Reduced field-of-view diffusion-weighted MRI in patients with cervical cancer

Introduction

Diffusion-weighted imaging (DWI) shows tissues characteristics based on water diffusion properties, which is related to tissue microenvironment, including tissue cellularity and the integrity of cell membranes.^1–3 It is well known that tissue apparent diffusion coefficient (ADC) value, which is calculated from DWI, allows quantitative evaluation of tissue diffusivity.⁴ With advances in MR techniques, DWI is widely included as a routine imaging protocol in many organs, including the female pelvis. The clinical application of DWI to cervical cancer has been investigated in many studies. Previous studies demonstrated that ADC value could be useful for differentiating cervical cancer from normal cervix as well as for prediction of the degree and histological type of cervical cancer.^5–9 Marc et al showed that using DWI resulted in higher reader confidence, sensitivity of tissue infiltration, and tumor-grading for cervical cancer, especially for less experienced reader.¹⁰ Park et al also reported that tumor ADC and parametrial invasion on MRI seemed to be independent predictors of pathologic parametrial invasion. Thus, adding DWI to MR imaging would improve accuracy for identifying low-risk patients for parametrial invasion, which is critical for appropriate treatment planning and improvement of patient outcomes.¹¹ In addition, DWI might have potential for assessing the therapeutic response to concurrent chemo-radiotherapy (CCRT) in advanced cervical cancer by measuring tumor ADCs or changes in tumor ADCs.^12–15 These studies were typically performed using single-shot echo-planar imaging (EPI) DWI. Despite these promising results, single-shot EPI has significant limitations; it is prone to susceptibility at air-tissue interfaces and has low spatial resolution.^16–19 As is well documented in prostate MR imaging in males, susceptibility artifacts at the interface of tissue with air in the rectum is also particularly problematic on DWI of the female uterine cervix.¹⁸

Recently, DWI sequences with reduced field-of-view (FOV) in the phase-encoding direction have been shown to improve image distortion with high spatial resolution in the spinal cord, breast, pancreas, and prostate gland.^20–27 To the best of our knowledge, application of reduced FOV DWI in the uterine cervix has not been reported. Therefore, we conducted this study to evaluate the qualitative image quality and quantitative ADC values of reduced FOV DWI in patients with cervical cancer in comparison to conventional DWI with 3 T MRI.

methods and Materials

Patients

This retrospective study had institutional review board approval, and informed consent was waived. We retrospectively searched the institutional pathologic database to find patients with biopsy-proven cervical cancer between April 2016 and February 2017. The search identified 44 patients with cervical cancer. The inclusion criteria for the patient group were: (1) biopsy-proven cervical cancer; (2) no history of conization or loop electrosurgical excision procedure before MRI, which could influence MRI interpretation; and (3) pelvic MR imaging including conventional DWI and reduced FOV DWI prior to initial treatment. Thus, a total of 22 patients (mean age, 53.9 years; range, 34–73 years) were included in our study (Figure 1). The clinical stage of cervical cancer was determined by obstetricians who used clinical examination, cystoscopy, and sigmoidoscopy according to the International Federation of Gynecology and Obstetrics classification.²⁸ Patients and tumor characteristics are shown in Table 1.

Download Figure

Table 1. Patients and tumor characteristics

No. of patients	22
Agea (range)	53.9 ± 9.1 years (34–73)
Tumor sizea (range)b	4.2 ± 3.2 cm  (1.5–12.5)
FIGO stage (no. of patients)
IA1	7
IBI	9
IB2	1
IIB	4
IVB	1
Histologic type
No. of squamous cell carcinoma	20
No. of adenocarcinoma	2

FIGO, International Federation of Gynecology and Obstetrics.

^aMean ± standard deviation.

^bTumors that are visible on MRI (n = 15).

Results

Qualitative analysis

Table 3 shows qualitative analysis scores between two DWI. For both observers, reduced FOV DWI achieved significantly better scores in anatomic detail, lesion conspicuity, presence of artifact, and overall image quality at both b = 0 s mm⁻² and b = 800 or 1000 s mm⁻² compared to conventional DWI (p < 0.05) (Figures 2 and 3), except for presence of artifact at b = 0 s mm⁻² with observer 2 (p = 0.083). The mean scores of both observers were also significantly higher on reduced FOV DWI than those on conventional DWI at both b values (p < 0.05). Interobserver agreement was fair to excellent on conventional DWI for anatomic detail, lesion conspicuity, presence of artifact, and overall image quality (κ = 0.450–0.815 at b = 0 s mm⁻² and κ = 0.388–0.712 at b = 1000 s mm⁻²). There was moderate to excellent agreement between the two observers on reduced FOV DWI for anatomic detail, lesion conspicuity, presence of artifact, and overall image quality (κ = 0.480–0.762 at b = 0 s mm⁻² and κ = 0.457–0.831 at b = 800 s mm⁻²) (Table 4).

Download Figure

Download Figure

Table 3. Comparison of qualitative analysis scores between reduced FOV DWI and conventional DWI

	Anatomic detail	Lesion conspicuity	Presence of artifact	Overall image quality
Observer 1
Reduced FOV DWI (b = 0 s mm−2)	3.82 ± 0.39	3.60 ± 0.51	3.23 ± 0.53	3.36 ± 0.58
Conventional DWI (b = 0 s mm−2)	3.00 ± 0.31	2.80 ± 0.41	3.00 ± 0.62	3.00 ± 0.44
P value	<0.001	0.001	0.025	0.011
Reduced FOV DWI (b = 800 s mm−2)	3.41 ± 0.50	3.60 ± 0.51	3.05 ± 0.58	3.36 ± 0.49
Conventional DWI (b = 1000 s mm−2)	3.05 ± 0.38	3.13 ± 0.52	2.82 ± 0.59	2.95 ± 0.38
P value	0.005	0.008	0.025	0.003
Observer 2
Reduced FOV DWI (b = 0 s mm−2)	3.68 ± 0.48	3.47 ± 0.64	3.00 ± 0.62	3.32 ± 0.65
Conventional DWI (b = 0 s mm−2)	3.00 ± 0.44	2.73 ± 0.46	2.86 ± 0.56	2.91 ± 0.43
Pp value	<0.001	0.001	0.083	0.003
Reduced FOV DWI (b = 800 s mm−2)	3.41 ± 0.59	3.60 ± 0.51	2.77 ± 0.69	3.23 ± 0.61
Conventional DWI (b = 1000 s mm−2)	2.82 ± 0.59	3.00 ± 0.53	2.59 ± 0.59	2.68 ± 0.57
P value	0.001	0.007	0.046	0.001
Mean
Reduced FOV DWI (b = 0 s mm−2)	3.75 ± 0.44	3.53 ± 0.57	3.11 ± 0.58	3.34 ± 0.61
Conventional DWI (b = 0 s mm−2)	3.0 ±0.37	2.77 ± 0.43	2.93 ± 0.50	2.95 ±0.43
P value	<0.001	<0.001	0.005	<0.001
Reduced FOV DWI (b = 800 s mm−2)	3.41 ± 0.54	3.60 ± 0.50	2.91 ± 0.64	3.29 ± 0.55
Conventional DWI (b = 1000 s mm−2)	2.93 ± 0.50	3.07 ±0.52	2.70 ±0.59	2.82 ±0.49
P value	<0.001	<0.001	0.003	<0.001

DWI, diffusion weighted imaging; FOV, field of view.

Data are mean ± standard deviation.

Table 4. Interobserver agreement for qualitative analysis scores

	Anatomic detail	Lesion conspicuity	Presence of artifact	Overall image quality
Reduced FOV (b = 0 s mm−2)	0.645 (0.296, 0.994)	0.762 (0.496, 1.000)	0.599 (0.307, 0.890)	0.480 (0.199, 0.760)
Reduced FOV (b = 800 s mm−2)	0.831 (0.624, 1.000)	0.722 (0.364, 1.000)	0.457 (0.148, 0.765)	0.748 (0.512, 0.984)
Conventional DWI (b = 0 s mm−2)	0.645 (0.196, 1.000)	0.815 (0.470, 1.000)	0.759 (0.507, 1.000)	0.450 (0.020, 0.880)
Conventional DWI (b = 1000 s mm−2)	0.505 (0.153, 0.856)	0.712 (0.344, 1.000)	0.626 (0.345, 0.906)	0.388 (0.058, 0.700)

DWI, diffusion weighted imaging; FOV, field of view.

Data are κ values.

Data in parentheses are 95% confidence intervals.

Discussion

Our results demonstrated that reduced FOV DWI achieved better image quality in terms of anatomic detail and lesion conspicuity compared to conventional DWI. Moreover, image artifacts were decreased on reduced FOV DWI. The reduced FOV DWI has previously been applied to other organs including the pancreas, breast, spinal cord, and prostate gland and has shown improvement of image quality with fewer artifacts.^20–27 Thus, our study was consistent with the results of previous reports and confirmed that the reduced FOV DWI sequence would have clinical value for cervical cancer. Obtaining DWI of the uterine cervix is particularly challenging due to the anatomical location between the air-containing rectum and urinary bladder, which are prone to susceptibility artifacts and distortion, which can subsequently affect the diagnostic performance, as with DWI of the prostate in males. In this respect, application of reduced FOV DWI on cervical imaging has potential advantages over conventional DWI using single-shot EPI.

Reduced FOV DWI sequence employs a 2-dimensional spatially selective echo-planar radiofrequency pulse and an 180° refocusing pulse reducing the FOV in the phase-encoding direction instead of conventional excitation in the single-shot EPI sequence. Consequently, this sequence facilitates high spatial resolution imaging with less susceptibility to distortion.^{24, 27} Implementation of this sequence also allows decreased partial volume averaging between tumor and normal tissue.^{22, 23} The better anatomic detail and lesion conspicuity observed in our study can be explained by high image resolution with decreased partial volume averaging of this technique. We believe this could lead to increased diagnostic performance in identifying small cervical cancer at an early stage, earlier detection of tumor recurrence, more accurate evaluation of tumor extent such as parametrial invasion, and improved assessment of treatment response. Further studies with small cervical cancer at early stage or post-treated tumor are needed in the future.

DWI with large FOV could help to detect pelvic lymph nodes. From this perspective, reduced FOV DWI could be less useful, although not covered in this study. However, according to the European Society of Urogenital Radiology guideline, T₁ weighted images without fat suppression are useful to evaluate for presence of lymphadenopathy.²⁹ Nevertheless, assessment of nodal involvement with cross sectional images that are largely dependent on the size of lymph nodes has significant limitations with low accuracy. Incorporating morphologic features such as margin irregularity, heterogeneity of nodal texture, and shape to the size, which are best seen at high-resolution T₂ weighted MR imaging have shown to be beneficial for differentiating benign and malignant nodes in patients with rectal cancer and may be applicable to those with cervical cancer.^{30, 33} Therefore, we think that it is best to use T₁ weighted images without fat suppression and high-resolution T₂ weighted imaging for assessing pelvic lymph nodes. Further research on the effect of FOV reduction, particularly regarding assessment of pelvic lymph nodes is needed.

In our study, we found no significant difference in the mean tumor ADC values between two DWI sequences, although reduced FOV sequence had a trend toward lower tumor ADC values. Our results were similar to those of previous reports.^20–22,34 On the contrary to our results, some studies showed a trend toward lower tumor ADC values for reduced FOV DWI compared to conventional DWI.^{23, 27} This difference is believed to be due to reduced partial volume averaging between tumor and normal tissue with the reduced FOV sequence and, consequently, more accurate ADC values.^{22, 23,27} In our study, mean tumor ADC values of conventional DWI sequences showed a trend toward higher values than those from the literature, which ranged from 0.757 to 1.11 × 10⁻³ mm² s⁻¹.^{5, 12,35} One possible explanation is that some of our study population had early stage cervical cancer with an invisible tumor compared with those of previous studies. Further, this could be partly explained by the complexity of ADC values that can be affected by multiple factors including T₂, b values, signal-to-noise ratio, and partial volume averaging.^{22, 26,36} Indeed, mean tumor ADC values of reduced FOV sequence were within the range of values from other studies.^{5, 12,35}

This study had several limitations. First, our study included only a small number of patients. Future studies are required with a larger population. Second, the parameters for reduced FOV and conventional DWI sequences were not exactly matched; in particular, the b-values of reduced FOV b = 0 and 800 s mm⁻² and those of conventional DWI were b = 0 and 1000 s mm⁻², which might have affected qualitative and quantitative image analyses. At higher b-values, the reduced FOV DWI acquisition would require increased averaging to boost the signal to noise (SNR), subsequently increasing scan time. In fact, it took more time to obtain reduced FOV DWI by increasing averaging to enhance the SNR in our study, which would affect the clinical use in practice. Further effort is needed to reduce scan time at higher b-values while maintaining SNR in the future. Third, interobserver variances in qualitative analysis have been reported^{21, 27} and also might have affected our results.

In conclusion, reduced FOV DWI showed better image quality in terms of anatomic detail and lesion conspicuity with fewer artifacts compared with conventional DWI in patients with cervical cancer. This could help radiologists to better assess cervical cancer in detail.