This paper introduces connectogram modeling of electroencephalography (EEG) signals as a novel approach to represent causal relationships and information flow between different brain regions. Connectograms are graphical representations that map the connectivity between neural nodes or EEG channels through lines and arrows of varying thickness and directionality. Here, interchannel phase connectivity patterns were analyzed by computing Granger causality to quantify the magnitude and direction of causal effects. The resulting weighted, directed connectograms displayed differences in functional integration between individuals with developmental dyslexia versus fluent readers when processing 4.8 Hz amplitude-modulated noise, designed to elicit speech encoding mechanisms. Machine learning classification was subsequently implemented to distinguish participant groups based on characteristic connectivity fingerprints. The methodology integrates signal filtering, instantaneous phase analysis via Hilbert transform, Granger causality computation between all channel pairs, automated feature selection using novel mutual information filtering, construction of directed weighted connectograms, and Gradient Boosting classification. Classification analysis successfully discriminates connectivity patterns, directly implicating theta and gamma bands (AUC 0.929 and 0.911, respectively) resulting from rhythmic auditory stimulation. Results demonstrated altered cross-regional theta
and gamma band oscillatory connectivity in dyslexia during foundational auditory processing, providing perspectives on multisensory and temporal encoding inefficiencies underlying language difficulties