Presentation
Search Abstracts | Symposia | Slide Sessions | Poster Sessions | Lightning Talks
Neural speech encoding predicts language outcome in preterm infants: a comparative study of synchronicity and gross-power measures
There is a Poster PDF for this presentation, but you must be a current member or registered to attend SNL 2023 to view it. Please go to your Account Home page to register.
Poster E92 in Poster Session E, Thursday, October 26, 10:15 am - 12:00 pm CEST, Espace Vieux-Port
Nikolay Novitskiy1, Shaoqi Pan1, Ching Man Lai1, Peggy Hiu Ying Chan1, Patrick Wong1; 1Brain and Mind Institute, The Chinese University of Hong Kong
Infants born preterm are more likely to have language developmental problems detected from preschool years (Zimmerman 2018). Infants born preterm are also more likely to show neuroanatomical deficits, particularly those associated with white matter processes (Dibble et al. 2021). A recent study has also found neural functional differences between term and preterm infants (Novitskiy et al. 2023). Using EEG to measure neural encoding of speech, term and preterm infants were found to differ mainly on synchronicity measures of the frequency-following response (FFR), as opposed to gross-power measures. These two types of measures are hypothesized to reflect white and grey matter processes, respectively. In the present study, we ask whether neural encoding of speech measured in infancy can forecast language development in preterm infants as observed in term-born infants (Wong et al. 2021). We further investigate whether synchronicity measures are more predictive of future language development than gross-power measures in preterm infants, due to a higher likelihood of synchronicity measures being disrupted by prematurity. Fifty-five preterm children born before 37 weeks of gestation (32.52±3.57 weeks) participated in the study. Their EEG was recorded during natural sleep at corrected age 6.97±4.02 months while they were passively listening to a sequence of three different Chinese tone syllables via the headphones. The children’s parents provided MacArthur-Bates Communicative Development Inventories (MCDI)-Cantonese version scores on average 7.69 months after the EEG recording, that is at the corrected age of 13.24±3.96 months. Three cross-validated support-vector machine (SVM) binary classification models were constructed: (1) with six FFR synchronicity features, including Pitch Strength, Response Consistency, FFR Signal-to-Noise Ratio (SNR), maximal inter-trial phase coherence (ITPC) and ITPC in mid- and low-frequency ranges, (2) with five FFR gross-power measures including Noise Root-Mean Square (RMS), Power in mid- and low-frequency ranges as well as the ratio of the Power in low- and mid-frequency ranges to the Power in the high-frequency range and (3) with all 11 measures included. The target class for binary classification were the children with vocabulary comprehension score in the lower 16th percentile of the population norm. Statistical significance was assessed with permutation and bootstrapping. Model 1 (with synchronicity features only) showed 0.799±0.060 in accuracy, 0.784±0.094 in sensitivity, 0.801±0.091 in specificity and 0.871±0.064 in area under the receiver-operator curve (AUC). Model 2 (gross-power features only) showed 0.768±0.067 in accuracy, 0.775±0.103 in sensitivity, 0.744±0.107 in specificity and 0.826±0.074 in AUC. Model 3 (all features) yielded 0.773±0.0628 in accuracy, 0.767±0.0994 in sensitivity, 0.762±0.102 in specificity, and 0.89±0.05 in AUC. All three models performed significantly better than chance in a permutation test (p<0.05). Bootstrapping test demonstrated no significant differences for any combination between synchronicity, gross-power and combined-measures models. Neural encoding of speech measured during infancy can predict language outcome in preterm infants. Despite our earlier findings that only synchronicity but not gross-power measures differ between term and preterm infants, both synchronicity and gross-power measures are equally predictive of preterm infants’ language outcome. References: Dibble et al.2021, J.Pediatr.232:48-58.e3.; Novitskiy et al.2023, Dev.Cogn.Neurosci. in press.; Wong et al.2021, Am.J.Speech-Language.Pathol.30(5):2241–50.; Zimmerman 2018, J.Speech.Lang.Hear.Res.61(1):53–65.
Topic Areas: Language Development/Acquisition, Computational Approaches