NCO-12RELATIONSHIPS BETWEEN TUMOR COMPARTMENT VOLUMES AND NEUROCOGNITIVE FUNCTIONING IN PATIENTS WITH TEMPORAL LOBE GLIOMA

Abstract

Listen

Translate article

BACKGROUND: To determine whether tumor compartment volumes (necrosis, edema, non-enhancing and enhancing) derived from automated tumor segmentation are predictive of neurocognitive impairment in patients with temporal lobe glioma. METHODS: Treatment naive patients with glioma (n = 49; 67% glioblastoma; 74% left temporal) completed neurocognitive testing [WAIS-III Digit Span (DS); Controlled Oral Word Association (COWA); HVLT-R Total Recall; Trail Making Test A & B (TMT)]. Demographically-adjusted scores that were at or below -1.5 SDs were considered impaired. Presurgical T1, gadolinium-enhanced T1, T2, and FLAIR MRI were used to define 3D tumor boundaries. Using these sequences, 4 normalized tumor compartment volumes (necrosis; edema; non-enhancing; enhancing) were computed with an automated, validated, volumetric segmentation pipeline. Random Forest classifier in Waikato Environment for Knowledge Analysis software (WEKA) was used to predict the impairment for each neurocognitive test using different combinations of the volumetric compartments. RESULTS: Neurocognitive impairment was relatively common across measures [DS, 22%; COWA, 20%; TMT-A, 17%; TMT-B, 54%; HVLT-R Total Recall, 55%]. The gadolinium-enhancing compartment volume alone yielded the best classification accuracy for COWA [AUC = 0.72, TP = 82%, FP = 27%], HVLT-R Total Recall [AUC = 0.71, TP = 65%, FP = 37%], and TMT-A [AUC = 0.51, TP = 75%, FP = 65%]. Non-enhancing compartment volume alone was most predictive of DS [AUC = 0.71, TP = 74%, FP = 46%] and necrosis volume alone was most predictive of TMT-B [AUC = 0.66, TP = 63%, FP = 38%]. Including combinations of varying compartments did not convey predictive advantage over the single predictor models. CONCLUSION: Differing tumor compartment volumes were predictive of neurocognitive impairment across a variety of domains. Specifically, gadolinium-enhancing volume most accurately classified impairment in processing speed, verbal fluency, and verbal learning, non-enhancing volume best classified attentional impairment, and necrosis showed greatest accuracy for executive functioning. These findings suggest that relationships between tumor volume and neurocognitive functioning may vary by tumor compartment.