MRI radiomics features of mesorectal fat can predict response to neoadjuvant chemoradiation therapy and tumor recurrence in patients with locally advanced rectal cancer

Predicting pathologic complete response in locally advanced rectal cancer patients after neoadjuvant therapy: a machine learning model using XGBoost

“…Previous studies have shown that the true pCR rate in those who were assessed as cCR ranged from 44 to 78%, and a proportion of approximately 10% pCR was misjudged as non-cCR. In addition, the sensitivity of biopsies was 12.9%, which was slightly consistent with 30.4% of the surgical specimens [32,33]. Moreover, current restaging methods such as digital rectal examination, endoscopy, computed tomography, MRI, or positron emission tomography-computed tomography were not capable of being surrogate methods of surgical pathological assessment to observe tumor response.…”

MRI‐Based Artificial Intelligence in Rectal Cancer

“…Compared with the definition of pCR based only on the tumor itself, other studies defined pCR as ypT0N0 (evaluating entire tumors and regional lymph nodes), which would be a better criterion considering the application of clinical needs 46 . Several studies 41,43,44 developed radiomic models that also performed high classification (AUC, 0.812–0.89) for predicting pCR defined as ypT0N0, with high specificity, which could be compared with the performance of radiologists 41,44 . The above studies delineated ROIs based on tumors, while several studies mined the information of mesorectal fat based on T2WI and built radiomic models to predict pCR after nCRT in LARC, which revealed that the heterogeneity of mesorectal fat surrounding tumors may also show great predictive power 44,45 .…”

Assessment of clinical and pathological complete response after neoadjuvant chemoradiotherapy in rectal adenocarcinoma and its therapeutic implications

“…In their study, radiomics features of the primary tumor were able to significantly increase the specificity (to 91%) and positive-predictive value (to 72%), while sensitivity and negative-predictive value for pCR were 100% [area under the curve (AUC) of 0.93] [ 43 ]. Moreover, recent studies revealed that MRI radiomics features of the mesorectum have also been associated with increased performance in predicting pCR [ 47 , 48 ] after neoadjuvant CRT in patients with LARC. By using eight MRI radiomic features of the mesorectal fat, Jayaprakasam et al were able to develop a predictive support vector machines model that achieved a diagnostic accuracy for pCR of 83.9%, a sensitivity of 78%, a specificity of 85.1%, a negative predictive value of 94.9%, a positive predictive value of 52.5% and an AUC of 0.89 [ 47 ].…”

“…Moreover, recent studies revealed that MRI radiomics features of the mesorectum have also been associated with increased performance in predicting pCR [ 47 , 48 ] after neoadjuvant CRT in patients with LARC. By using eight MRI radiomic features of the mesorectal fat, Jayaprakasam et al were able to develop a predictive support vector machines model that achieved a diagnostic accuracy for pCR of 83.9%, a sensitivity of 78%, a specificity of 85.1%, a negative predictive value of 94.9%, a positive predictive value of 52.5% and an AUC of 0.89 [ 47 ]. Thus, in present, DW–MRI, associated with MRI-based radiomics of the primary tumor and mesorectal fat can predict pCR with high accuracy, representing an important complementary tool to the DRE and rectoscopy.…”