Lung tumour radiotherapy treatment response assessment using Active Breathing Coordinated (ABC) Diffusion-Weighted Magnetic Resonance Imaging

Evangelia Kaza¹, Matthew Blackledge¹, David John Collins¹, Erica Scurr², Helen McNair³, Richard Symonds-Tayler¹, Fiona McDonald², Martin Osmund Leach¹, and Dow-Mu Koh²

¹The Institute of Cancer Research and Royal Marsden Hospital, London, United Kingdom, ²The Royal Marsden NHS Foundation Trust, London, United Kingdom, ³Department of Radiotherapy, Royal Marsden NHS Foundation Trust and Institute of Cancer Research, London, United Kingdom

Synopsis

Imaging with an Active Breathing Coordinator (ABC) modified for MR use was performed on lung cancer patients to acquire spatially matching diffusion-weighted images (DWI) before, during and after Radiotherapy. DWI spatially matched the CT and depicted mediastinal nodal involvement as well as internal tumour heterogeneity. ADC maps provided information about changes in solid and fluid components throughout therapy. Treatment response was evaluated by applying multi-parametric tumour heterogeneity characterisation using Gaussian Mixture Modelling. Differences in ADC and volume behavior of separate cancerous tissue components at various treatment time points may indicate tumour sub-volumes and provide detailed cancer characterisation.

Introduction

An Active Breathing Coordinator (ABC, Elekta Oncology Systems, Crawley, UK), ensuring reproducible breath-holds at a predefined air volume and duration, has been modified for MR use and demonstrated good organ position reproducibility when applied to lung cancer MRI ¹. Given the prospects of Diffusion-Weighted Imaging (DWI) as a diagnostically useful oncologic imaging tool ², we investigated the potential of performing DWI under MR-ABC control for treatment response assessment.

Methods

Eight lung cancer patients were scanned in a 1.5T Siemens Aera using MR-ABC with the same positioning, tattoo alignment and ABC settings as during planning CT. Two patients were scanned before, during and after radiotherapy. In each of these visits, two DW echo planar imaging sequences (EPI1: b 200 smm^-2, 5 mm slice thickness, 2 averages; EPI2: b 100, 400, 750 smm^-2, 6 mm slice thickness, 1 average) were acquired in ABC-controlled breath holds with the same volume threshold applied in radiotherapy (RT). For every MR-ABC visit, apparent diffusion coefficient (ADC) maps were produced from EPI2. Solid tumour regions of interest (ROIs) were drawn on the b100 slices and reviewed by a senior radiologist. Signal intensity on b100 images (SI-b100) was normalised so that the mean value within 5 ROIs drawn around the spinal cord was equalised to 200.

Multi-parametric tumour heterogeneity characterisation ³ was applied to the solid tumour ROIs for each MR-ABC visit. A two-component Gaussian Mixture Model (GMM) was applied to the joint distribution of SI-b100 and ADC, initialised through user-selected seeds on a scatter plot and refined using the Expectation Maximization (EM) algorithm for parameter estimation ⁴. One component was used to model solid disease whereas the other represented outliers due to a different tissue class. The posterior probability of a voxel belonging to one of the two classes was derived, providing tissue classification maps. The ROI area corresponding to each class was noted for every slice and the overall class volume was calculated. Changes of mean ADC values and volumes between the three time points were assessed for each class.

Results

Example results are displayed for a lung adenocarcinoma patient. Fig 1 shows a good spatial agreement without any image registration between a planning CT a) and EPI1 image b) acquired before RT under ABC control, even though the aggressive tumour grew between scans. DWI detected mediastianal nodal involvement, blood vessels within the cancer and small volume pleural effusion, not discernible by CT. Comparing another diagnostic CT slice c) to EPI1 under ABC control d) after RT reveals improved contrast, internal tumour heterogeneity and boundary distinction with EPI1.

ADC maps under ABC control pre, during and post RT (Fig 2) are spatially matched and can be compared to reveal changes in tumour size, fluid content and indicate probable fibrosis following irradiation. Voxel distributions in the solid tumour volume of interest (VOI) demonstrate shape and size variations of the clusters during and after treatment, suggesting tissue alterations following therapy (Fig 3). The derived probability maps (Fig 4) indicate an increase in tissue heterogeneity with irradiation. Table 1 shows a continuous solid component volume reduction throughout treatment, while the volume of the second tissue class increased during treatment before diminishing afterwards. Mean ADC increased during but then remained constant after RT in the solid component. In contrast, the mean ADC of the second class increased during treatment but decreased again to approximately its initial value afterwards. The second to first tissue class ratio increased slightly from 11% pre to 13% through, and decreased to 5% post RT.

Discussion

Performing lung DWI under ABC control provides a good organ and tumour position reproducibility between imaging sessions, which can be applied to study tumour response during treatment. A DW-EPI with high spatial resolution allows morphological cancer imaging, offering better contrast and additional diagnostic information to CT. ADC maps help to characterise lesions and their comparison between treatment time points reveals irradiation-induced effects. Tumour heterogeneity can be assessed using GMM to evaluate treatment response. The differential ADC and volume behavior of different cancerous tissue components at varying RT time points may reflect the heterogeneity of response to therapy.

Conclusion

Spatially matching DW images acquired with ABC control at various treatment stages provide additional diagnostic information to CT and allow for response assessment and quantitative characterisation of tumour heterogeneity regarding response.

Acknowledgements

EPSRC grant EP/H046410/1; CRUK and EPSRC support to the Cancer Imaging Centre at ICR and RMH in association with MRC and Department of Health C1060/A10334, C1060/A16464; CRUK grant C46/A3970 to the ICR Section of Radiotherapy. NHS funding to the NIHR Biomedical Research Centre and the Clinical Research Facility in Imaging. MOL is an NIHR Senior Investigator.

References

1. Kaza E. et al. First application of an Active Breathing Coordinator. Phys Med Biol. 2015; 60:1681-96.

2. Turkbey B. et al. Diffusion-weighted MRI for detecting and monitoring cancer: a review of current applications in body imaging. Diagn Interv Radiol 2012; 18:46-59.

3. Rata M. et al. Whole body quantitative, multi-parametric characterisation of tumour heterogeneity for response evaluation. Proc. 21st Intl. Soc. Mag. Reson. Med. 2013, 0592.

4. Pedregosa et al. Scikit-learn: Machine Learning in Python. J Mach Learn Res 2011; 12:2825-2830.

Figures

Fig 1. a) CT of a lung adenocarcinoma patient under ABC control (17 days pre-RT). b) Corresponding EPI1 image in ABC breath holds with the same settings (5 days pre-RT). c) Same patient CT without ABC (86 days post-RT). d) Corresponding EPI1 image under ABC control (73 days post-RT).

Fig 2. ADC maps under ABC control of a lung adenocarcinoma patient treated with 64 Gy in 32 fractions over 45 days; a) 5 days before, b) 23 days through, c) 3 months post-RT. Comparison suggests solid tumour shrinkage, fluid content increase and possible fibrosis. No image registration was applied.

Fig 3. Scatter plots of normalised b100 image intensity (SI-b100) against the ADC value of each voxel in the solid tumour VOI of the same lung adenocarcinoma patient before, during and after RT. The centers of the two assumed clusters appear in red and green, as initialised by the user.

Fig 4. Colour-coded tissue probability maps overlaid on the zoomed ADC maps of matched slices under ABC control for the same patient a) before, b) through, c) after RT. Red and green represent two tissue classes: solid tumour and outliers due to tissue heterogeneity, respectively.

Table 1. Mean ADC value and volume of each tissue class for every MR-ABC visit of the same lung adenocarcinoma patient, and their percentage differences between visits.

Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)

2800