Although the introduction of phased-array coil systems has improved the accuracy of MRI in staging rectal cancer, even more recent studies report accuracies of only 67–86 % for T-staging [
53‐
56]. These disappointing results are primarily due to the poor differentiation of T1/2 cancer from so-called borderline T3 cancer, where it is often not possible to distinguish true mesorectal tumor invasion from desmoplastic reactions (Fig.
3) [
49,
54,
57]. Desmoplastic reactions are reactive tissue alterations which often occur in the immediate surrounding of tumors, most frequently resulting in fibrotic extensions that may contain tumor cells or not. The failure to differentiate between desmoplastic reactions and tumor growth is not specific to MRI but is also a well-known problem in rectal cancer staging with EUS [
27]. Clinically and therapeutically, however, this differentiation is of minor importance. As already mentioned, it is much more important to precisely describe the relationship of the tumor to the mesorectal fascia, representing the anticipated resection plane for TME in order to assess the likelihood of a tumor-free CRM. Several recent studies have confirmed that MRI is highly suited to provide this information [
54,
57‐
60]. In a study of 43 patients, Bissett et al. [
59] found good agreement between preoperative MRI and histopathology with regard to the demonstration of tumor penetration through the mesorectal fascia (accuracy: 95%). These results are underlined by the studies of Beets-Tan et al. [
54,
61], who investigated 76 patients and likewise found preoperative MRI to be highly accurate in assessment of the CRM. The agreement was 100% in T4 tumors, and 97% and 93% for both readers in tumors with a histologically determined tumor-free CRM >10 mm. Regression analysis for histologically determined margins of 1–10 mm demonstrated that a tumor-free resection margin of 2 mm was predicted with an accuracy of 97% if the distance between tumor and mesorectal fascia measured by MRI was at least 6 mm. It is noteworthy that this study likewise showed only moderate results with regard to T-staging (accuracy of 83% and 67% for the two readers) [
54,
61]. In a study of 98 patients published by Brown and co-workers in 2003, the agreement between MRI and histology in assessment of the CRM was 92% [
60]. These figures indicate that MRI allows accurate prediction of the CRM status after resection. The expected CRM can be described as involved if tumor invasion of the mesorectal fascia is visible or the tumor has a proximity of 1 mm or less to the mesorectal fascia. A tumor-free CRM can be assumed with a high degree of accuracy if the shortest distance from the maximum tumor extension, a mesorectal tumor deposit or a suspect lymph node in the mesorectum is more than 6 mm [
54]. The role of tumors that extend towards the mesorectal fascia to a distance of less than 5 mm on MR images remains controversial.