A Comparison of Diagnostic Performance of Signal Intensity and Volume Related MRI Multiparameters for Assessing Different Response of Rectal Cancer to Neoadjuvant Chemoradiotherapy

Background: “Wait-and-see”, has been proposed as a possible method of treatment in patients with locally advanced rectal cancer (LARC) after chemoradiotherapy (CRT), MR is important to predict the pathological tumor regression grade(TRG) to preoperative CRT. This study aims to evaluate the diagnostic value of signal intensity (SI) and volume (V) change rate in magnetic resonance imaging (MR) and determine which ones perform best as a potential biomarker for predicting pathological TRG to preoperative CRT in patients with LARC. Methods: A retrospective analysis of 82 patients with LARC, for whom clinical and imaging data were retrieved from our institute was conducted between Oct 2017and Oct 2019. Patients underwent pre- and post-CRT T2-weighted (T2W), diffusion-weighted (DW)/apparent diffusion coecient (ADC) and contrast-enhanced T1-weighted (ceT1W). V, difference of volume between pre-CRT and post-CRT tumor ( △ V), V of tumor reduction rate (% △ V), as well as SI of tumor (SIt), SI of muscle (SIm), relative SI ratio of tumor/muscle (SIR), changed difference SIR between pre- and post-CRT SIR ( △ SIR), SIR of tumor changed rate (% △ SIR) on T2W, ADC and ceT1W were measured. All of LARC after CRT were conrmed pathologically and classifed into histologic TRG: TRG 0 (complete response), TRG 1 (moderate response), TRG 2 (minimal response), TRG 3 (poor response). Descriptive statistics and areas under the receiver operating characteristic curves (ROC) were generated to compare performance of % △ V and % △ SIR on T2W, DW, ceT1W for distinguishing between different pathological TRG. Result: Of the 82 patients, TRG 0 (16), TRG 1 (15), TRG 2 (35), TRG 3 (16).Except for ADC-% △ SIR, the remaining % △ V and % △ SIR on T1W, ADC/DWI, ceT1W showed statistics signicance between four groups. There was and DW-MRI correlated well with histopathological percent tumour regression in the resected specimen, and was superior to post-CRT T2 tumour volumetry[8]. In present study show diagnostic performance of volume on DW was slightly greater than T2W and ceT1W, but no statistic difference between AUC of which parameters; because on DW images, viable tumor remnants are more easily recognized, as they appear hyperintense compared with the low signal intensity (SI) of the surrounding non neoplastic tissue, which is in contrast with previous studies showed post-CRT DW MR volumetry with AUC of 0.93 provided high diagnostic performance in assessing CR and was signicantly more accurate than T2-weighted MR volumetry[13].


Background
The multidisciplinary treatment of locally advancedrectal cancer (LARC) has markedly improved and led to better patient outcomes over the last three decades. The reasons for this are multifactorial, but one important factor is the use of Neoadjuvant chemoradiation (CRT) [1,2]. Tumors after CRT will have different TRG and downsizing, and has been shown to be an independent and important prognosticating factor for survival [3]. Therefore, it is important to predict the tumour regression grade (TRG) before surgery because of approximately 10-30% of rectal cancer patients achieve pathological complete remission (pCR) after CRT [1,2,4]. A wait-and-see policy has been proposed as a possible method of treatment in clinical complete responders [2,4].
MR as optional non-invasive examination has also played an increasingly important role and become the gold standard for rectal cancer staging and assessment of response to neoadjuvant treatment as restaging after CRT [4]. mrTRG as a potential biomarker for predicting pTRG to preoperative CRT and disease-free survival and overall survival in patients with LARC, was reported [1,[5][6][7][8][9][10]. However, mrTRG suffers from different interreader agreement. Siddiqui et al reported a wide range of κ values (0.14-0.82), and the assessment of the response of rectal cancers to chemoradiation therapy may be performed effectively using mrTRG [7].
while Hotker et al and Sclafani et al reported the agreement between mrTRG and pTRG is low (0.10, 0.24, respectively) and mrTRG cannot be used as a surrogate of pTRG [4,8], because it is di cult to distinguish between the signal intensity (SI) of brosis and viable tumors by visual inspection [3,11]. Consequently, the signi cance of quantitative analysesand functional imaging studies, along with morphological evaluations have been proposed to assess the TRG after CRT including SI and volume (V) on MR related pammeters respectively, and showed promising results for predicting pCR [8,[11][12][13][14][15][16]. Meanwhile studies have reported SI and V showed potential limited additional diagnostic value for pathological good responders [2,8,13,17].
In previous study, SI and V was used to discriminate pCR or response groups from non-pCR or non response groups, and not for discrimination between TRG groups. To our knowledge, different TRG after CRT has DFS and OS in patients withLARC [7,9], the determination of a TRG before surgery would in uence the subsequent treatment choice, so an accurate clinical assessment of response becomes essential. However, MR quantitative was evaluated to distinguish different TRG has not been reported, we hypothesized that relative SI (SIR) and V change rate on MR sequence included T2WI, DW/ADC, ceT1W between after and after CRT was also associated with TRG, and to evaluate the diagnostic value of their and determine which ones perform best as a potential biomarker for predicting pTRG to preoperative CRT in patients with LARC.

Materials And Methods
This retrospective study was approved by the relevant institutional review board and the need to obtain informed consent was waived. Clinical data were obtained consecutively from the Hospital Database between october 2017 and october 2019.

Magnetic resonance imaging protocol
All patients were performed atmultiple MR unit included 1.5 and 3.0 T by using a phasedarray body coil, Without any bowel preparation. Oblique axial or axial T2-weighted (T2W), contrast-enhanced T1-weighted (ceT1W) images and ADC images were retrieved from the picture archiving and communication system (PACS, Carestream, Canada). Routine rectal MR protocol and image acquisition parameters are presented in Appendix A1. The rst MRI examination was performed to assess tumor stage, the preoperative MRI to assess treatment response and restage after CRT.

Volumetric Image Evaluation
One gastrointestinal radiologists with respective 7 years of expertise in rectal cancer diagnosis, calculated tumor V and SI by manually tracing the tumor boundaries on the axial images and placing free-hand regions-of-interest (ROIs), which provided the sectional area of the lesion for each tumor-containing section (Fig. 2). On the T2W images, tumor was de ned as areas of isointense or hyperintense signal as compared with the relatively lower hypointense signal of the normal adjacent muscular rectal wall. On the DW images, areas of high SI, compared with the normal bowel wall or background of lower SI tissue, were considered as tumor, portions of the tumor showing a high DWI signal along with a high ADC were avoided, so as not to include T2 shinethrough in our ADC data; On the ceT1W images, Areas of enhanced high SI, compared with the normal bowel wall, were considered as enhanced tumor. On post-CRT T2-weighted MR images, areas of markedly low SI at the location of the primary tumor bed were interpreted as brosis. As the risk for residual tumor in these brotic areas is known to be ± 50%, they were also included in the volumetric and SI measurements [13].
Whole-tumor volume was then calculated by multiplying each crosssectional area by section thickness (Fig. 2V1-V4). Post-CRT measurements were performed with comparison to pre-CRT MR images to ensure ROIs were placed within same axial level of the location of the primary tumor, The ROIs with maximum area of tumor obtained on single sections of axial T2W, ADC, nonceT1W, ceTW. The intestinal lumen and the artifact areas were avoided after ROI selection. In some patients, high signal-intensity zones were not identifed on post-CRT DW images, and then the ROIs were positioned at the location of the tumor bed before CRT (Fig. 3). The SI of iliopsoas muscle (SIm) was used as reference tissue carefully avoiding any intramuscular fat [8,[19][20][21].
The Δwas de ned as the change in V and SI value between pre-CRT and post-CRT measurements, while %Δ was de ned as the change rate. V and SI was calculated using the following formula: Another radiologist with 10 years of expertise in gastrointestinal diagnostics independently assessed each pre and post-treatment MRI, blinded to clinical and histopathological information, patients and MRI image characteristics that were evaluated (Table 1)

Patient and Treatment Characteristics
There was no statistical difference in gender, age, MR unit, tumor differentiation, pre-CRT T stage, interval between pre-and post-CRT MR, interval between post-CRT restaging MR and surgery between TRG groups ( Table 1); The median time between pre-and post-CRT MR, and the restaging MRI and surgery was 92.5 days (range, 31-498 days) and 12 days (range, 1-63 days), respectively. All of patients were staged on MR images pre-CRT as cT3-4 (cT3: 59.77% vs cT4: 40.23%). On the other hand, There was statistical difference in post-CRT MR T restaging and pathological ypT stage between TRG groups (both P < 0.001), respectively, agreement of which was low (Kappa = 0.191); post-CRT T restage show 43.9% of all were downstaged, but in fact, 70.73% of all were downstaged on pathology (Table 2). 2. T2W, DW, ceT1W MR SI value and Volumetry There was no statistical difference in Vpre, SImpre, SItpre, SIRpre, SImposton T2W, DW, ceT1W and nonceT1W between groups, as well as SI, SIR, △SIR on ADC between groups. There was statistical difference in nonceT1W-SItpost (p = 0.035), but no statistical difference in nonceT1W-SIRpost between groups (p = 0.126)(in Appendix A2).

Discussion
The focus of this study was to clarify the value of manual tumor %△V, %△SIR and combined of their estimations obtained by T2W, DWI, ceT1W for predicting TRG after CRT in the patients with LARC, and results show that good diagnostic performance of T2W-%△V, DW-%△V, ceT1W-%△V, T2W-%△SIR, ceT1W-%△SIR, and DW-%△V * T2W-%△SIR for discriminating TRG 0 from TRG 2 and TRG 3, T2W-%△V, DW-%△V, ceT1W-%△V for TRG 0 and TRG 1, T2W-%△SIR, ceT1W-%△SIR for TRG 1and TRG 3; despite there was no statistical difference in AUC of all ROC, DW-%△V * T2W-%△SIR had highest AUC; but all measurement of V and SIR were not helpful for distinguishing between TRG 1 and TRG 2, TRG 2 and TRG 3. To the best of ourknowledge, no prior studies have comprehensively assessed and compared the performances of V change rate and SIR volume change rate on T2W, DWI, ceT1W to predict different grade response after CRT.
In present study, agreement of post-CRT MR T restaging and pathological ypT stagewas low (Kappa = 0.191), this is consistent with the results of some prior studies [3,4,8]. Tumor volume and signal intensity has been proven to be an important prognostic indicator for achange of tumors during CRT and response of after CRT. Van den et al and Doenja et al reported changes in rectal tumor morphology ( brosis) and volume visually evaluated can already be observed during CRT [11,16]. Volume was reported to correlate well with downstaging of rectal cancer [14,25], and Lambregts et al reported post-CRT DWI volumetry offers the best results for the detection of patients with a CR after CRT with an area under the curve of 0.92, sensitivity of 70%, and speci city of 98% [14]. In line with previous studies, volume has good diagnostic performance discriminating TRG 0 from TRG 1, TRG 2 and TRG 3 with AUC of 0.921 ~ 0.984, sensitivity of 75 ~ 93.75%, and speci city of 85.71 ~ 100%.
Hotker el al reported tumour volumetry on post-treatment DCE-MRI and DW-MRI correlated well with histopathological percent tumour regression in the resected specimen, and was superior to post-CRT T2 tumour volumetry [8]. In present study show diagnostic performance of volume on DW was slightly greater than T2W and ceT1W, but no statistic difference between AUC of which parameters; because on DW images, viable tumor remnants are more easily recognized, as they appear hyperintense compared with the low signal intensity (SI) of the surrounding non neoplastic tissue, which is in contrast with previous studies showed post-CRT DW MR volumetry with AUC of 0.93 provided high diagnostic performance in assessing CR and was signi cantly more accurate than T2-weighted MR volumetry [13].
Tumour volume may not or slightly change for poor response after CRT, but may had a brotic transformation that was unidenti ed by visual in tumor [26]. Signal intensity rate on MR parameter by quantitative as monitoring therapy response after CRT and diagnostic value for tumor characterization and differentiation [19][20][21]27]. Wan et al reported T2WI signal intensity related parameters with AUC of 0.694 ~ 0.762, sensitivity of 68.2%~77.3%, speci city of 63.6%~77.0 are potential predictors for pCR in LARC after CRT [12]; Value of intra-tumor heterogeneity evaluated by DW for predicting TRG to CRT in lower rectal and measurements of ADC change induced by CRT may have considerable diagnostic value for the estimation of CR, was reported [27,28]. DCE-MRI in rectal cancer is promising mainly for prediction and assessment of response to CRT [29]. Our results are in contrast with previous studies, %△SIR on T2W, ceT1W is a promising diagnostic tool for CR and non-CR, TRG 1 and TRG 3, respectively. Our ADC-%△SIR results show useless, and this discrepancy in published data is probably due to the use of variety protocols, different ROI selection, and general factors contributing to magnetic eld inhomogeneity suchas pH, hydration status, and susceptibility effects [27,28,30,31].
In previous study, SIR and V on MR sequence was used to discriminate pCR or response groups from non-pCR or non response groups, and not for discrimination between TRG groups [2,3,8,9,12,16,18,27,28,31]. In our study, SIR and V change rate on T2W, DW/ADC, ceT1W rst were evaluated to differentiate different response grade after CRT, and considerable results were obtained that T2W-%△V, DW-%△V, ceT1W-%△V, T2W-%△SIR, ceT1W-%△SIR, and DW-%△V * T2W-%△SIR has higher diagnostic performance with accuracy of 82.35%~100% to predict TRG after pCRT compared to previous result [2, 9, 11-16, 23, 25, 27-29], except ceT1W for TRG 0 and TRG2; and among of all, despite that was no statistical difference, DW-%△V * T2W-%△SIR was highest for TRG with AUCs of 0.954 ~ 1.000, sensitivity of 93.75%~100%, speci city of 97.14 ~ 100%, accuracy of 96.08%~100%. %△V and %△SIR not distinguished TRG 1and TRG 2, TRG 2 and TRG 3, which could be due to that volume and brotic transformation of tumor for grade close to grade has little change or there was a ROI selection deviations after CRT. Volume be able to distinguish TRG 0 and TRG 1, which could be due to that ROI area from TRG 0 smaller than ROI area from TRG 1 which in addition to the tumor bed, volume for ROI containa little tumor.
Despite the interesting ndings, our study has some limitations. First, This was a retrospective study with asmall sample size between TRG groups, which may limit the statistical power and generate a statistical bias. Further validation may be needed by prospective study withlarge sample size to prove our hypothesis. Second, Although all data need to be collected using the same type of MRI and protocol, it was not possible to unify our data because they were obtained from multiple MR unit. Unfortunately, intraobserver and interobserver differences were not evaluated, but SIR, △SIR, △V, %△ and %△SIR on MR sequence of preoperative after CRT for prediction TRG were calculated so as to reduce variability of imaging [2,14,27]. Reproducibility of relative SI is relative better than SI on MR sequence was reported [2, 13, 15, 19-21, 30, 31]. Hotker et al reported DCE-MRI volumetry demonstrating better inter-reader agreement [8]. Third, select of ROI and comparison of MR before and after CRT had subjectivity; to overcome these issues, we evaluated a relatively largest area of cancerous tissue [28]. Blazic et al reported the use of single-section and whole-tumor volume methods had similar accuracy in predicting CR based on post-CRT measurement ADC change and saved time compared to whole-tumor volume methods [27].

Conclusion
In conclusion, in consideration of many limitations such as T2W-%△V, DW-%△V, ceT1W-%△V, T2W-%△SIR, ceT1W-%△SIR, and DW-%△V* T2W-%△SIR with good diagnostic performance showed promising results regarding TRG prediction after CRT in LARC, especially DW-%△V* T2W-%△SIR, which needs to be proven by prospective large sample cohorts. The advantage of establishing preoperative TRG as the reference standard assessment is to effectively offer management speci cally tailored to patients, predict prognosis after CRT, and may include the option of non-operable management or potentially further chemoradiotherapy.  Examples of the pre-CRT and corresponding post-CRT of quantitative volumetric and signal intensity measurements performed in a 63-year-old male patient with advanced rectal cancer who rst was histologically con rmed moderately differentiated adenocarcinoma, diagnosed T4 based on MR and post-CRT diagnosed T2. Whole-tumor volume was calculated by multiplying each crosssectional area by section thickness (V1-V4), and signal intensity of tumor was measured by relatively largest ROI of cancerous tissue obtained on single same sections of axial T2W, ADC, nonceT1W, ceTW, as well as for measurement of signal intensity of muscle (The inset in the upper leftcorner). Post-CRT measurements of signal intensity were performed with comparison to pre-CRT MR images to ensure ROIs were placed within same axial level of the location of the primary tumor. Interval between pre-and post-CR MR was 115 day and interval between post-CRT MR and surgery was 4 day. Compared pre-CRT MR and post-CRT MR, tumor had decreased 71.08% in T2W-%△V, 64.89% in DW-%△V, 68.72% in ceT1W%△V, -27.81% in T2W-%△SIR, -65.24% in ceT1W-%△SIR, respectively, and nal histologically con rmed ypT2 and TRG 2 (P).

Figure 2
Examples of the pre-CRT and corresponding post-CRT of quantitative volumetric and signal intensity measurements performed in a 63-year-old male patient with advanced rectal cancer who rst was histologically con rmed moderately differentiated adenocarcinoma, diagnosed T4 based on MR and post-CRT diagnosed T2. Whole-tumor volume was calculated by multiplying each crosssectional area by section thickness (V1-V4), and signal intensity of tumor was measured by relatively largest ROI of cancerous tissue obtained on single same sections of axial T2W, ADC, nonceT1W, ceTW, as well as for measurement of signal intensity of muscle (The inset in the upper leftcorner). Post-CRT measurements of signal intensity were performed with comparison to pre-CRT MR images to ensure ROIs were placed within same axial level of the location of the primary tumor. Interval between pre-and post-CR MR was 115 day and interval between post-CRT MR and surgery was 4 day. Compared pre-CRT MR and post-CRT MR, tumor had decreased 71.08% in T2W-%△V, 64.89% in DW-%△V, 68.72% in ceT1W%△V, -27.81% in T2W-%△SIR, -65.24% in ceT1W-%△SIR, respectively, and nal histologically con rmed ypT2 and TRG 2 (P).

Figure 3
Examples of the pre-CRT and corresponding post-CRT of quantitative signal intensity measurements performed in a 52-year-old male patient with advanced rectal cancer who rst was histologically con rmed moderately differentiated adenocarcinoma, diagnosed T4 based on MR and post-CRT diagnosed T2.The patient has a well-de ned, almost circular tumor mass (Pre-CRT).
Post-CRT, tumor signal-intensity zones were not identifed on T2W, ADC, nonceT1W, ceT1W images, and then the ROIs were positioned at the location of the tumor bed before CRT, comparison to pre-CRT MR images. Interval between pre-and post-CR MR was 91 day and interval between post-CRT MR and surgery was 9 day. Comparedpre-CRT MR and post-CRT MR, tumor had decreased 97.77% in T2W-%△V, 94.29% in DW-%△V, 81.37% in ceT1W%△V, 46.21% in T2W-%△SIR, -5.76% in ceT1W-%△SIR, respectively, and nal histologically con rmed ypT0 and TRG 0 (P).

Figure 3
Examples of the pre-CRT and corresponding post-CRT of quantitative signal intensity measurements performed in a 52-year-old male patient with advanced rectal cancer who rst was histologically con rmed moderately differentiated adenocarcinoma, diagnosed T4 based on MR and post-CRT diagnosed T2.The patient has a well-de ned, almost circular tumor mass (Pre-CRT). Post-CRT, tumor signal-intensity zones were not identifed on T2W, ADC, nonceT1W, ceT1W images, and then the ROIs were positioned at the location of the tumor bed before CRT, comparison to pre-CRT MR images. Interval between pre-and post-CR MR was 91 day and interval between post-CRT MR and surgery was 9 day. Comparedpre-CRT MR and post-CRT MR, tumor had decreased 97.77% in T2W-%△V, 94.29% in DW-%△V, 81.37% in ceT1W%△V, 46.21% in T2W-%△SIR, -5.76% in ceT1W-%△SIR, respectively, and nal histologically con rmed ypT0 and TRG 0 (P).