Search Abstracts | Symposia | Slide Sessions | Poster Sessions
White matter architecture associated with speech production outcomes following surgical removal of brain tumours
Poster Session C, Friday, October 25, 4:30 - 6:00 pm, Great Hall 3 and 4
Greig de Zubicaray1, Kate Drummond2, Emma Murton2, Benjamin Ong3, Rosalind Jeffree4, Katie McMahon1; 1Queensland University of Technology, 2Royal Melbourne Hospital, 3Princess Alexandra Hospital, 4Alfred Hospital
Introduction: Brain tumours are associated with an increased risk of aphasia. Post-operative impairments may occur due to surgical removal of tissue, residual tumour or both.[1] Brain tumours may also result in destruction, displacement and/or deformation of adjacent white matter tracts due to their mass effect.[2] Using diffusion imaging fibre tractography, we explored whether the shape features of 6 language tracts are associated with speech production outcomes post-operatively: arcuate (AF), uncinate (UF), frontal aslant (FAT), inferior longitudinal (ILF), inferior frontal occipital (IFOF) and third subdivision of the superior longitudinal (SLF-III) fasciculi. Methods: 39 patients who underwent surgery to remove a primary left-hemisphere tumour 6 to 24 months previously were scanned with either single-shell high angular resolution diffusion imaging (HARDI; N = 24) or dual shell neurite orientation dispersion and density imaging (NODDI; N = 15) acquisition sequences. Language function was assessed with the Comprehensive Aphasia Test (CAT) which revealed 54% and 32% of patients, respectively, scored below the aphasia cutoff on the verb naming and spoken picture description subtests. Tract shape metrics (total volume, number of streamlines, curl, elongation, surface area, span, mean length and irregularity) were extracted using DSI Studio.[3] Acquisition-specific effects were harmonised using empirical Bayes estimation via neuroCombat,[4] with the parameters estimated from the unaffected right hemisphere applied to the left-hemisphere data. We performed exhaustive regression model searches to identify subsets of tract/shape metrics that predicted scores on the two CAT production tests, followed by k-fold cross-validation to select the best model to avoid overfitting. We then entered sex, age, education, WHO tumour grade and radiotherapy (treated vs. none) as control variables, followed by the best-fitting shape model in a separate step to determine their unique contribution to predicting performance. Results: 37% of the variance in verb naming was significantly predicted by tract features: performance was positively associated with AF volume, FAT curl and IFOF surface area, and negatively associated with IFOF volume, ILF curl and AF surface area. Conversely, 90% of the variance in spoken picture description was predicted by multiple tract features: Performance was positively associated with AF and IFOF volume, FAT and UF length, AF, IFOF, ILF and SLF-III curl, FAT surface area, IFOF, ILF and SLF-II span, and IFOF, ILF number of streamlines. Negative associations were observed with FAT volume, AF, ILF, SLF-III, and UF surface areas, AF and SLF-III number of streamlines, AF, IFOF and ILF length. Conclusions: These results illustrate the patterns of white matter tract architecture responsible for spoken language production outcomes following surgical resection of primary brain tumours. They extend previous reports of lesion deficits by revealing the tract characteristics associated with better outcomes, which may prove useful for prognoses and planning appropriate language therapies following surgery. References: [1] de Zubicaray, G. et al. (2023). Brain and language, 239, 105244. [2] Lucci, G., et al. (2022). Biomechanics and Modeling in Mechanobiology, 21(5), 1483–1509. [3] Swinburn, K., et al., (2004). NY: Psychology Press. [3] Yeh F. C. (2020). NeuroImage, 223, 117329. [4] Fortin, J. et al. (2017). NeuroImage, 161, 149–170.
Topic Areas: Disorders: Acquired, Language Production