Diffusion Kurtosis Imaging Versus DWI in the Clinical Assessment of Rectal Carcinoma: A Retrospective Study

Background: To perform an analysis of mean diffusivity(MD) (cid:0) mean kurtosis(MK) demonstrating the diagnostic value of diffusion kurtosis imaging (DKI) and diffusion weighted imaging (DWI) with respect to rectal carcinoma. Methods: A total of thirty-nine rectal carcinoma cases and thirty-nine healthy subjects (Normal control group) were enrolled in our study. All the subjects underwent multi-parameter (DWI, DKI )magnetic resonance examination. The acquired images were individually analysed by two readers. The obtained images were input into the corresponding software, then an analysis of the subjects’ apparent diffusion coecient(ADC), MD and MK values was performed. A receiver-operating charasterictic (ROC) analysis was used to assess the diagnostic eciency of the MK, MD and ADC parameters. The Mann-Whitney U test was used to contrast the parameters in both groups. Spearman correlation analysis was used to analyse the correlation between ADC and MD, MK. The Kappa consistency test was used to evaluate the consistency between each reader’s evaluation. Results: Reecting their diagnostic values with respect to rectal carcinoma, the AUC for MK, MD, and ADC were 0.911, 0.888, and 0.827 (all P <0.05), respectively. Using 0.59, 2.15×10 (cid:0) 3 mm 2 /s, 1.35×10 (cid:0) 3 mm 2 /s as thresholds, the sensitivities of MK, MD, ADC were 89.50%, 78.90%, and 76.30%, respectively; meanwhile their respective specicities were 84.20%, 73.70%, and 73.70%.The ADC was directly proportional to MD ( r =0.994, P <0.05) and inversely proportional to MK ( r = (cid:0) 0.460, P <0.05). Analysis of the imaging data revealed consistent results from both readers,Kappa=0.737. Conclusion: The ADC (cid:0) MK and MD parameters were effective in diagnosing rectal carcinoma. Moreover, the MK and MD parameters were found to provide even more valid information regarding the microenvironment with a higher diagnostic performance.


Introduction
Rectal carcinoma is a common malignant tumor of the digestive tract. In the United States, colorectal cancer is the third most common type of cancer and the third leading cause of cancer mortality [1] .Recent, evidence has revealed that the popularity of screening and changes in treatment strategies have delayed the progression of disease and decreased the overall mortality rates of patients with rectal carcinoma [2].Correct diagnosis and assessment of rectal cancer are crucial factors when determining the choice of treatment.
According to the diffusion-weighted imaging(DWI) single index and diffusion-tensor imaging(DTI) models, the diffusion of water molecules follows a Gaussian distribution and water molecules are found to move freely and unrestricted in biological tissues [3]. DWI has been shown to be effective for the qualitative and quantitative analysis of lesions, and can be used for the diagnosis and staging of rectal cancer at the molecular level [4].Diffusion kurtosis imaging(DKI) is based on the non-gaussian diffusion theory [5]; it is primarily applied in the neurological eld[6],with its prominent role being tumor identi cation and grading. However, the technique is gradually being applied in the diagnosis of prostate lesions and disease affecting other organs [7]. Up to now no comparison had been made between DKI and apparent diffusion coe cient(ADC) in terms of colorectal cancer diagnosis, thus we aimed to investigate the ADC values of DWI and DKI parameters in the diagnosis of rectal cancer, and the relationship between ADC values and DKI parameters.

Patients
The pathological results were considered as the gold standard for diagnosis, a retrospective analysis was conducted on the data of patients (carcinoma of the rectum group)admitted to our hospital from January to June 2017 with a con rmed diagnosis of rectal cancer by surgical pathology, including twenty-nine males and ten females (Mean age = 54.6 ± 12.9 years; age range = 25 to 80 years).Case inclusion criteria: Patients who had not received surgery or chemoradiotherapy for at least a week before the examination; The image had no motion artifacts Postoperative pathological biopsy con rmed adenocarcinoma of the rectum. Case exclusion criteria The image's artifacts were serious and did not meet the diagnosis requirements No surgical treatment was performed within 10 days of the examination Chemoradiotherapy was performed before surgery The postoperative pathological results reported mucinous adenocarcinoma adenoma and in ammation The lesion was too small to conduct an ROI measurement Thirty-nine healthy volunteers matched with the rectal carcinoma group in terms of age and sex were enrolled as the normal control group, including twenty-nine males and ten females(Mean age = 53.1 ± 11.2 years; age range = 25 to 80 years old).Inclusion criteria: Tolerance test; No intestinal disease;No chemoradiotherapy; MRI showed no obvious abnormalities. This study was approved by the medical ethics committee of our hospital, and all the subjects signed an informed consent.

Data Extraction
The MRI images were saved in a DICOM format. ADC values were automatically generated after DWI sequence scanning, the DKI raw date was post-processed through a prototyped Diffusion Toolbox software (Siemens Healthineers), then the mean diffusivity(MD) mean kurtosis(MK) parameters were generated. Two radiologists with more than 5 years of working experience in the diagnosis of pelvic diseases conducted an ROI analysis (Area range 7.12 to 9.68 mm 2 )on the largest surface area covered by the tumor to measure the MK and MD values, avoiding necrotic and cystic areas as much as possible. Whenever a great difference between the two measurements made by the two radiologists was identi ed, a corresponding discussion was initiated and a unanimous decision made. Additionally, two radiologists with more than 5 years of working experience in the diagnosis of pelvic diseases also performed an ROI analysis(Area range 6.58 to 9.47 mm 2 )on the largest surface area covered by the tumor to measure the ADC values, avoiding necrotic and cystic areas as much as possible. Whenever a great difference between the two measurements made by the two radiologists was identi ed, a corresponding discussion was initiated and a unanimous decision made. All the parameters were measured for 3 times and the mean value was taken.

Statistical Analysis
The SPSS 21.0 software was used for statistical analysis,and measurement data were expressed as ±s The collected data did not follow a normal distribution. The Mann-Whitney U test was used to compare the differences between the two groups. Using the pathological results as the gold standard, ROC curves were drawn, and the respective values for the AUC were calculated. Threshold values were determined according to the most probable index, and the diagnostic e cacy of each parameter for carcinoma of the rectum was evaluated. The Spearman rank correlation was used to analyse the relationship between ADC values and MD MK values. The kappa test was used to evaluate the consistency of the ROI results measured by the two physicians.A value of kappa lower than 0.20 was interpreted as poor agreement,0.41-0.60 as moderate,0.61-0.80 as substantial,and 0.81-1 as almost perfect agreement according to Cohen's kappa coe cient[8] The difference was statistically signi cant if P < 0.05

Results
The consistency of the ROI measured by the two diagnostic radiologists was good, Kappa=0.737.
Postoperative pathology con rmed that among the thirty-nine cases of rectal carcinoma,11 cases were highly differentiated adenocarcinoma,13 cases were moderately differentiated adenocarcinoma, and 15 cases were poorly differentiated adenocarcinoma. Pathological T stage was ≤T2 in 8 cases T3 in 18 cases and T4 in 13 cases.
Compared with the control group, an increment in the MK values and a decrement in the MD ADC values were observed in the rectal carcinoma group, with statistically signi cant differences, as displayed in table 1 and gure 1 and 2.  ADC values re ect the state of limited diffusion of water molecules. The higher the density of tissue cells, the more limited the diffusion of molecules, the lower the ADC value and vice versa. Malignant cells elicit a fast differentiation rate, thus the diffusion of water molecules in such lesions are limited by the restriction of macromolecular substances and cell membranes [10],resulting in a decrease in the ADC value. This study's results reveal that ADC values in the rectal carcinoma group were lower than those depicted in the control group(P < 0.001) which was consistent with the previous theoretical hypotheses [11].Sun YS et al [12] con rmed that the optimal b value of DWI in rectal cancer was 1000 s/mm 2 which could better overcome the in uence of perfusion and T2 penetration effects. However, measurement of the ADC value is based on the fact that the diffusion of water molecules conforms to the gaussian distribution. Withal, the human body is composed of several tissues, hence the cell types, cell density and blood supply of each tissue are different. The diffusion of water molecules follows a nongaussian distribution [12] DKI represents the plane echo sequence of a single excitation, with 7 different b values and 3 scanning directions perpendicular to each other;The MK value is calculated by the apparent kurtosis coe cient (AKC) in each direction. The larger the MK value is, the more complex the structure is in the target organization, so it is used to measure the complexity of the organizational structure the larger the MK value is, the more complex the structure is in the target organization, so it is used to measure the complexity of the organizational structure [13]. Moreover, malignant lesions have an abundant supply of interstitial blood vessels, so it was speculated that the MK values of malignant lesions should be higher than those of normal tissues, which was con rmed by the results of this study. On the other hand, MD values were similar to the ADC values of the DWI single index model, which re ects the diffusion state of water molecules in the human body. Malignant cells are compactly arranged, which limits the diffusion of water molecules, leading to a subsequent decrease in their MD values. Based on the aforementioned observations, we speculate that the MD values of malignant lesions should be lower than those of normal tissues. In this study, differences in ADC, MD and MK values between the two groups were statistically signi cant, indicating that the three parameters were all effective parameters for the diagnosis of rectal carcinoma and could re ect the changes of lesion tissue structures at the microstructure level.
The ROC analysis conducted in this study suggested that the AUC for the MK value was the largest and that of the ADC value was the smallest. The sensitivity and speci city values of MK value were higher than those of the MD and ADC values, so DKI was more effective in diagnosing rectal carcinoma. At present, the preoperative diagnosis of rectal cancer is mainly based on the combination of DWI and T2WI [14], and DKI can also provide similar diffusion images.  50 year-old man, normal volunteer. a. Axial T2WI shows normal rectal wall divided into three layer. bd.MK, MD, and ADC maps, respectively; MK, MD, ADC values were 0.485, 2.37×10 3 mm/s2, and 1.61×10 3 mm/s2 respectively.