145 General Discussion 8 radiation could help in the selection of patients who will benefit most from chemoradiation while at the same time avoiding side effects for those who are expected to be non-responders. It could also open up new opportunities to further tailor neoadjuvant treatment strategies based on the anticipated treatment benefit. To identify potential candidates for organ-preservation after completion of CRT, MRI is typically used in conjunction with endoscopy [14], which has been shown to offer the most reliable diagnostic performance [24]. In the restaging setting, MRI interpretation is hampered by the presence of fibrosis. Still, radiological staging templates for restaging are similar to those used for primary staging. These are primarily ‘TNM’ (Tumour Node Metastases) based [4, 25]. The TNM staging manual does not offer any specific recommendations on how to apply staging when the tumour bed has undergone fibrotic transformation as a result of CRT [26]. Our multicentre and multireader study in Chapter 5 has shown that TNM-based staging – in particular T-staging – is of very limited value in the restaging setting. The overall concordance between the T-stage as assessed on MRI after CRT (the ymrT-stage) and the final ypT-stage at histopathology was only 40-43% and in up to 44% of patients the yT-stage was overstaged. More experienced radiologists showed reasonable accuracy to determine the yT-stage in patients who responded poorly to CRT (i.e. patients with predominant residual tumour), but results for expert as well as less expert readers were very poor when patients showed predominant fibrosis after CT. Restaging reports therefore need to be better tailored to take these limitations into account and answer the key clinical questions to offer useful treatment guidance after CRT. In specific, we need to distinguish between poor responders with gross residual disease who will definitely require surgical resection, and good responders who may be candidates for organ-preservation and require further clinical and endoscopic evaluation to guide final treatment planning. In recent years, different grading systems have been published to visually assess and classify the local tumour response on MRI after CRT [27–30]. In Chapter 6 we set out to compare and validate these different methods among a group of 22 radiologists with different clinical expertise levels. Results showed that methods incorporating diffusion-weighted imaging (DWI) showed the most favourable results when combining diagnostic performance, interreader agreement and reader preference. These findings support previous literature showing that DWI plays a crucial role in response evaluation and should be incorporated into standard protocols [31]. Another important finding of Chapter 6 is that it is important to ensure high-quality imaging and prevent artefacts, as this can significantly affect diagnostic performance. In Chapter 7 we shifted our focus to pre-treatment response prediction. So far, most published literature on pre-treatment response prediction has focused on multiparametric prediction models incorporating quantitative ‘image biomarkers’ such as tumour volume, apparent diffusion coefficient (ADC), perfusion and image texture parameters. Research in this field has further taken flight owing to the rise of radiomics and other
RkJQdWJsaXNoZXIy MjY0ODMw