Utilising functional imaging to predict survival in paediatric brain tumours. (12th October 2019)

• Record Type: Journal Article
• Authors: Grist, James; Withey, Stephanie; MacPherson, Lesley; Oates, Adam; Stephen Powell, Mr; Novak, Jan; Abernethy, Laurence; Pizer, Barry; Grundy, Richard; Bailey, Simon; Mitra, Dipayan; Arvantis, Theodoros; Auer, Dorothee; Avula, Shivaram; Peet, Andrew
• Is Part Of: Neuro-oncology. Volume 21(2019)Supplement 4
• Page Start: iv5
• Page End: iv5
• Publication Date: 2019-10-12
• Subjects: Brain Neoplasms -- Periodicals; Brain -- Tumors -- Periodicals; Brain -- Cancer -- Periodicals; Nervous system -- Cancer -- Periodicals; 616.99481
• DOI: 10.1093/neuonc/noz167.018
• Languages: English
• ISSNs: 1522-8517

Abstract

Abstract: Introduction: Brain tumours are a common cause of death in the paediatric population. We have previously shown that MR imaging and spectroscopy can be used to non-invasively differentiate between tumour types. Here, we demonstrate that functional imaging can be highly predictive of survival and grade in a paediatric cohort. Methods: Perfusion (PWI) and diffusion weighted imaging (DWI) were performed in a multi-site (Birmingham Children's Hospital, Royal Victoria Infirmary, Alder Hey, Nottingham) cohort ([grade, 5-year survival alive:dead number] = [I, 15:1], [II, 5:1], [III, 2:3], [IV, 8:11]). ROIs were drawn on T2 imaging and functional imaging features (mean, standard deviation, skewness, and kurtosis) were derived. Supervised machine learning was used to predict 5-year survival and tumour grade from features. ANOVA and post-hoc tests were used to assess differences in features between grade and 5-year survival status. Results: 5-year survival was predicted with 89%, 85%, and 87% accuracy with all imaging, perfusion, or diffusion features, respectively. A significant difference in perfusion was found between surviving and diseased participants (1.71 ± 0.82 vs 2.62 ± 1 mL/100g/min, respectively, p < 0.05). A significant difference in ADC (mm 2 s -1 ) between tumour grades was found (1 vs 4 (1533 ± 458 vs 857 ± 239), 4 vs 3 (857 ± 239 vs 1197 ± 137), 4 vs 2 (857 ± 239 vs 1440 ± 557), corrected p < 0.05). Conclusion: We have shown that perfusion and diffusion imaging features can be used to non-invasively assess tumour grade and estimate 5-year survival status in a cohort of paediatric brain tumours.