Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Comparing Indirect Tract-Based Disconnection Metrics for Post-Stroke Language Deficits: A Methodological Evaluation of LQT and BCBToolkit
Poster Session A, Friday, September 12, 11:00 am - 12:30 pm, Field House
Cong Du1, Nina F. Dronkers1,2, Maria V. Ivanova1; 1University of California, Berkeley, 2University of California, Davis
Understanding the extent of white matter tract damage is critical for predicting post-stroke language deficits. Indirect structural disconnection-symptom mapping methods—using structural lesion data and probabilistic white matter atlases—have become increasingly popular as practical alternatives to direct methods such as diffusion imaging. Previous studies using different indirect metrics have reported mixed findings regarding the association between tract damage and language deficits, suggesting that the predictive power of these metrics may vary across linguistic domains. Variability in algorithmic implementations across studies introduces further discrepancies, potentially contributing to inconsistent results. The present study aims to directly compare three widely used indirect disconnection metrics—tract lesion load (percent of the tract covered by the lesion), binary disconnection (whether the tract is disconnected or not), and severity of disconnection (percent of streamlines damaged by the lesion)—as implemented in two popular toolkits: the Lesion Quantification Toolkit (LQT; Griffis et al., 2021) and the Brain Connectivity and Behavior Toolkit (BCBToolkit; Foulon et al., 2018), providing systematic methodological guidance for future research. We analyzed data from 113 individuals with aphasia resulting from left hemisphere stroke and examined the relationship between white matter tract damage and post-stroke language deficits. Language performance was assessed using the five subtest scores from the Western Aphasia Battery–Revised (WAB-R; Kertesz, 2007). Indirect disconnection metrics were derived from LQT and BCBToolkit for six language-relevant tracts: the arcuate fasciculus (AF), superior longitudinal fasciculus (SLF), inferior longitudinal fasciculus (ILF), inferior fronto-occipital fasciculus (IFOF), frontal aslant tract (FAT), and uncinate fasciculus (UF). In a subset of 32 participants with available diffusion-weighted imaging (DWI), we additionally extracted direct metrics of tract integrity—volume, normalized volume, and hindrance modulated orientational anisotropy (HMOA)—to assess their correspondence with the indirect estimates. The three indirect disconnection metrics yielded significantly different estimates of tract damage across all tracts, underscoring substantial methodological variability in disconnection quantification across toolkits. In relation to behavior, both severity of disconnection and tract lesion load explained unique variance in WAB-R scores beyond lesion volume for the ILF and IFOF. For the SLF, only severity of disconnection showed a significant association, while tract lesion load emerged as the sole predictor for the AF and FAT. In the DWI subsample, only HMOA was significantly associated with language outcomes after controlling for lesion volume. Among the indirect measures, tract lesion load demonstrated the strongest consistency with HMOA, suggesting greater alignment with diffusion-based indices of tract integrity. These findings demonstrate that methodological choices—such as the selection of disconnection metric and analysis toolkit—as well as the tract under investigation, can substantially influence the strength and specificity of structure–function relationships in post-stroke aphasia. While tract lesion load aligned more closely with DTI-derived measures of tract integrity and showed more consistent associations with language outcomes than the other metrics, its predictive power varied depending on the tract and the linguistic domain assessed. By comparing multiple metrics across toolkits and tracts, our study provides practical guidance for applying indirect disconnection methods and underscores the need to match measurement strategies to research aims and neuroanatomical targets.
Topic Areas: Methods,