Diffusion kurtosis imaging assessment of the response to radiotherapy in a VX2 bone tumor model: an animal study

Abstract

Background

Neoadjuvant radiotherapy plays a vital role in the treatment of malignant bone tumors, and non-invasive imaging methods are needed to evaluate the response to treatment.

Purpose

To assess the value of diffusion kurtosis imaging (DKI) for monitoring early response to radiotherapy in malignant bone tumors.

Material and Methods

Treatment response was evaluated in a rabbit VX2 bone tumor model (n = 35) using magnetic resonance imaging (MRI), DKI, and histopathologic examinations. Subjects were divided into three groups: pre-treatment, post-treatment, and control groups. The post-treatment group was subclassified into good response and poor response groups according to the results of histopathologic examination. Apparent diffusion coefficient (ADC) and DKI parameters (mean diffusion coefficient [MD] and mean kurtosis [MK]) were recorded. The relationship between ADC, DKI parameters, and histopathologic changes after radiotherapy was determined using Pearson’s correlation coefficient. The diagnostic performance of these parameters was assessed using receiver operating characteristic analysis.

Results

MD in the good response group was higher after treatment than before treatment (P < 0.001) and higher than that in the poor response group (P = 0.009). MD was highly correlated with tumor cell density and apoptosis rate (r = −0.771, P < 0.001 and r = 0.625, P < 0.001, respectively). MD was superior to other parameters for determining the curative effect of radiotherapy, with a sensitivity of 75.0%, specificity of 100.0%, and area under the curve of 0.917 (P < 0.001).

Conclusion

The correlations between MD, tumor cell density, and apoptosis suggest that MD could be useful for assessing the early response to radiotherapy in rabbit VX2 malignant bone tumors.

Keywords

Magnetic resonance imaging neoplasms bone

Introduction

Malignant bone tumors represent a small proportion of cancers, although these tumors are associated with high disability, mortality, and metastasis rates, which seriously endangers human health (1). Neoadjuvant radiotherapy and chemotherapy play vital roles in the comprehensive treatment of malignant bone tumors (2,3). However, ineffective treatment may increase the risk of radiation injury or promote the formation of resistant clones. Histopathological specimens are commonly used to monitor the clinical efficacy of treatment. However, histological changes in the tumor can only be assessed by puncture biopsy or postoperatively, and biopsy can only assess the condition of the tissue sample and is an invasive procedure. Therefore, the development of a noninvasive imaging alternative is desirable to monitor the early response to neoadjuvant chemoradiotherapy.

Diffusion-weighted imaging (DWI) is a non-invasive magnetic resonance imaging (MRI) modality that is useful for distinguishing benign and malignant musculoskeletal lesions and for monitoring the response of bone tumors to treatment (3 –6). In 2005, Jensen et al. (7) proposed a diffusion kurtosis imaging (DKI) model as an expansion of DWI, which can measure the diffusion characteristics of water molecules under a non-Gaussian displacement distribution and more sensitively reflect changes in tissue microstructure. Recently, the DKI model has been applied in research in nasopharyngeal carcinoma (8), glioma (9), rectal cancer (10), and other neoplastic lesions. The aim of the present study was to investigate whether DKI parameters could predict the treatment response to radiotherapy of VX2 malignant bone tumors in rabbits, and to investigate the relationship between DKI parameters and histopathological features.

Material and Methods

Animal model

The experiment was conducted according to the guidelines for studies using laboratory animals and approved by our hospital ethics committee. The 41 purebred male New Zealand white rabbits (weighing 2.0–3.0 kg) that were used in this study were provided by the Qingdao Food and Drug Administration. The rabbits were housed with a standard diet, and their circadian rhythms were synchronized with 12-h light/dark cycles. The malignant bone tumor model was established as described in previous studies (11,12). Exclusion criteria were: tumor diameter <10 mm at three weeks; death from natural or iatrogenic causes; absence of complete MR or pathology data; and poor MR image quality due to artifacts caused by breathing and movement.

Study design

The primary observational outcome index was the therapeutic effect. Chun et al. (13) suggested that the reduction in tumor cell density (TCD) can be used to assess the treatment response. A treatment-induced decrease in TCD of > 40% was considered a good response, whereas a decrease of < 40% was considered a poor response. According to preliminary experiments, the percentage TCD decreased by > 40%, with the treatment group having a density of 60%, whereas the control group had a density of 10% of the proportion. Assuming power = 0.8 and alpha = 0.05, and a treatment-to-control ratio of 1:1, we assigned no less than 13 animals per group. Accordingly, the malignant bone tumor models were randomly divided into pre-treatment, post-treatment, and control groups (n = 13, 15, and 13, respectively). MRI was performed three weeks after tumor implantation. In the pre-treatment group, MRI scans were performed on the first day after three weeks of tumor implantation (day 1). The post-treatment group received a single dose of 10 Gy radiotherapy, and rabbits were evaluated by longitudinal MRI scan after three days. Subjects were divided into good response and poor response groups according to the treatment response. The control group underwent MRI on day 3 without treatment to assess the natural course of the tumor. After the MR examination, the subjects were immediately euthanized for pathological examination. The flow chart is shown in Fig. 1.

Fig. 1.

Study workflow. MRI was performed three weeks after tumor implantation. Day 1 = the first day after three weeks of tumor implantation, Day 3 = the third day after three weeks of tumor implantation.

MRI protocol

MRI scans were processed on a Magnetom Prisma 3.0 MR system (Siemens Healthcare GmbH, Erlangen, Germany) using an eight-channel rabbit coil (Chenguang Healthcare, Shanghai, PR China). Ketamine (1.5 mg/kg) was injected into the thigh muscles of the rabbits for sedation, and rabbits were then placed in the supine position and allowed to breathe freely. Plain MR scans were performed after sedation (axial, sagittal, coronal T1-weighted [T1W] image sequence, and sagittal T2-weighted [T2W] image fat suppression sequence) with the following scanning parameters: field of view (FOV) = 160 × 160 mm; slice thickness = 2 mm; and intersection gap = 1 mm. The readout segmentation of long variable echo-trains (RESOLVE) sequence was used for DKI scanning. Imaging was performed in the sagittal plane. The DKI parameters were as follows: repetition time/echo time (TR/TE) = 3000/55 ms; slice thickness = 3 mm; matrix = 128 × 100; number of excitations (NEX) = 2; diffusion-gradient direction = 3; b values were 0, 1000, 1500, and 2000 s/mm². The scanning time was 7 min 53 s.

Histopathologic evaluation

The animals were sacrificed immediately after the MR examination under deep anesthesia. The right tibia containing the malignant bone tumor was immersed in formalin for at least 48 h. The specimens were then cut to a thickness of 3 mm along the sagittal plane, and the maximum sagittal plane of the tumor corresponding to the MR image was selected. The tumor position on the specimen was delineated by the grid coordinates as shown in Fig. 2 to perform a point-by-point comparison with the tumor orientation in the corresponding MR image (12). Tissue specimens were obtained and fixed in formalin, decalcified with ethylene diamine triacetic acid (EDTA), and embedded in paraffin for hematoxylin and eosin (H&E) staining on large microscope slides. The terminal deoxynucleotidyl transferase (TdT)-mediated deoxyuracil nucleoside triphosphate (dUTP) nick-end labeling (TUNEL) method was used following established protocols to detect apoptotic cells in each specimen. TUNEL kits were obtained from Roche Applied Science in Penzberg, Germany.

Fig. 2.

Maximal sagittal plane of the tumor and the corresponding DKI-sequence imaging. The square grids in the two figures form a coordinate system, which allowed point-to-point comparisons with the same coordinate values. DKI, diffusion kurtosis imaging.

CaseViewer and Image J software were used to process the histopathological data. Histopathologic sections were evaluated, and a consensus was obtained by two blinded pathologists (HZ, LLS). The tumor nuclei were evaluated by point-to-point comparison with pathological sections at high magnification (200×). TCD was calculated using the following formula: TCD = Area_{tumor cell nucleus}/Area_{statistical field}. Cell apoptosis was assessed as described above and expressed as the percentage of positive TUNEL-stained cells among 200 tumor nuclei from selected fields at high magnification (400×). The results were calculated by randomly selecting five visual fields and calculating average values, which were expressed as percentages.

MRI analysis

MRI data were imported into a Siemens workstation (Syngo.via) and analyzed using the Body Diffusion Toolbox (Siemens Healthcare GmbH, Erlangen, Germany). The diffusion kurtosis images were imported into the software to obtain the final image calculated and fitted by the kurtosis model (MD map and MK map). The ADC map used for analysis was calculated from two diffusion coefficient values (b = 0, 1000 s/mm²). With reference to the gross specimen sections, two radiologists (WKS, CD) delineated the regions of interest (ROIs) by consensus with respect to the sagittal ADC map representing the matching area of the tumor, excluding the macroscopically visible necrotic or cystic areas. The ROIs were then automatically copied onto the corresponding MD and MK map. The standard ROI was 5 mm in diameter and 19 mm² in area and was drawn three times to obtain the final average value (Fig. 3).

Fig. 5.

Representative scatterplot showing the relationships between ADC and DKI parameters and the histological features. ADC, apparent diffusion coefficient; DKI, diffusion-kurtosis imaging.

Statistical analysis

Statistical analysis was performed using SPSS v22.0 (IBM Corporation, Armonk, NY, USA) and MedCalc v19.0.7 (MedCalc Inc, Mariakerke, Belgium). A P value <0.05 was considered statistically significant. The Shapiro–Wilk test was used to verify whether the data conformed to a normal distribution. The independent samples t-test with Bonferroni correction was used to compare the differences between the ADC and DKI parameters, TCD, and cellular apoptosis in the tumors in each group. Pearson’s correlation coefficient was used to analyze the correlations between pathological indices and the corresponding ADC and DKI parameters. In the absence of a linear relationship, the significance between groups was assessed by Spearman’s correlation analysis. The area under the curve (AUC) of the receiver operator characteristic (ROC) curve was used to evaluate the efficacy of each MR parameter for predicting the response to treatment between the good response and poor response groups. A greater AUC value indicated a higher predictive efficacy. The corresponding AUC values, sensitivity, and specificity were calculated, including 95% confidence intervals (CI). The optimal cut-off values were determined according to the most approximate Youden’s index (sensitivity + specificity − 1). The results are expressed as the mean ± SD.

Results

Successful models were generated in 41 rabbits, of which two died naturally, three died during anesthesia, and one died after receiving radiotherapy. Finally, there were 11 subjects in the pre-treatment group, 14 in the post-treatment group, and 10 in the control group. The subgroups according to the treatment response constituted eight rabbits in the good response group and six in the poor response group.

Relevant MRI and histological parameters in the pre-treatment, post-treatment, and control groups are listed in Table 1, and the results of the statistical analysis of the quantitative parameters are shown in Fig. 4. Table 2 summarizes the relationships between ADC, DKI parameters, TCD, and cellular apoptosis determined by Pearson’s correlation coefficient analysis (Fig. 5). MD and ADC were negatively correlated with TCD (r = −0.771, P < 0.001 and −0.760, P < 0.001, respectively) and positively correlated with apoptosis (r = 0.625, P < 0.001 and r = 0.543, P = 0.001, respectively). MK was not significantly correlated with cell density and apoptosis. Figs. 3 and 6 show the calculated ADC, MD, and MK maps and the corresponding histopathologic images.

Table 1.

Summary of various MRI and histological parameters in each group.*

Variable	Pre-treatment	Post-treatment		Control	P1	P2	P3	P4
Variable	Pre-treatment	Good response	Poor response	Control	P1	P2	P3	P4
MD (×10^–3mm²/s)	0.74 ± 0.09	0.96 ± 0.09	0.82 ± 0.06	0.72 ± 0.06	<0.001	0.242	0.009	0.713
MK	0.86 ± 0.06	0.78 ± 0.05	0.82 ± 0.03	0.84 ± 0.11	0.010	0.148	0.163	0.510
ADC (×10^–3mm²/s)	0.63 ± 0.06	0.75 ± 0.03	0.72 ± 0.02	0.65 ± 0.06	<0.001	0.006	0.053	0.376
TCD	44.32 ± 5.50	21.46 ± 1.69	28.83 ± 2.52	41.20 ± 1.78	<0.001	<0.001	0.003	0.304
Apoptosis	8.04 ± 1.33	12.00 ± 2.39	9.02 ± 1.12	7.09 ± 0.77	<0.001	0.148	0.005	0.922

Values are given as mean ± SD.

*P < 0.05 is considered to be statistically significant.

ADC, apparent diffusion coefficient; MD, mean diffusion coefficient; MK, mean kurtosis; P1, pre-treatment vs. good response; P2, pre-treatment vs. poor response; P3, good response vs. poor response; P4, pre-treatment vs. control; TCD, tumor cell density.

Fig. 3.

DKI parameters and histopathological images of rabbit VX2 malignant bone tumor models. The images represent before (a–c) and after (d–f) radiotherapy. (a, d) ADC map; (b, e) MD map; (c, f) MK map. Both MD and ADC in the tumor increased, whereas MK decreased after radiotherapy. ADC, apparent diffusion coefficient; DKI, diffusion kurtosis imaging; MD, mean diffusion; MK, mean kurtosis.

Table 2.

Correlation between histopathology of tumor and imaging indexes.

Variable	TCD (%)		Apoptosis (%)
Variable	r	P	r	P
MD (×10^–3mm²/s)	–0.771	<0.001	0.625	<0.001
MK	0.284	0.098	–0.226	0.191
ADC (×10^–3mm²/s)	–0.760	<0.001	0.543	0.001

P < 0.05 is considered to be statistically significant.

ADC, apparent diffusion coefficient; MD, mean diffusion coefficient; MK, mean kurtosis; TCD, tumor cell density.

Fig. 4.

Box plots show the results of ADC and DKI parameter analysis in the pre-treatment, good response, poor response, and control groups. (a) Compared with pre-treatment and control groups, the MD of the post-treatment group increased, and the difference between the good response and the poor response groups was statistically significant. (b) The MK of the good response group was significantly lower than the pre-treatment value, and there was no significant difference in MK between good response and poor response groups. (c) The ADC of the post-treatment group was significantly higher than that of the pre-treatment and control groups, and there was no significant difference between the good response and poor response groups. *Statistically significant. ADC, apparent diffusion coefficient; DKI, diffusion kurtosis imaging; MD, mean diffusion; MK, mean kurtosis.

Fig. 6.

The histopathological images of rabbit VX2 malignant bone tumor models. The images represent before (a, b) and after (c, d) radiotherapy. (a, c) Corresponding H&E images (magnified 200×). Before radiotherapy, the tumor cell density was high and cells were distributed in clusters. After treatment, cytoplasmic necrosis increased and nuclear pyknosis or disappearance of tumor was observed, and the density of tumor cells decreased significantly. (b, d) Corresponding TUNEL staining images (magnified 400×). The number of apoptotic cells increased significantly after radiotherapy, and apoptotic bodies were detected. H&E, hematoxylin and eosin.

ROC analysis was performed to determine the accuracy of DKI (Table 3). MD showed the highest diagnostic performance for assessing radiotherapeutic response (AUC = 0.917, P < 0.001, sensitivity = 75.0%, specificity = 100.0%). MK and ADC were 0.708 and 0.813, respectively (Fig. 7).

Table 3.

Diagnostic performance for DKI-MRI indexes in assessing radiotherapeutic response.

Variable	Cut-off	AUC	95% CI	Sensitivity (%)	Specificity (%)	P
MD (×10^–3 mm²/s)	0.91	0.917	0.646–0.997	75.00	100.00	<0.001
MK	0.79	0.708	0.413–0.913	75.00	83.33	0.185
ADC (×10^–3 mm²/s)	0.74	0.813	0.521–0.965	66.67	90.48	0.017

P < 0.05 is considered to be statistically significant.

ADC, apparent diffusion coefficient; AUC, area under the receiver operating characteristic curve; CI, confidence interval; MD, mean diffusion coefficient; MK, mean kurtosis.

Fig. 7.

The result of ROC curve analysis to show the efficacy of various parameters to predict the outcome of VX2 malignant bone tumor treatment response. ROC, receiver operating characteristic.

Discussion

Non-invasive monitoring and evaluating the tumor response to treatment during the early stages of radiotherapy is critical to adapt specific therapeutic schedules, which can improve patients’ outcomes (3). In the present study, MD and ADC showed significant dynamic changes before and after treatment, and there were significant differences in MD between the good response group and the poor response group. In addition, MD showed a good correlation with cell density and apoptosis, suggesting the value of MD for monitoring the early response to radiotherapy in the rabbit VX2 malignant bone tumor model used in this study.

The MD value increased significantly during treatment. This could be attributed to post-radiotherapeutic changes, such as tumor lysis, loss of cell membrane integrity, and increased extracellular space, which promotes the movement of water molecules, thereby increasing the diffusion of water (14). Liu et al. (15) reported that the MD of osteosarcoma increases gradually during treatment, which is consistent with the present findings. We also explored the association between pathological changes in VX2 malignant bone tumors and MD changes during the same period to confirm the value of MD for predicting tumor response. The results indicated that tumors in model subjects that did not receive radiation showed a higher malignant behavior than tumors in treated subjects, indicated by higher cell density and smaller extracellular space. These parameters showed strong negative correlations with MD, which was consistent with the results in previous studies (8,10,16).

Apoptosis plays an important role in tumor development (17). In the present study, the number of apoptotic cells was higher in the post-treatment group than in the pre-treatment group, and apoptosis was negatively correlated with MD. A possible explanation for this result is that radiation triggers a pre-programmed cell death process that causes tumor cells to die spontaneously; this increases the extracellular space, which facilitates intra-tumoral water diffusion and increases tumor MD. Thus, an increase in post-therapeutic MD may indicate that radiotherapy accelerates tumor cell apoptosis and may enable the early detection of tumor response.

In this study, the good response group showed higher MD values than the poor response group after radiotherapy, whereas there were no significant differences in ADC between the two groups. In addition, ROC curve analysis indicated that MD was superior to ADC for identifying the early response to radiotherapy. There are two possible explanations for this result. First, the DKI sequence suggested that non-Gaussian displacement distribution could be used as an indicator of the diffusion behavior of water molecules to represent tissue heterogeneity (18,19). MD derived from DKI has a high b value to significantly deviate from the linear distribution, which may provide additional information to that obtained from ADC (7). Second, the cellular microstructure of the lesions was more complex in the good response group than in the poor response group. MD can reflect the molecular heterogeneity of water at the nanometer level and represent the true molecular diffusion, which is less affected by tissue characteristics, such as the curvature of blood vessels and extracellular space, which modify the diffusion coefficient (20).

In our study, the MK value was significantly lower in the post-treatment group than in the pre-treatment group. Tumor cells in the treatment group exhibited a lower cellularity and nuclear atypia, as well as a reduction in the interface of cellular tissues, resulting in a decrease in MK. The difference in the MK value between the good response and poor response groups was statistically significant. A possible explanation for this result is that tumor cells in the poor response group had more hypoxic and acidic environments, which may lead to necrosis and fibrosis. The related changes increase the structural complexity of the tissue and decrease the effectiveness of radiotherapy. These findings indicate that MK may be effective for evaluating changes in the tumor microenvironment, which is consistent with previous studies (21,22). The diagnostic performance of MK was inferior to that of MD and had lower specificity, which could be attributed to the fact that MD is more easily influenced by treatment-induced variations. Further research is needed to confirm its role in predicting treatment response.

As mentioned in the introduction, DKI has been used in the study of neoplastic lesions in various systems, but musculoskeletal tumors have been rarely evaluated. Musculoskeletal tumors can occur in any part of the body. However, the related image noise and diffusion-weighted image artifacts are different because of the effects of sensitivity artifacts caused by location, air, or bone. It is particularly important to obtain adequate signal-to-noise ratios (SNR) with high b values to ensure that the DKI parameter calculations are accurate (23). In our study, a 3-T MR scanner was used with a machine gradient of 80/200. Excellent gradient performance can greatly shorten the TE time, thereby allowing for ultra-high b-value imaging and ensuring the SNR of the image. Furthermore, the diffusion technology we used was RESOLVE imaging, which uses segmented readouts to shorten the TE time and which improved the SNR. Simultaneously, the SNR was controlled by adding the noise level parameter during the DKI scanning process, and images acquired in the lower value range were removed. In addition, we strictly followed previously published methods for data post-processing (19) and performed SNR control during acquisition to permit reliable DKI calculations.

The present study has some limitations. First, the ROI was selected and drawn manually, which may lead to selection bias. To minimize error, we obtained multiple measurements and calculated the average. Second, this study assessed the physiologic and histologic characteristics of animal models, which may not accurately reflect the conditions in human malignant bone tumors. Further research is necessary before the results of animal studies can be applied to clinical practice. In addition, the final value might be affected by perfusion because the value 1000 > b > 0 was not used in this study; this will be studied in future research. Finally, we used animal models in a limited time span, and tumor cell responses to long-term treatment were not assessed. Therefore, increasing the observation period is necessary in future studies to investigate the relationship between DKI and long-term changes after radiotherapy.

In conclusion, quantitative DKI parameters may reflect changes in rabbit VX2 malignant bone tumors in response to radiotherapy, to a certain extent. Our results indicated that MD was more effective in showing changes in cell characteristics after tumor treatment, suggesting that MD could be used to monitor the early response to treatment.

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received the following financial support for the research, authorship, and/or publication of this article: This work was supported by the National Natural Science Foundation of China (Grant no. 8167070553).

ORCID iDs

Cheng Dong

Wenjian Xu

References

Hameed

Dorfman

Primary malignant bone tumors–recent developments. Semin Diagn Pathol 2011; 28:86–101.

Koizumi

Inoue

Yamazaki

, et al. Perioperative fractionated high-dose rate brachytherapy for malignant bone and soft tissue tumors. Int J Radiat Oncol Biol Phys 1999; 43:989–993.

Baidya Kayal

Kandasamy

Khare

, et al. Intravoxel incoherent motion (IVIM) for response assessment in patients with osteosarcoma undergoing neoadjuvant chemotherapy. Eur J Radiol 2019; 119:108635.

Zhang

Chen

Zhang

, et al. Use of monoexponential diffusion-weighted imaging and diffusion kurtosis imaging and dynamic contrast-enhanced-MRI for the differentiation of spinal tumors. Eur Spine J 2020; 29:1112–1120.

Bauerle

Seyler

Munter

, et al. Diffusion-weighted imaging in rectal carcinoma patients without and after chemoradiotherapy: a comparative study with histology. Eur J Radiol 2013; 82:444–452.

Hayashida

Yakushiji

Awai

, et al. Monitoring therapeutic responses of primary bone tumors by diffusion-weighted image: Initial results. Eur Radiol 2006; 16:2637–2643.

Jensen

Helpern

Ramani

, et al. Diffusional kurtosis imaging: the quantification of non-gaussian water diffusion by means of magnetic resonance imaging. Magn Reson Med 2005; 53:1432–1440.

Zhong

Shi

Chen

, et al. Diffusion kurtosis imaging of a human nasopharyngeal carcinoma xenograft model: Initial experience with pathological correlation. Magn Reson Imaging 2018; 47:111–117.

Haopeng

Xuefei

Yan

, et al. Diffusion kurtosis imaging differs between primary central nervous system lymphoma and high-grade glioma and is correlated with the diverse nuclear-to-cytoplasmic ratio: a histopathologic, biopsy-based study. Eur Radiol 2020; 30:2125–2137.

10.

Genovesi

Filippone

Ausili

Cefaro G

, et al. Diffusion-weighted magnetic resonance for prediction of response after neoadjuvant chemoradiation therapy for locally advanced rectal cancer: preliminary results of a monoinstitutional prospective study. Eur J Surg Oncol 2013; 39:1071–1078.

11.

Chen

Jia

Malignant bone tumor intramedullary invasion: evaluation with dual-energy computed tomography in a rabbit model. J Comput Assist Tomogr 2015; 39:70–74.

12.

Chen

Zhang

Pang

, et al. The differentiation of soft tissue infiltration and surrounding edema in an animal model of malignant bone tumor: evaluation by dual-energy CT. Technol Cancer Res Treat 2019; 1;18:1533033819846842.

13.

Feng

Bian

, et al. Simultaneous changes of magnetic resonance diffusion-weighted imaging and pathological microstructure in locally advanced cervical cancer caused by neoadjuvant chemotherapy. J Magn Reson Imaging 2015; 42:427–435.

14.

Yuan

Qiao

Qiang

JW.

Diffusion-weighted imaging and diffusion kurtosis imaging for early evaluation of the response to docetaxel in rat epithelial ovarian cancer. J Transl Med 2018; 16:340.

15.

Liu

, et al. Monitoring Response to Neoadjuvant Chemotherapy of Primary Osteosarcoma Using Diffusion Kurtosis Magnetic Resonance Imaging: Initial Findings. Korean J Radiol 2019; 20:801–811.

16.

Liu

Sun

, et al. Clinical application of diffusion-weighted magnetic resonance imaging in uterine cervical cancer. Int J Gynecol Cancer 2015; 25:1073–1078.

17.

Smith

, et al. A comparative assessment of preclinical chemotherapeutic response of tumors using quantitative non-Gaussian diffusion MRI. Magn Reson Imaging 2017; 37:195–202.

18.

Tang

Sun

, et al. The value of diffusion kurtosis imaging in assessing pathological complete response to neoadjuvant chemoradiation therapy in rectal cancer: a comparison with conventional diffusion-weighted imaging. Oncotarget 2017; 8:75597–75606.

19.

Jensen

Helpern

JA.

MRI quantification of non-Gaussian water diffusion by kurtosis analysis. NMR Biomed 2010; 23:698–710.

20.

Rosenkrantz

Padhani

Chenevert

, et al. Body diffusion kurtosis imaging: Basic principles, applications, and considerations for clinical practice. J Reson Imaging 2015; 42:1190–1202.

21.

Song

, et al. The value of diffusion kurtosis magnetic resonance imaging for assessing treatment response of neoadjuvant chemoradiotherapy in locally advanced rectal cancer. Eur Radiol 2017; 27:1848–1857.

22.

Zheng

Chen

Zheng

, et al. Diffusion kurtosis imaging and tumour microstructure for monitoring response to radiotherapy in human nasopharyngeal carcinoma xenografts. Jpn J Clin Oncol 2020; 50:548–555.

23.

Ding

Tan

Mao

, et al. Differentiating between malignant and benign renal tumors: do IVIM and diffusion kurtosis imaging perform better than DWI? Eur Radiol 2019; 29:6930–6939.