Chemical Shift-Encoded Sequence (IDEAL-IQ) and Amide Proton Transfer (APT) MRI for Prediction of Histopathological Factors of Rectal Cancer

To investigate whether parameters from IDEAL-IQ/amide proton transfer MRI (APTWI) could help predict histopathological factors of rectal cancer. Preoperative IDEAL-IQ and APTWI sequences of 67 patients with rectal cancer were retrospectively analyzed. The intra-tumoral proton density fat fraction (PDFF), R2* and magnetization transfer ratio asymmetry (MTRasym (3.5 ppm)) were measured according to the histopathological factors of rectal cancer. The relationship between MR parameters and histopathological factors were analyzed, along with diagnostic performance of MR parameters. PDFF, R2* and MTRasym (3.5 ppm) were statistically different between T1+T2/T3+T4 stages, non-metastatic/metastatic lymph nodes, lower/higher tumor grade and negative/positive status of MRF and EMVI (p < 0.001 for PDFF, p = 0.000–0.015 for R2* and p = 0.000–0.006 for MTRasym (3.5 ppm)). There were positive correlations between the above parameters and the histopathological features of rectal cancer (r = 0.464–0.723 for PDFF (p < 0.001), 0.299–0.651 for R2* (p = 0.000–0.014), and 0.337–0.667 for MTRasym (3.5 ppm) (p = 0.000–0.005)). MTRasym (3.5 ppm) correlated moderately and mildly with PDFF (r = 0.563, p < 0.001) and R2* (r = 0.335, p = 0.006), respectively. PDFF provided a significantly higher diagnostic ability than MTRasym (3.5 ppm) for distinguishing metastatic from non-metastatic lymph nodes (z = 2.407, p = 0.0161). No significant differences were found in MR parameters for distinguishing other histopathological features (p > 0.05). IDEAL-IQ and APTWI were associated with histopathological factors of rectal cancer, and might serve as non-invasive biomarkers for characterizing rectal cancer.


Introduction
Colorectal cancer, a common alimentary tumor, accounts for 30-35% of cancer-related death [1]. The surgical operation of total mesorectal excision is applicable for patients at the early stage, while the neoadjuvant chemo-radiotherapy is usually considered for patients with locally advanced rectal cancer [2]. The pathological features of rectal cancer could help decide the selection of surgical operation or neoadjuvant chemo-radiotherapy for its treatment. Previous studies [3,4] reported that patients with rectal cancer of higher T stage, lymph node involvement, poor differentiation and extramural venous invasion (EMVI) have found to be related to local recurrence and distant metastasis. Hence, it is valuable to make an accurate preoperative assessment of histopathological features of rectal cancer, which might facilitate treatment strategies and improve long-term survival rate of patients.
Previous studies [5,6] demonstrated that the adipose tissue, adipocytes and preadipocytes in the culture medium could promote the proliferation of colon cancer cells. The adipose tissue owns the characteristics similar to an endocrine organ [7] and influences

MRscanning Protocols
A 3.0T MR scanner (Discovery MR750, GE Healthcare, Waukesha, WI, USA) was applied for rectal cancer patients. A 32-channel phased-array coil was utilized in the scanning process. Each studied subject was intramuscularly injected with 5 mg racanisodamine hydrochloride 20 min before MR examination to prevent artifacts caused by intestinal peristalsis. Our MR imaging protocol was included as follows: T1-weighted imaging (WI), T2WI, diffusion-weighted imaging (DWI), IDEAL-IQ and APTWI ( Table 1). The DWI, IDEAL-IQ and APTWI sequences were scanned vertically to the long axis of rectal lesions by utility of sagittal T2WI as reference.

MRscanning Protocols
A 3.0T MR scanner (Discovery MR750, GE Healthcare, Waukesha, WI, USA) was applied for rectal cancer patients. A 32-channel phased-array coil was utilized in the scanning process. Each studied subject was intramuscularly injected with 5 mg racanisodamine hydrochloride 20 min before MR examination to prevent artifacts caused by intestinal peristalsis. Our MR imaging protocol was included as follows: T1-weighted imaging (WI), T2WI, diffusion-weighted imaging (DWI), IDEAL-IQ and APTWI ( Table 1). The DWI, IDEAL-IQ and APTWI sequences were scanned vertically to the long axis of rectal lesions by utility of sagittal T2WI as reference.

Image Processing and Analysis
All the APTWI images were transferred to a GE workstation (AW 4.6, GE healthcare, USA), and the corresponding images were processed and analyzed using the APT software in the functional kit. The region of interest (ROI) was contoured by delineating the margin of tumor tissue on the axial APT image using axial high-resolution T2WI and DWI images as reference. Every ROI was placed on the corresponding APT image (single slice) at the largest section of each tumor to cover the solid region of tumor as much as possible. The necrotic, cystic and bleeding areas should be avoided for ROI delineation.
The APTWI-derived parameter (MTRasym) was calculated by utility of the following equation: (MTRasym (3.5 ppm) refers to the magnetization transfer ratio at 3.5 ppm, S0 indicates the signal intensity of APTw without applying the saturation pulse, and Ssat is the signal intensity of APTw after applying the saturation pulse).
As for the IDEAL-IQ imaging technique, the proton density fat fraction (PDFF) maps and R2* maps were automatically produced by the corresponding post-processing software installed in our MR scanner. Then, all the functional MR images were sent to the PACS (Picture Archiving and Communication System) system in our working computers. The freehand ROIs were drawn along the margin of the rectal tumor on each slice consecutively by covering the whole rectal lesion. The delineation of ROIs on the PDFF and R2* maps was based on T2WI and DWI as references, and the regions with intra-luminal gas, necrosis, cystic change and bleeding should be eschewed.
The above MR data were measured by two radiologists specializing in abdominal radiology with five and eleven years' experience. They independently assessed all the MR images and delineated ROIs without knowing the clinicopathological information of patients. The final values of IDEAL-IQ and APTWI-derived parameters were the averaged values of all the measurements on all slices.

Histopathological Analysis
Two gastrointestinal pathologists with six and thirteen years of experience, independently evaluated all the histological examinations. All the specimens, removed at surgical resection, were fixed with 10% formalin, cut into 4-µm sections, and stained with hematoxylin and eosin (H.E.). The pathological characteristics of rectal cancer were as follows: tumor invasion of rectal wall (T stage), lymph node involvement (N stage), histological grade, status of mesorectal fascia (MRF) invasion and extramural venous invasion (EMVI).
With regard to the gland/tubule formation and tissue structure in rectal tumors, the histological tumor grade was categorized into G1 (well-differentiated tumor with over 95% gland formation), G2 (moderately differentiated tumor with 50-95% gland formation) and G3 (poorly differentiated tumor with 0-49% gland formation). G1 and G2 tumors belonged to low-grade tumors and G3 tumors were classified as high-grade tumors, based on the WHO criteria for tumor grading system. A positive MRF status indicates that tumors/malignant lymph nodes are very near to the mesorectal fascia-within 1 mm of distance. Positive EMVI status refers to the finding of tumor components within an extramural endothelium-lined space, which is surrounded by a rim of muscle structure or contains red blood cells [27].

Statistical Analysis
The Kolmogorov-Smirnov test was utilized to test the quantitative parameters of mean PDFF, R2* and MTRasym (3.5 ppm) for normal distribution. The independent sample t test or Mann-Whitney U test was performed to assess the differences of each parameter in relation to different status of histopathological features of rectal cancer. The different status of histopathological features refers to: pT1-2 versus pT3-4, pN0 versus pN 1-2, low grade (G1-2) versus high grade (G3), negative versus positive status of MRF and EMVI. The association of MRI parameters with histopathological features was analyzed by the Spearman correlation coefficient, which was also utilized for evaluating the relationship between PDFF/R2* and MTRasym (3.5 ppm).
The receiver operating characteristic (ROC) curve was performed to assess the diagnostic performances of these MR parameters for prediction of different status of histopathological features of rectal cancer. The differences in the area under the ROC curve (AUC) of each parameter for assessment of histopathological features were compared using the Delong test. The parameters including the AUC, sensitivity, specificity, and Youden index were derived from the ROC curve analysis.
The intra-class correlation coefficient (ICC) was also performed to analyze the interobserver repeatability between two radiologists on measurements of IDEAL-IQ and APTWI derived parameters. The statistical data analyses were performed using SPSS 19.0 (IBM, Armonk, NY, USA) and MedCalc 11.4.2 softwares (Mariakerke, Belgium). p values of <0.05 were considered statistically significant.

Clinic and Pathologic Characteristics of Studied Population
The study cohort comprised 67 patients (42 males, 25 females, age range: 30-82 years, median: 58 years). There were 15, 22 and 30 patients with rectal cancer located at the upper, middle and lower segments of the rectum according to the location of rectal tumors, respectively. The IDEAL-IQ and APTWI parameters of different histopathological factors of rectal cancer were averaged in Table 2. The characteristics of all patients were displayed in Table 2.

Discussion
In our study, we performed IDEAL-IQ and APTWI on patients with rectal cancer, and obtained the intra-tumoral fat fraction (PDFF), R2* and MTRasym (3.5 ppm) from tumor regions. Regarding the available MR data, we found that these MR parameters were associated with histopathological features of rectal cancer, such as T/N stage, histological differentiation, MRF and EMVI. The AUC of PDFF was significantly greater than that of MTRasym (3.5 ppm) in distinguishing metastatic from non-metastatic lymph nodes, but no other statistical differences were noted between IDEAL-IQ and APTWI parameters in prediction of the other histopathological features of rectal cancer. The IDEAL-IQ and APTWI are advantageous in imaging adipose tissue and tumor protein, and both functional techniques could be applied to clinical workflow by combining with immunohistochemical analysis in the future radiology diagnosis, although IDEAL-IQ and APTWI techniques need to be optimized in both image quality and scanning parameters.
We demonstrated that higher PDFF and R2* values were found in rectal tumors with higher T stage, lymph node metastasis, higher tumor grade, and positive status of MRF and EMVI. Adipose tissue is an endocrine organ, which could secrete cytokines and adipokines by adipocytes [28,29]. These cytokines and adipokines might finally promote inflammation and tumorigenesis of rectal cancer [30]. The adipose tissue microenvironment is closely associated with tumor formation and progression because of adipokines and fatty acids provided to tumor cells [31]. Moreover, accompanied angiogenesis also plays a key role in aggravation of histopathological features of rectal cancer with poor prognosis. Therefore, based on the findings from our investigation, it was assumed that intra-tumoral fat composition might potentially be related to the evolution process of tumor growth and prognosis of rectal cancer.
As for R2* parameter derived from the IDEAL-IQ sequence, there was a positive correlation between R2* and histopathological features of rectal cancer with advanced T/N stage, higher tumor grade, and positive status of MRF and EMVI. The R2* parameter was previously reported to quantify tumor hypoxia in animal and human studies [13,32,33]. Previous investigation indicated the increased blood deoxyhemoglobin content could result in elevated R2* [16]. The advanced stage tumors are manifested by gland formation, conspicuous cell atypia in tumor cells, nuclear polymorphism and increased tumor cell density of rectal cancer, which is depleted of oxygen in the tumor microenvironment by tumor consumption [34,35], accompanied by an increased level of deoxyhemoglobin concentration and R2* values. Moreover, tumor hypoxia is often accompanied by angiogenesis, which further aggravates the hypoxic status of tumor microenvironment and facilitates tumor progression and metastasis [36]. So the R2* parameter could be a potential biomarker for discrimination of the different status of T/N stage, tumor grade, MRF and EMVI of rectal cancer.
With regards to APTWI, significantly higher values of MTRasym (3.5 ppm) were found in rectal cancer with higher T stage, nodal involvement, higher tumor grade and positive status of MRF/EMVI. APTWI is a relatively novel molecular MR sequence, which could detect tumor proteins and peptides. APTWI is founded on the chemical exchange saturation transfer process and the signal intensity of APTWI is produced from the exchange of protons between endogenous protein/peptides and bulky water. Rectal cancer with the above histopathological features shows biological behaviors with aggressive manners, and is often featured by large tumor size, increased tumor cells/cellular density and excessive cell proliferation rate [37]. High-level tumor cell proliferation necessitates growing protein synthesis, leading to abnormal accumulation of peptides and proteins in the local tumor microenvironment. Hence, an augmented tumor proliferation pattern and tumor-related protein synthesis are the primary reasons for the enhanced APT signal (high MTRasym (3.5 ppm) value) by APTWI. Besides, tumor angiogenesis was reported to be another important factor, which could influence the APT signal and increase the MTRasym (3.5 ppm) value in rectal cancer with higher T/N stage, poor histological differentiation and positive status of MRF/EMVI status [19]. Abnormal vessel hyperplasia can often be present in rectal cancer with EMVI positive status, and increased vasculature permeability/blood perfusion are noted, which could be the potential reason for high APTWI signal. Based on the above findings, we hypothesize that APTWI could image protein synthesis by tumor cell proliferation, and that protein-associated angiogenesis in the tumor microenvironment is closely intertwined with a high APT signal.
More interestingly, we found that PDFF and R2* were moderately and mildly correlated to MTRasym (3.5 ppm), respectively. Previous studies [38,39] indicated that adipose tissue and hypoxia could help facilitate tumor growth. This result of association of PDFF and R2* with MTRasym (3.5 ppm) was consistent with previous findings. This finding indirectly implied that intra-tumoral fat tissue (PDFF) and hypoxia (R2*) might promote tumor cell proliferation, indicated by MTRasym (3.5 ppm) from APTWI. However, the above results were found based on the MR data from the rectal cancer patients collected from our hospital with a limited study number, so more patients should be enrolled to test the repeatability and validity of our findings.
We also evaluated the diagnostic performance of IDEAL-IQ and APTWI in analysis of histological features of rectal cancer. PDFF was significantly superior over MTRasym (3.5 ppm) for distinguishing metastatic from non-metastatic lymph nodes from our ROC analysis. No statistical differences between the diagnostic abilities of PDFF, R2* and MTRasym (3.5 ppm) were found in discriminating histological features of rectal cancer. The above results might be due to the following reasons: (1) it was under exploration for finding factors for causing variations of peptides and proteins in the tumor microenvironment; (2) the APTWI still needs improvement for scanning parameters, slices and scanning time; (3) APTWI could be influenced by the T1 relaxation time, local environment temperature and tissue pH, and the measurement of the parameter from APTWI might potentially be affected by the above factors.
Several limitations are present in our study. First, our sample size of studied patients was limited, and more patients should be enrolled in future investigations; second, the scanning time for APTWI was too long, and the long scanning time might cause potential occurrence of artifacts; third, the homogeneity of regional MR signal intensity is essential for APT imaging of the rectum, and intra-luminal gas existence and bowel movement would lead to uneven MR signal intensity and influence the APT signal, even though racanisodamine hydrochloride was utilized to reduce bowel movement. Fourth, only patients with rectal cancer were chosen for IDEAL-IQ and APT-weighted imaging, and other histological types of rectal tumors should be taken into consideration in future studies. Finally, the publicly available databases of IDEAL-IQ/APTWI-derived parameters and associated histopathological factors of rectal cancer did not exist, thus limiting the generalizability of our findings.
In conclusion, IDEAL-IQ and APTWI are proved to be applicable to imaging rectal cancer based on our investigation. Both sequences could be used as noninvasive modalities for discrimination of histopathological features of rectal cancer.

Informed Consent Statement:
The written informed consent was waived for all patients prior to their enrollment in this investigation because of the retrospective nature of our study.

Data Availability Statement:
The data associated with the findings of our investigation are available from the corresponding author (Z.L.) upon reasonable request.

Conflicts of Interest:
The authors declared that they had no conflicts of interest.