6 and 16 h after birth respectively (p < 0.0001). In the screened population 6/50 (12%) CCHDs were missed and recognized after discharge because of symptoms. Two of the six missed cases failed the pulse oximetry screening, but were overlooked (echocardiography not performed before discharge). If these cases had been recognized, 4/50 (8%) would have been missed compared to 11/48 (23%) in the non-screened population (p = 0.05). Of the cases missed, 14/17 (82%) had left-sided obstructive lesions. CONCLUSION: First-day of life pulse oximetry screening provides early in-hospital detection of CCHDs and may reduce the number missed and diagnosed after discharge.

First day of life pulse oximetry screening to detect congenital heart defects.

OBJECTIVE: To evaluate the efficacy of first day of life pulse oximetry screening to detect congenital heart defects (CHDs). STUDY DESIGN: We performed a population-based prospective multicenter study of postductal (foot) arterial oxygen saturation (SpO(2)) in apparently healthy newborns after transfer from the delivery suite to the nursery. SpO(2) < 95% led to further diagnostic evaluations. Of 57,959 live births, 50,008 (86%) were screened. In the screened population, 35 CHDs were [corrected] classified as critical (ductus dependent, cyanotic). CHDs were prospectively registered and diagnosed in 658/57,959 (1.1%) [corrected] RESULTS: Of the infants screened, 324 (0.6%) failed the test. Of these, 43 (13%) had CHDs (27 critical), and 134 (41%) had pulmonary diseases or other disorders. The remaining 147 infants (45%) were healthy with transitional circulation. The median age for babies with CHDs at failing the test was 6 hours (range, 1-21 hours). For identifying critical CHDs, the pulse oximetry screening had a sensitivity rate of 77.1% (95% CI, 59.4-89.0), specificity rate of 99.4% (95% CI, 99.3-99.5), and a false-positive rate of 0.6% (95% CI, 0.5-0.7). CONCLUSIONS: Early pulse oximetry screening promotes early detection of critical CHDs and other potentially severe diseases. The sensitivity rate for detecting critical CHDs is high, and the false-positive rate is low.