Objective:

To investigate the applicable value of diffusion kurtosis imaging (DKI) in assessing the pathological complete response (pCR) to neoadjuvant chemoradiation therapy (CRT) in locally advanced rectal cancer (LARC).

Methods:

Fortyconsecutive patients diagnosed with LARC were prospectively enrolled and underwent MRI before and after CRT on 3.0 T MRI. Apparent diffusion coefficient (ADC)

mean diffusion (MD)

and mean kurtosis (MK) values of the tumor were measured in pre-CRT and post-CRT phases and compared with histopathologic findings after total mesorectal excision (TME). According to the athological results

the patients were divided into pCR group and non-pCR group. The diffusion weighted imaging (DWI) and DKI parameters were compared between pCR and non-pCR groups

using the nonparametric Mann-Whitney U test. The Kruskal-Wallis test was used to assess differences in the means between pCR and non-pCR groups. The sensitivity

specificity

positive predictive value (PPV)

negative predictive value (NPV)

accuracy and area under the receiver operating characteristic (ROC) curve (AUC) for the evaluation of pCR were also calculated and compared between DKI and DWI models.

Results:

For a total of 40 rectal lesions (pCR

n

=10; non-pCR

n

=30)

the MKpre and MKpost values in pCR group (0.720.10 and 0.560.06

respectively) were significantly lower than those in non-pCR group (0.870.11 and 0.670.08

respectively) (

P

0.001). The ADCpost and the ratio of apparent diffusion coefficient (ADCratio) values were significantly higher in pCR group (1.310.13 and 0.640.40

respectively) than those in non-pCR group (1.150.18 and 0.360.29

respectively) (

P

=0.011 and

P

=0.026

respectively). In addition

the MDpost and the ratio of mean diffusion (MDratio) (2.510.34

vs

. 1.990.30

P

=0.001; 0.820.51

vs

. 0.370.34

P

=0.003

respectively) were significantly different

whereas the ADCpre

MDpre

and the ratio of mean kurtosis (MKratio) were not significantly different between the two groups (

P

=0.499

0.510 and 0.589

respectively). The area under the receiver operating characteristic curve (AUROC) for the assessment of pCR was greater using MKpost (0.893

cutoff value=0.590 5) compared with other parameters. And overall accuracy of MKpost (90%) was the highest.

Conclusion:

Both DKI and conventional DWI exhibit potential for predicting treatment response to neoadjuvant CRT in rectal cancer. The DKI parameters

especially MKpost

shows higher specificity than conventional DWI in assessment of pCR and non-pCR in the patients with LARC

while the pre-CRT ADC and MD values are not reliable.