Deep reasoning neural network analysis to predict language deficits from psychometry‐driven DWI connectome of young children with persistent language concerns

Abstract

This study investigated whether current state‐of‐the‐art deep reasoning network analysis on psychometry‐driven diffusion tractography connectome can accurately predict expressive and receptive language scores in a cohort of young children with persistent language concerns (n = 31, age: 4.25 ± 2.38 years). A dilated convolutional neural network combined with a relational network (dilated CNN + RN) was trained to reason the nonlinear relationship between “dilated CNN features of language network” and “clinically acquired language score”. Three‐fold cross‐validation was then used to compare the Pearson correlation and mean absolute error (MAE) between dilated CNN + RN‐predicted and actual language scores. The dilated CNN + RN outperformed other methods providing the most significant correlation between predicted and actual scores (i.e., Pearson's R/p‐value: 1.00/<.001 and .99/<.001 for expressive and receptive language scores, respectively) and yielding MAE: 0.28 and 0.28 for the same scores. The strength of the relationship suggests elevated probability in the prediction of both expressive and receptive language scores (i.e., 1.00 and 1.00, respectively). Specifically, sparse connectivity not only within the right precentral gyrus but also involving the right caudate had the strongest relationship between deficit in both the expressive and receptive language domains. Subsequent subgroup analyses inferred that the effectiveness of the dilated CNN + RN‐based prediction of language score(s) was independent of time interval (between MRI and language assessment) and age of MRI, suggesting that the dilated CNN + RN using psychometry‐driven diffusion tractography connectome may be useful for prediction of the presence of language disorder, and possibly provide a better understanding of the neurological mechanisms of language deficits in young children.

Deep learning in neuroimaging of epilepsy

“…In line with the two previous articles, three other works [69][70][71] propose DL models to predict the expressive and receptive language scores of children with epilepsy from the DWI. These authors trained ANN on learning the nonlinear relationship between structural connectivity (input) and language score (output).…”

“…In summary, these seven articles show the potential of tractographybased DL models for surgical planning and for the prediction of postsurgical outcome in patients with DRE. Some papers [67][68][69][70][71] support the hypothesis that tractography-based ANNs have great potential as a biomarker of language function. This approach makes it possible to optimize and individualize therapeutic interventions for the treatment of language dysfunction in children with epilepsy.…”

“…Jeong et al [69] reported R = 0.82 and MAE = 7.77 for expressive scores; and R = 0.75 and MAE = 7.40 for receptive scores. The second study [70] combined two different architectures: a CNN with a relational network (CNN + RN). In this work, the input to the network consisted only of the connectivity values between the expressive and receptive regions delineated by the psychometric-driven DWI.…”

“…Transfer learning can transfer knowledge to a new environment and can overcome the limitations of small medical imaging sample sets for training and testing in ANN, but only three papers applied this approach [42,43,49]. Some works applied data augmentation techniques [40][41][42][43]47,59,[63][64][65][66][69][70][71], however, this can potentially lead to contamination between different data partitions, as data points coming from the same patients can be found in both the validation and test data sets. In some studies that applied this technique, it is not specified whether the data partitioning was done with care to avoid such contaminations.…”