0258

Classification of Pediatric Brain Tumours using Apparent Diffusion Coefficient – a Multi-Centre Study

Jan Novak^1,2, Niloufar Zarinabad^1,2, Theodoros N Arvanitis³, Lesley MacPherson⁴, Daniel Rodriguez ^5,6, Richard Grundy⁶, Dorothee Auer^7,8, Tim Jaspan⁶, Shivaram Avula⁹, Laurence Abernethy⁹, Patrick Hales¹⁰, Ramneek Kaur¹⁰, Darren Hargrave¹¹, Dipayan Mitra¹², Simon Bailey¹³, Nigel Davies¹⁴, Christopher Clark¹⁰, and Andrew Peet^2,15

¹Institute of Cancer and Genomic Sciences, University of Birmingham, Birmingham, United Kingdom, ²Oncology, Birmingham Children's Hospital, Birmingham, United Kingdom, ³Institute of Digital Healthcare, University of Warwick, Coventry, United Kingdom, ⁴Radiology, Birmingham Children's Hospital, Birmingham, United Kingdom, ⁵Division of Clinical Neuroscience, University of Nottingham, Nottingham, United Kingdom, ⁶The Children‘s Brain Tumour Research Centre, University of Nottingham, Nottingham, United Kingdom, ⁷Sir Peter Mansfield Imaging Centre, University of Nottingham, Nottingham, United Kingdom, ⁸Department of Neuroradiology, Nottingham University Hospital Trusts, Nottingham, United Kingdom, ⁹Radiology, Alder Hey Children’s NHF Foundation Trust, Liverpool, United Kingdom, ¹⁰Developmental Imaging and Biophysics Section, University College London, London, United Kingdom, ¹¹Haematology and Oncology Department, Great Ormond Street Children's Hospital, London, United Kingdom, ¹²The Newcastle upon Tyne Hospitals NHS Foundation Trust, Newcastle, United Kingdom, ¹³Sir James Spence Institute of Child Health, Royal Victoria Infirmary, Newcastle upon Tyne, United Kingdom, ¹⁴Radiation Protection Services, University Hospitals Birmingham NHS Foundation Trus, Birmingham, United Kingdom, ¹⁵Institute of Cancer and Genomic Sciences, University of Birmingham, BIRMINGHAM, United Kingdom

Synopsis

The purpose of the study was to investigate the potential of apparent diffusion coefficient (ADC) to discriminate between paediatric cerebellar tumours on a multicentre basis using histograms. A total of 117 paediatric patients with cerebellar tumours (55 Medulloblastomas, 36 Pilocytic Astrocytomas and 26 Ependymomas) were imaged using diffusion weighted imaging across 12 different hospitals using a total of 18 different scanners. Overall classification accuracies for the ADC histogram metrics were 86 % using Naïve Bayes and 84 % for Random Forest classifier.

Purpose

The purpose of the study was to investigate the potential of apparent diffusion coefficient (ADC) to discriminate between paediatric cerebellar tumours on a multicentre basis using histograms.

Methods

Prior to this study all patients were consented for research to the UK Children’s Cancer and Leukemia Group Functional Imaging Group data base in a UK National Health Service Research Ethics Committee-approved study. Five primary treatment centres provided data for the study including Nottingham, Newcastle, Great Ormond Street, London, Alder Hey Liverpool and Birmingham Children’s Hospital. A total of 117 paediatric patients with cerebellar (55 Medulloblastomas, 36 Pilocytic Astrocytomas and 26 Ependymomas) were imaged using diffusion weighted imaging across 12 different hospitals using a total of 18 different scanners using non-standardised protocols with a range of parameter values. This is the current total of patients on the CCLG database with the diagnoses and DWI. Rarer tumour types found in the cerebellum were excluded. ADC maps were produced using in-house software written in the Python program language using b0 and b1000 images. Regions of interest were drawn manually by a research fellow for the whole tumours on the b0 image excluding areas of large cyst using MRIcro. Data was extracted and placed into 180 bins for histogram analysis. Statistics were evaluated using a combination of the R statistical package (The R Foundation) and SPSS (version 22, IBM). Classification of the tumours was performed using Orange software package(1) with histogram metrics inputted including the Mean, Median, Variance, Skew, Kurtosis and the 10th, 25th, 50th and 75th quantiles. Initial feature selection was undertaken via principal component analysis to account for 95% of the variance. Two different classification methods were employed: 1. Naïve Bayes (NB) and 2. Random Forest (RF). 10-fold cross validation was used to examine the predictive value of the classifiers.

Results

Figure 1 shows example images from the three different tumour types included in this study. The metrics extracted from the tumour ROI histograms are shown in Table 1. An Anova of the group mean, median and the quantiles showed significant differences between all three tumour groups (p < 0.001). As an example the distribution of the mean values is illustrated in the box plot in Figure 2. Average histograms for the three tumour types are shown in Figure 3. The classification results using the histogram metrics are shown in Table 2. The overall (NB: 85% and RF: 84%) and balanced (NB: 89% and RF: 86%) classification accuracies for the two methods were similar. The only notable difference was the lower sensitivity for Ependymomas using NB.

Dicussion

This aim of this study was to determine if ADC values can be used to discriminate between the most common types of paediatric cerebellar tumours on a multicentre basis. We found that there was a highly statistically significant difference between the Mean, Median and the 10th, 25th, 50th and 75th quantiles when observing Pilocytic Astrocytomas, Medulloblastomas and Ependymomas (p < 0.001). These differences are in line with previous studies that have shown significant differences between Medulloblastomas and Ependymomas(2,3) which are usually the two tumour types which overlap in terms of ADC metrics. The average histograms presented in Figure 2 show that there are differences between the three main tumour types if assessed as groups and they have distinct appearances. The standard deviation represented by the error bars, suggests there is the potential for significant overlap on a case-by-case basis. Interestingly, the Medulloblastomas appear to be the most homogeneous tumours with respect to their histograms. The Pilocytic Astrocytomas show an almost bimodal distribution with respect to the shape of the average histogram. We have shown that a cohort of patients scanned on a number of different scanners, in different hospitals and using heterogeneous protocols can still discriminate between tumour types. Our results suggest that rigorous harmonisation of DWI protocols may not be necessary for the production of reliable biomarkers. This has already been demonstrated in healthy volunteers(4). The classification rates are not as high as has previously been seen in the literature for ADC histogram analysis with Rodriguez Gutierrez et al.(2) quoting overall classification rates of 91.4% and Bull et al.(5) 93.75%. However our study is much larger than the aforementioned studies with a more heterogeneous data input with regards to hospitals, scanners and acquisition protocols which may account for the slightly lower rates.

Conclusions

We have shown in this study that we are able to discriminate between the most common types of paediatric cerebellar brain tumour types using histogram analysis of Apparent Diffusion Coefficient maps.

Acknowledgements

This work was funded the National Institute for Health Research (NIHR) by means of a Research Professorship,the Paediatric Experimental Cancer Medicine Centre, and Free Radio in conjunction with Help Harry Help Others (HHHO).

References

1. Demsar J, Curk T, Erjavec A, Gorup C, Hocevar T, Milutinovic M, Mozina M, Polajnar M, Toplak M, Staric A, Stajdohar M, Umek L, Zagar L, Zbontar J, Zitnik M, Zupan B. Orange: Data Mining Toolbox in Python. J Mach Learn Res 2013;14:2349-2353.

2. Rodriguez Gutierrez D, Awwad A, Meijer L, Manita M, Jaspan T, Dineen RA, Grundy RG, Auer DP. Metrics and textural features of MRI diffusion to improve classification of pediatric posterior fossa tumors. AJNR American journal of neuroradiology 2014;35(5):1009-1015.

3. Schneider JF, Confort-Gouny S, Viola A, Le Fur Y, Viout P, Bennathan M, Chapon F, Figarella-Branger D, Cozzone P, Girard N. Multiparametric differentiation of posterior fossa tumors in children using diffusion-weighted imaging and short echo-time H-1-MR spectroscopy. J Magn Reson Imaging 2007;26(6):1390-1398.

4. Grech-Sollars M, Hales PW, Miyazaki K, Raschke F, Rodriguez D, Wilson M, Gill SK, Banks T, Saunders DE, Clayden JD, Gwilliam MN, Barrick TR, Morgan PS, Davies NP, Rossiter J, Auer DP, Grundy R, Leach MO, Howe FA, Peet AC, Clark CA. Multi-centre reproducibility of diffusion MRI parameters for clinical sequences in the brain. NMR in biomedicine 2015;28(4):468-485.

5. Bull JG, Saunders DE, Clark CA. Discrimination of paediatric brain tumours using apparent diffusion coefficient histograms. European radiology 2012;22(2):447-457.

Figures

Figure 1: Example MR images from paediatric brain tumour patients. This first column shows T1-weighted images following the injection of gadolinium contrast agent. The second column shows T2-weighted images and the final column shows apparent diffusion coefficient maps calculated from diffusion-weighted images. (a)-(c) are taken from a patient with a Pilocytic Astrocytoma, (d)-(f) are from a patient with an Ependymoma and (g)-(i) were acquired from a patient with a Medulloblastoma.

Figure 2: Showing some details of the patients included in the study including age and the distribution of the diagnoses. A summary of histogram metrics from the three tumour types is also included in the table

Figure 3: A boxplot showing the distribution of the means from the ADC histograms for Ependymomas, Pilocytic Astrocytomas and Medulloblastomas. Centre lines show the medians; box limits indicate the 25th and 75th percentiles as determined by R software; whiskers extend 1.5 times the interquartile range from the 25th and 75th percentiles, outliers are represented by dots.

Figure 4: Showing average histograms for the most common paediatric cerebellar brain tumours. The error bars show the standard deviation. The bottom right shows the histograms for the different tumour types overlaid with the black representing the Medulloblastoma, the red representing the Ependymoma and the blue representing the Pilocytic Astocytoma.

Figure 5: Shows the classification results using both Naïve Bayes and Random Forest using the Mean, Median, Variance, Skew, Kurtosis and the 10th, 25th, 50th and 75th quantiles extracted from the tumour ROI histograms.

Proc. Intl. Soc. Mag. Reson. Med. 25 (2017)

0258