LAUSR

AIM To analyse changes in post-neoadjuvant follow-up magnetic resonance imaging (MRI) staging accuracy for malignant adenopathies in rectal cancer, by comparing size criteria with morphological criteria using high-resolution T2-weighted sequences, as well as variations when adding diffusion-weighted imaging. METHODS AND MATERIALS The present study was a cross-sectional study of a database including 46 1.5-T MRI examinations (2011-2016) from patients with biopsy-proven rectal cancer and chemoradiotherapy treatment before surgery. All cases were reviewed by three radiologists individually, who were blinded to any clinical information. The radiologists were experienced in rectal cancer (3-6 years) and evaluated the presence of malignant nodes in each patient. Malignancy was determined using morphological, size (5 mm), and diffusion criteria separately, as well as morphology plus diffusion. Each case was assessed four times: (1) evaluation of morphological criteria; (2) size criteria; (3) evaluation only using diffusion (b-values 50, 400, and 800); and (4) diffusion plus morphological criteria. Histological staging of surgical specimens was the reference standard. Statistical analysis included accuracy (area under the receiver operating characteristic [ROC] curve [AUC]), sensitivity, specificity, and positive/negative predictive values (PPV/NPV) for each radiologist, and group agreement (Fleiss' kappa). RESULTS Mean values using morphological criteria were: AUC 0.78, sensitivity 77.7%, specificity 73.8%, PPV 66.1%, NPV 85.2%. Using size criterion: AUC 0.75, sensitivity 62.9%, specificity 83.2%, PPV 74.1%, NPV 80%. Added diffusion yielded no improvement, and yielded worse results by itself. CONCLUSIONS Although morphological criteria showed better results in accuracy, sensitivity, and NPV, size criterion yielded the best specificity and PPV. Adding diffusion did not demonstrate a clear advantage over the criteria by themselves. Thus, mixed size-morphology criteria could have the greatest diagnostic value for follow-up N-staging.