The purpose of this study was to evaluate the potential value of whole-tumour MT imaging to assess tumour response after neoadjuvant chemoradiation for rectal cancer. We compared the mean MTR and various histogram parameters derived from the MTR between patients with a good response after CRT (TRG 1–2) and those with a poor response (TRG 3–5). Results showed that MT imaging of the whole tumour volume after CRT could predict a TRG1–2 at histology with AUCs up to 0.88 (for the 95th percentile), thereby confirming that MTR may be useful for assessing fibrosis.
Interestingly, the best results were obtained for the histogram parameters, and in particular for the 95th percentile, which resulted in high AUCs for both readers 1 and 2 (0.75 and 0.88, respectively). The 95th percentile represents the value above which the highest 5 % of the MTR measurements are within a given ROI. In other words, the 95th percentile represents the most fibrotic areas within the ROI. It is anticipated that the 95th percentile will be higher in patients with a good response (with predominant fibrosis) and lower in patients with a poor response. This is indeed supported by the findings of the current study. It also confirms our secondary hypothesis that, compared to analysis of only mean MTR, histogram data analysis is advantageous in that it gives us better insight into the distribution pattern of MTR and therefore of the heterogeneity of the tumour area. Moreover, with a good intraclass correlation coefficient of 0.80, the 95th percentile appears to be more robust than mean MTR values, which showed only a moderate ICC of 0.50. Interobserver reproducibility may thus be a limiting factor when assessing MTR. For the mean MTR measurements, the difference observed between patients with TRG1–2 and TRG3–5 was, in fact, smaller than the limits of agreement for the difference in MTR measurements between the two readers as derived from the Bland-Altman analysis. Although ICC was considerably better for the 95th percentile, the mean difference between the two response groups was still smaller than the differences in measurements between the two readers. While this is limiting with regard to clinical utility, it may be partially explained by the current study methodology, in which ROI placement was performed directly on the MTR maps (see Fig. 1), which lack anatomical detail, making them challenging for radiologists to interpret. Although corresponding T2-weighted images were used for anatomical reference during the delineation process, it may be worthwhile investigating whether it would be helpful to draw the ROIs and copy them from another sequence, such as the standard T2-weighted images. A second disadvantage of quantitative imaging such as MT imaging is that it can be a somewhat incomprehensible concept for clinicians to interpret. A simpler, qualitative approach such as an MRI-based scoring system may be easier for clinicians to understand. The Royal Marsden research group have investigated the mrTRG, an MRI-based scoring system of tumour regression grade, which is similar in construction to the pathological Mandard model . In their report, the degree of fibrosis determined at pathology was significantly associated with the mrTRG, albeit with moderate interobserver agreement (0.55) . To date, this qualitative approach has not been reproduced or validated by other groups. For future studies, it would be interesting to compare this qualitative approach with MT imaging for the assessment of fibrosis.
The results of the current study are in line with a previous proof-of-principle study in 26 patients in which a single “sample” slice of the MT sequence was obtained after CRT. In that study, the single-slice MTR map was correlated with a corresponding whole-mount histopathology section, and individual areas of fibrosis, tumour, and the normal rectal wall were compared in detail with the underlying histology. This resulted in a significant difference in mean MTR between areas of pathologically proven fibrosis and non-fibrotic areas, including residual tumour . We did not perform such a detailed whole-mount correlation with histopathology, but evaluated the overall MTR of the entire tumour area (obtained from multislice MT imaging) to discriminate between good (TRG1–2) and poor responses (TRG3–5) to assess the value of measuring MTR from a more clinical, patient-based perspective. Our results confirmed that whole-tumour MTR was helpful for differentiating between response groups (i.e. patients with varying degrees of fibrosis). Nevertheless, MTR measurements mainly reflect an estimation of the amount of fibrosis, and specific information regarding the presence and extent of residual tumour cells is limited. In this respect, it might be interesting to combine MT imaging with other tools such as diffusion-weighted MR imaging that specifically target detection of residual tumour in order to obtain a more complete understanding of the overall tumour response. Diffusion-weighted imaging has shown great promise in various studies as a sensitive technique for detecting residual tumour within post-radiation fibrosis, with AUCs in the range of 0.87–0.93 [13, 14, 23, 24]. As such, the two techniques may be of complementary value in assessing the effects of chemoradiotherapy and evaluating the various aspects of treatment response.
The MT ratios in our study are higher than typically reported in the literature. This is likely explained by a combination of factors. First, we used on on-resonance MT pre-pulse, since an on-resonance pulse has been suggested to result in a more prominent MT effect (resulting in higher MT ratios) than that of an off-resonance MT pulse [25, 26], which is more commonly reported in the current literature. Other factors that may have contributed to the relatively high MT ratios in our study include the use of a multislice sequence and the dynamic acquisition of the MT sequence (simultaneous acquisition of images with and without MT pulse in one sequence). The fact that variables such as acquisition parameters can affect the MTR reveals the need for standardization and calibration of the MT-sequence.
There are some limitations to our study design. First, the patient group was relatively small. Second, there were some small differences within the study group in the neoadjuvant treatment scheme. And finally, histopathological confirmation was not obtained in all patients. Eight patients were followed according to a watch-and-wait protocol, with a recurrence-free follow-up period of 12–21 months, which was considered a surrogate endpoint for a complete response (TRG 1).
In conclusion, MT imaging is a promising tool for assessing tumour response after neoadjuvant chemoradiation in rectal cancer. In addition to measurement of mean MTR, histogram data assessment (in particular the 95th percentile) proved beneficial and resulted in high AUCs for differentiating between patients with a good and poor response. A potential limitation of MTR is the limited interobserver agreement, which may be improved by adjusting the measurement protocol, an item that should be addressed by future studies.