Predicting the pathological response to chemoradiotherapy of non-mucinous rectal cancer using pretreatment texture features based on intravoxel incoherent motion diffusion-weighted imaging

Abstract

Objectives To investigate the performance of the mean parametric values and texture features based on intravoxel incoherent motion diffusion-weighted imaging (IVIM-DWI) on identifying pathological complete response (pCR) to neoadjuvant chemoradiotherapy (nCRT) in locally advanced rectal cancer (LARC). Methods Pretreatment IVIM-DWI was performed on 41 LARC patients receiving nCRT in this prospective study. The values of IVIM-DWI parameters (apparent diffusion coefficient, ADC; pure diffusion coefficient, D; pseudo-diffusion coefficient, D* and perfusion fraction, f), the first-order, and gray-level co-occurrence matrix (GLCM) texture features were compared between the pCR (n = 9) and non-pathological responder (non-pCR, n = 32) groups. Receiver operating characteristic (ROC) curves in univariate and multivariate logistic regression analysis were generated to determine the efficiency for identifying pCR. Results The values of IVIM-DWI parameters and first-order texture features did not show significant differences between the pCR and non-pCR groups. The pCR group had lower Contrast and DifVarnc values extracted from the ADC, D, and D* maps, respectively, as well as lower Correlat D value. Higher Correlat D* , Correlat f , SumAverg ADC , and SumAverg D values were observed in the pCR group. The area under the ROC curve (AUC) values for the individual predictors in univariate analysis ranged from 0.698 to 0.837, with sensitivities from 43.75% to 87.50% and specificities from 66.67 to 100.00%. In multivariate analysis, Correlat D* (P < 0.001), DifVarnc ADC (P = 0.024), and DifVarnc D (P < 0.001) were the independent predictors to pCR, with an AUC of 0.986, a sensitivity of 93.75%, and a specificity of 100.00%. Conclusion Pretreatment GLCM analysis based on IVIM-DWI may be a potential approach to identify the pathological response of LARC.

Advanced analytics and artificial intelligence in gastrointestinal cancer: a systematic review of radiomics predicting response to treatment

“…In the majority of included studies, the predictive value of radiomics was analyzed in patients diagnosed with colorectal cancer (CRC). A total of 27/60 (45%) studies described prediction of response to treatment in patients with primary CRC, in particular in patients with locally advanced rectal cancer (LARC) [ 37 – 63 ]. Pathologic response to chemoradiotherapy (CRT) is most frequently assessed, and radiomic features are predominantly derived from MRI (Supplementary, Table 1 ).…”

“…The predictive value of individual radiomic features was analyzed in 10 studies, while radiomics-based prediction models were constructed in the remaining 17 studies. Different types of individual radiomic features were associated with complete pathologic response to CRT [ 37 , 39 – 45 , 47 , 63 ]. The predictive value of entropy, kurtosis, skewness, and tumor volume were most frequently described [ 37 , 40 – 44 ].…”

Texture analysis for chemotherapy response evaluation in osteosarcoma using MR imaging

“…To the best of our knowledge there is no study related to statistical TA on IVIM DWI parametric maps for evaluating chemotherapeutic response in osteosarcoma. Prior studies analyzing GLCM features in IVIM DWI parametric maps of locally advanced rectal cancer 30 and nasopharyngeal carcinoma 31 baseline was associated with a better response to chemoradiotherapy. Therefore, statistical textural features quantifying significantly low perfusion and lower angiogenic heterogeneity in tumor might be potential markers of chemoresistance and poor outcome among patients with osteosarcoma at the time of presentation.…”

“…However, textural features are sensitive to image acquisition parameters used in different protocols and to intra‐ and inter‐scanner variations in the case of MRI 49 . To solve this problem, gray‐level normalization filter dynamics limited to μ ± 3 σ , proposed by Collewet et al 44 and supported by prior studies 30,31 performing TA using IVIM‐DWI data, were applied before TA. TA was applied to functional parametric maps as these are in absolute units and unbiased by the scanner and acquisition protocols used.…”

“…Recently, texture analysis (TA) 19 of MR images has shown promising results in evaluating the response to therapy in different types of cancers 20,21 such as brain, 22 breast, [23][24][25] bone, [26][27][28] rectum, 29,30 nasopharynx 31 and soft tissue sarcoma. 32 TA uses mathematical formulae to characterize the spatial distribution of signal intensity variations in an image and evaluate its quantitative parameters.…”