Automated contactless radar-derived movement index for outpatient motor surveillance

Repeatable outpatient screening tools are needed to support early identification of infants at neurodevelopmental risk. We assessed the feasibility of an automated, contactless radar-based movement analysis for outpatient screening in infants without overt neurological concerns.

Infants born at 29–41 weeks of gestation underwent outpatient assessments between 37–60 weeks of postmenstrual age; infants with congenital anomalies, major brain injury, or clinical instability were excluded. Seventy-seven infants contributed 100 assessments. Each visit included a 450-second frequency-modulated continuous-wave (FMCW) radar recording. A pretrained model classified 45 non-overlapping 10-second epochs for asymmetric movements and cramped-synchronized general movements, and the NeuroRiskAbility (NRA) index was computed. NRA-based classification was compared with the outpatient clinical classification based on neurological and developmental assessment, including developmental concerns warranting follow-up.

With prespecified parameters (α = 0.5, cutoff 20), concordance with clinical classification was 58%. Exploratory within-cohort recalibration (α = 0.4, cutoff 24) increased concordance to 93% (κ = 0.86). NRA values showed no monotonic association with postmenstrual age (ρ = 0.045, p = 0.658). Among 23 infants assessed twice, 14 (61%) changed clinical classification between visits.

A brief, non-contact radar-based model can generate an automated movement-based index with high concordance to outpatient clinical classification, supporting feasibility for serial surveillance to identify infants requiring closer follow-up in early infancy.

Contactless frequency-modulated continuous-wave (FMCW) radar enables brief outpatient recordings to quantify spontaneous infant movements and generate an automated movement-based index with high concordance to routine outpatient clinical classification.