The Connectivity Fingerprint of the Fusiform Gyrus Captures the Risk of Developing Autism in Infants with Tuberous Sclerosis Complex

Abstract

Tuberous sclerosis complex (TSC) is a rare genetic disorder characterized by benign tumors throughout the body; it is generally diagnosed early in life and has a high prevalence of autism spectrum disorder (ASD), making it uniquely valuable in studying the early development of autism, before neuropsychiatric symptoms become apparent. One well-documented deficit in ASD is an impairment in face processing. In this work, we assessed whether anatomical connectivity patterns of the fusiform gyrus, a central structure in face processing, capture the risk of developing autism early in life. We longitudinally imaged TSC patients at 1, 2, and 3 years of age with diffusion compartment imaging. We evaluated whether the anatomical connectivity fingerprint of the fusiform gyrus was associated with the risk of developing autism measured by the Autism Observation Scale for Infants (AOSI). Our findings suggest that the fusiform gyrus connectivity captures the risk of developing autism as early as 1 year of age and provides evidence that abnormal fusiform gyrus connectivity increases with age. Moreover, the identified connections that best capture the risk of developing autism involved the fusiform gyrus and limbic and paralimbic regions that were consistent with the ASD phenotype, involving an increased number of left-lateralized structures with increasing age.

Keywords: autism; connectivity fingerprint; fusiform gyrus; tuberous sclerosis complex.

Figure 1

Distribution of AOSI for female and male participants in our patients cohort. There was no evidence of statistically significant difference in AOSI between the groups (P > 0.05).

Figure 2

Synthetic view of the processing pipeline used to compute the fusiform gyrus connectivity fingerprint from anatomical and DW MRI data.

Figure 3

Fully automatic brain parcellation of four different TSC patients, both with FreeSurfer (Version: stable 6.0.0, 23 January 2017) (a-d) and with our multitemplate fusion parcellation approach (e-h).

Figure 4

(a) DCI in a 1-year patient. (b) Zoom in the corona radiata showing DCI's ability. DCI’s ability to capture crossing in the white matter at early as 1 year of age.

Figure 5

The AQ12 filtered tract streamlines for an infant with low AOSI (first row) and high AOSI (second row) together with the corresponding brain structures (FUSG is in yellow) used to compute the fingerprint vectors.

Figure 6

Prediction of the AOSI total score from the DCI fixel-based connectivity fingerprint of the fusiform gyrus at (a) 1 year old, (b) 2 years old, and (c) 3 years old.

Figure 7

ROC curves to get insight into sensitivity and specificity in binarily classify high AOSI scores (AOSI > 10) at (a) 1 year old, (b) 2 years old, and (c) 3 years old.

Figure 8

Control experiment: Ability of the motor system to capture the risk of developing autism. (a) Tractography of the motor system. (b-d) Prediction of the AOSI total score from the connectivity fingerprint at 1, 2 and 3 year old, respectively Note that the flat line in (d) is explained by the inability of the Elastic Net algorithm to find any structure in the data, leading to β_1…β_120 < 1e − 8.