Multi-parametric MRI at 3.0 Tesla for the Prediction of Treatment Response in Rectal Cancer
Trang Pham1,2,3, Michael Barton1,2,3, Dale Roach4, Karen Wong1,2,3, Daniel Moses2,5, Christopher Henderson2,6,7, Mark Lee1, Robba Rai1, Benjamin Schmitt8, and Gary Liney1,3,9,10

1Radiation Oncology, Liverpool Hospital, Sydney, Australia, 2Faculty of Medicine, University of New South Wales, Sydney, Australia, 3Ingham Institute for Applied Medical Research, Sydney, Australia, 4Faculty of Physics, University of Sydney, Sydney, Australia, 5Radiology, Prince of Wales Hospital, Sydney, Australia, 6Anatomical Pathology, Liverpool Hospital, Sydney, Australia, 7Faculty of Medicine, Western Sydney University, Sydney, Australia, 8Siemens Healthcare Pty Ltd, Sydney, Australia, 9Faculty of Radiation and Medical Physics, University of Wollongong, Wollongong, Australia, 10University of New South Wales, Sydney, Australia

Synopsis

A complete protocol using quantitative diffusion weighted imaging (DWI) and dynamic contrast enhanced (DCE) imaging in combination, and a voxel-by-voxel histogram analysis strategy was successfully developed for multi-parametric MRI prediction of treatment response in rectal cancer. In good responders, the week 3 histograms showed a combined shift in distribution of ADC of voxels to higher values and Ktrans of voxels to lower values compared to the pre-CRT. Multi-parametric histogram analysis of ADC and Ktrans appears to be a promising and feasible method of assessing tumour heterogeneity and its changes in response to CRT in rectal cancer.

PURPOSE

Current functional MRI techniques have shown promising results for prediction of response to chemoradiotherapy (CRT) in rectal cancer, but lack sufficient accuracy for clinical use. There is a wide variation in performance of functional MRI in response prediction reported. Most studies describe single parameter values from either diffusion or perfusion MRI. Single parameter measurements, such as mean ADC or Ktrans, do not reflect tumour heterogeneity. The purpose of this study was to prospectively evaluate the role of quantitative diffusion weighted imaging (DWI) and dynamic contrast enhanced (DCE) imaging used in combination for multi-parametric voxel-wise prediction of treatment response in rectal cancer.

METHODS

This study used a voxel-by-voxel multi-parametric histogram analysis strategy to assess tumour heterogeneity and its changes in response to combined chemotherapy and radiotherapy (CRT). Twenty patients with locally advanced rectal cancer undergoing preoperative CRT prospectively underwent MRI on a 3T wide bore Siemens Skyra at 3 time-points: Pre-CRT, week 3 CRT, and post-CRT. The study protocol consisted of: (i) T2-weighted images (ii) DWI using RESOLVE, which has been previously shown to be robust with respect to geometrical distortions1. Images were acquired with b-values 50 and 800s/mm2 and 1 & 3 averages. ADC maps and calculated b=1400s/mm2 images were produced as part of protocol. (iii) DCE consisted of pre-contrast VIBE scans with flip angles 2º and 15º in order to calculate native T1, followed by gadoversetamide (0.1mM/kg) injection and 60 phases using TWIST with a 5s temporal resolution. ADC and Ktrans parameter maps were registered to T2-weighted images. Semi-automated segmentation was used to define the volume of interest from hyperintense tumour on the b-value=1400 images. A voxel-by-voxel technique was used to produce colour coded histograms of ADC and Ktrans, as well as combined scatterplots for each time-point. CRT response was defined according to histopathology tumour regression grade (TRG) (AJCC 7th Edition)2.

RESULTS

Of 20 patients, 1 had clinical stage T2N2M0, 5 had T3N0M0, 4 had T3N1M0, 7 had T3N2M0, and 3 had T4N2M0. Eight patients had a good response (TRG0-1) and 11 patients had a poor response (TRG2-3) to CRT. Pathology for 1 patient is pending. A complete protocol and analysis strategy was successfully developed which has utilized commercial, in-house developed and works-in-progress (OncoTreat) software. In good responders, the week 3 histograms and maps showed both a shift in distribution of ADC of pixels to higher values and Ktrans of pixels to lower values compared to the pre-CRT histogram. The ADC histograms for a good responder shown in Figure 1 demonstrated an increase in the absolute ADC values of voxels over the time-points. For the same patient, the majority of Ktrans voxel values were high (red) pre-CRT, and by week 3-CRT the Ktrans histogram demonstrated a marked reduction in the absolute Ktrans values of voxels (Figure 2). Figure 3 shows the scatterplots demonstrating changes in combined ADC and Ktrans of voxels of segmented region over the time-points for a good responder and a poor responder.

DISCUSSION

We found the calculated b-value=1400 images useful for visualization of tumour. For DCE analysis, pre-registration of flip angle to dynamic images was a crucial step in producing pixel-by-pixel T1 map, to ensure accurate voxel-by-voxel calculation of Ktrans and had to be performed outside of commercial software owing to its limitations. Patient 1 (good responder) had majority of voxels with high Ktrans pre-CRT. A possible explanation for this is that the high Ktrans is due to a well perfused oxic tumour, which is predictive of good radiotherapy response. In contrast, Patient 2, a poor responder, had low Ktrans values pre-CRT, without much change the in values of voxels over the time-points (Figure 3). The low Ktrans values in this patient may be due to poor perfusion representing a hypoxic tumour, which is predictive of a radio-resistant tumour and poor response.

CONCLUSION

Multi-parametric histogram analysis of ADC and Ktrans appears to be a promising and feasible method of assessing tumour heterogeneity and its changes in response to CRT in rectal cancer.

Acknowledgements

No acknowledgement found.

References

1Liney G et al. Quantitative evaluation of diffusion-weighted imaging techniques for the purposes of radiotherapy planning in the prostate. Br J Radiol. 2015. DOI: http://dx.doi.org/10.1259/bjr.20150034

2Edge S, Byrd D, Compt C et al (Eds). AJCC Cancer staging manual 7th edition. New York Springer 2010

Figures

Figure 1: Patient 1 Good Responder (AJCC TRG1) ADC colour-coded maps and voxel-by-voxel histograms for entire segmented region shown for pre-CRT (top panel), week 3 CRT (middle panel) and post-CRT (bottom panel). Colour code: blue voxels – ADC <1000x10-6mm2/s, green voxels – ADC 1000-1500x10-6mm2/s, red voxels – ADC > 1500x10-6mm2/s.

Figure 2. Patient 1- Good Responder (AJCC TRG1). Ktrans colour-coded maps and voxel-by-voxel histograms for the entire segmented region shown for pre-CRT (top panel), week 3 CRT (middle panel) and post-CRT (bottom panel). Colour code: blue voxels – Ktrans values<100x10-3mL/g/min, green voxels – Ktrans100 -500x10-3 mL/g/min, red voxels – Ktrans > 500x10-3 mL/g/min.

Figure 3: Scatterplots of ADC and Ktrans in combination for every voxel for each time-point. Percentages of voxels in each quadrant is shown with most dominant quadrant highlighted. Top panel: Scatterplots for Patient 1 - Good responder (AJCC TRG 1). Bottom panel: Scatterplots for Patient 2 – Poor Responder (AJCC TRG 2)



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
2646