Diffusion tensor imaging of the anal canal at 3 tesla: Feasibility and reproducibility of anisotropy measures

Authors


  • This article was published online on 29 November 2011. Subsequently, an error was identified and the article was corrected on 27 January 2012.

Abstract

Purpose:

To assess the feasibility and reproducibility of 3-tesla diffusion tensor imaging (DTI) of the anal canal.

Materials and Methods:

DTI was performed in 25 men with no clinical history of anal canal disease undergoing MRI for prostate cancer. Analysis of fractional anisotropy (FA), relative anisotropy (RA), and apparent diffusion coefficient (ADC) were determined for the epithelial/subepithelial layer, internal sphincter, external sphincter, and puborectalis. The directionality of diffusion was recorded from color-coded tractography maps. Obturator internus and gluteus maximus served as reference muscles. Mean (SD) of values for FA, RA, and ADC were compared using analysis of variance. Intra and inter-rater agreement and test reproducibility (n = 5) was assessed by Bland-Altman statistics.

Results:

Mean (SD) for the epithelial/subepithelial layer, internal, external sphincter, and puborectalis were as follows: FA: 0.283 (0.099); 0.337 (0.049); 0.415 (0.072); and 0.407 (0.062), respectively. RA: 0.241 (0.094); 0.292 (0.050); 0.371 (0.083); 0.361 (0.067), respectively; and ADC: 1.49 (0.23); 1.59 (0.19); 1.51 (0.28); and 1.54 (0.29) × 10−3mm2/s, respectively. Good overall intra and inter-rater agreement and test–retest reproducibility was noted (coefficient of variation of 4.8–19.4% and 5.9–12.9%, respectively).

Conclusion:

Anisotropy is evident in the anal canal with good inter-rater agreement and test reproducibility. J. Magn. Reson. Imaging 2012;35:820–826. © 2011 Wiley Periodicals, Inc.

DIFFUSION TENSOR IMAGING (DTI) is a MRI technique that assesses the directionality of water diffusion in tissues, informing on its underlying microstructure and microdynamics (1, 2). Water diffusion in tissues is a three dimensional process and the degree of anisotropy depends on the underlying biophysical structure. Diffusion anisotropy is evident in the brain (3) and has been observed in other tissues including muscle, heart, uterus, and prostate (4–7). DTI has been shown to be clinically useful in the brain, for example, in the assessment of demyelinating disorders (8), brain maturation (9–11), and brain tumors (12–14). The technique has also been applied to pelvic organs, and has shown initial promise for the evaluation of prostate cancer (15–17). Due to the technical challenges of performing DTI of the bowel, including artifacts from peristalsis and respiration, and presence of bowel gas, there has been a paucity of bowel studies to date.

DTI assessment of the anal canal has clinical potential, by providing additional microstructural information. The anal canal has an important physiological function in maintaining bowel continence. Sphincter function may be disrupted following childbirth in 10% of women (18), and by disease such as anal canal cancer (19) or fistulous disease (20). Anatomical imaging, e.g., endoanal ultrasound (US), endoanal or pelvic phased array MRI currently complements physiological studies by depicting gross anatomical sphincter disruption (21–24). More subtle microstructural changes might be further depicted by DTI anisotropy measures. To date, the anisotropy of the anal canal has not been investigated. Data of the reproducibility of such measures are also lacking. The purpose of this study was to assess the feasibility of DTI of the anal canal, to establish the range of anisotropic measurements, and to determine the degree of intra-and inter rater agreement and test–retest agreement.

MATERIALS AND METHODS

Patients

Twenty-five consecutive male patients (mean age, 69 years; range, 58–86 years) with prostate cancer undergoing staging MRI (Stage T1: 7; T2: 16; T3; 2; all N0) at 3 tesla (T) over a 10-month period between November 2008 and May 2009 were assessed. Exclusion criteria were as follows: (i) standard exclusion criteria for MRI; (ii) previous clinical history of anal disease or dysfunction (n = 2); (iii) previous anorectal surgery (n=0); (iv) previous chemoradiation to the pelvis (n = 3); (v) presence of an anal canal lesion or signal abnormality on morphological MRI (n=12). Institutional review board approval with waiver of informed consent was obtained for this retrospective analysis.

MRI

Diffusion weighted imaging including diffusion tensor sequences were part of the standard comprehensive examination protocol for these patients at our center which also included morphological sequences, dynamic contrast enhanced MRI and spectroscopy. MRI was performed at 3T (Tim Trio, Siemens Healthcare, Erlangen, Germany) using a multichannel phased array coil. Diffusion weighted imaging using b values of 0, 1000, 1400, and 1600 s/mm2 was performed. The DTI sequence was the final sequence, acquired in the axial plane, encompassing the length of the anal canal, with the following parameters: single shot echo planar sequence with parallel acquisition, parallel imaging factor 2; TR 2900 ms; TE 80 ms; FOV 260 mm, matrix 99 × 128, NEX 3, section thickness 5 mm, b values: 0 and 800 s/mm2; 12 diffusion gradient orientations; pixel bandwidth 1600 Hz, spectral fat suppression, acquisition time: 3 min.

Diffusion Tensor Analysis

Fractional anisotropy (FA) is defined as follows (1, 25):

equation image

Where (D) = 1/3 (λ1+ λ2+ λ3), the mean diffusivity (apparent diffusion coefficient, ADC), and lambda, λ = eigenvalue

Relative anisotropy (RA), the ratio of the variance of the computed eigenvalues (lambda, λ) to their mean, is defined as follows (1, 25):

equation image

Where (D) = 1/3 (λ1+ λ2+ λ3), the mean diffusivity (apparent diffusion coefficient, ADC).

Image Analysis

Images were transferred to a standard commercial workstation (Leonardo, Siemens Healthcare, Erlangen, Germany) and analyzed independently by two board certified radiologists with an interest in abdominal MR imaging, (with 7 and 13 years MRI experience), using standard commercial software for diffusion tensor analysis (Neuro3D, Siemens Healthcare, Erlangen, Germany). Fused registered T2 and b0 images were viewed initially on the workstation using the 3D viewing platform, to confirm the position of anatomical structures of interest, and to facilitate subsequent region of interest placements.

The diffusion tensor software automatically generated and displayed the following pixel maps: b0, fractional anisotropy (FA), relative anisotropy (RA), and diffusivity (apparent diffusion coefficient; ADC). FA and RA provide complementary information. FA measures the fraction of total diffusivity that can be ascribed to anisotropic diffusion (i.e., directionality; values range from 0 to 1, where 1 = highly directional). RA represents the ratio of the anisotropic part to its isotropic part, or a normalized standard deviation (values range from 0 to √2) (1, 25). Pixel interpolation and a 40% threshold (to reduce background signal) for display of each parameter map were used by the software.

Measurements were obtained in the axial plane at mid-canal level for the following anal canal components: epithelial/subepithelial layer, internal anal sphincter, and external anal sphincter. Measurements were also obtained for the puborectalis (pubovisceralis) muscle separately. The obturator internus and gluteus maximus muscles served as controls.

The anal canal components were defined as follows (Fig. 1): (i) Epithelial/subepithelial layer: Central portion of the canal consisting of the lining mucosa/epithelium and subepithelial layer; (ii) Internal anal sphincter: Slightly hyperintense circular structure surrounding the central epithelial/subepithelial layer on T2 weighted sequences, the inferior border terminating 1cm above the inferior edge of the sphincter complex; (iii) External anal sphincter: Outer cylinder of the anal canal isointense to muscle on T2 weighted imaging, the upper half integrating with the sling-like puborectalis (pubovisceralis). Posterior fibers are continuous with the anococcygeal ligament and anterior fibers with the transverse perineal muscle and perineal body.

Figure 1.

T2 weighted axial (a), fused T2 weighted and b=0 s/mm2 diffusion weighted axial (b) and b=0 s/mm2 diffusion weighted axial (c) images demonstrating the anal canal components evaluated: epithelial/subepithelial layer (ESE); internal anal sphincter (IAS); external anal sphincter (EAS); and fat within the intersphincteric space (IS).

ROIs were positioned on the b0 images, which had been previously registered and fused to corresponding T2-weighted images to confirm the exact anatomical position of the structures of interest. These ROIs were replicated automatically by the software onto the corresponding displayed FA, RA, and ADC maps. The largest possible region of interest for each anal canal component was delineated freehand (mean pixel area, 239 mm2; range, 15.6 mm2 to 1076 mm2). For the reference muscles, a circular region of interest (mean, 112 mm2) was placed centrally within the muscle belly. The mean values were displayed for each ROI for each parameter, FA, RA, and ADC (×10−3mm2/s), were recorded for each component. The directionality of anisotropy was color-coded on the parameter maps as follows: antero-posterior (green), medio-lateral (red) and cranio-caudal (blue). Thus, the dominant directionality for each structure was recorded from the color-coding of the structures of interest as follows: antero-posterior, medio-lateral, and cranio-caudal.

Analysis was repeated in the same manner with > 8 weeks in between analyses to assess intra-rater agreement. Test–retest reproducibility was assessed in 5 patients where repeated studies (within 24 h) had been obtained as part of quality assurance. The signal to noise ratio, the ratio of signal intensity of the anal canal to background on the trace images, was also assessed to gauge image quality.

Statistical Analysis

Means (standard deviation, SD) were calculated for the FA, RA, and ADC values for each anatomical structure. Parameters were compared using analysis of variance; statistical significance was at 5%. Posttesting was performed using the Bonferroni method. Intra-and inter-rater agreement and test-retest reproducibility was assessed using Bland-Altman statistics: the mean difference, 95% limits of agreement (mean difference − 2 SD and mean difference + 2 SD), and within subject coefficient of variation (measurement error relative to the parameter value) were assessed.

RESULTS

Anisotropy Measures

DTI was feasible in each patient, with parametric maps corresponding to underlying anatomy (Fig. 2). Signal to noise ratio (SNR) was acceptable with a mean (SD) SNR of 6.55 (2.81). Mean (SD) values for FA, RA, and ADC for the anal canal components and reference muscles are given in Table 1.

Figure 2.

T2-weighted axial image (a) and corresponding diffusion weighted b0 s/mm2 (b), ADC (c) and color coded FA (d) maps at mid anal canal level. Abbreviations: epithelial/subepithelial layer (ESE); internal anal sphincter (IAS); external anal sphincter (EAS); intersphincteric space (IS).

Table 1. Means (standard deviation, SD) of fractional anisotropy (FA), relative anisotropy (RA), and apparent diffusion coefficient (ADC) for the anal canal and pelvic musculature
Anal canalFARAADC × 10−3mm2/s
Epithelium/ subepithelium0.283 (0.099)0.241 (0.094)1.49 (0.23)
Internal anal sphincter0.337 (0.049)0. 292 (0.050)1.59 (0.19)
External anal sphincter0.415 (0.072)0.371 (0.083)1.51 (0.28)
Puborectalis0.407 (0.062)0. 361 (0.067)1. 54 (0.29)
Mean0.361 (0.089)0.317 (0.088)1.53 (0.21)
Reference muscles   
Obturator internus0.480(0.089)0. 437(0.095)1.24 (0.33)
Gluteus maximus0.374 (0.052)0. 325(0.062)1.59 (0.28)
Mean0.427 (0.087)0.381 (0.093)1.42 (0.36)

Analysis of variance demonstrated FA and RA differed for components of the anal canal (P ≤ 0.0001), whereas ADC did not (P = 0.44). Post-testing demonstrated that FA was significantly lower for the epithelial/subepithelial layer than the internal or external sphincter/puborectalis (Fig. 3; P < 0.0001 to P = 0.02). Similar findings were noted for RA (Fig. 3; P < 0.0001 to P = 0.03). FA and RA were also significantly different for the internal anal sphincter and the external anal sphincter/puborectalis (Fig. 3; P < 0.001 to P = 0.03).

Figure 3.

Box plot (median, quartile and range) of values for FA (a), RA (b), and ADC (c) for the different anal canal components. Two-tailed P-values from Bonferroni testing are shown

Overall FA and RA values of the anal canal were lower than those of skeletal muscle (P < 0.0001). Fiber orientation mapping demonstrated that the predominant direction of diffusivity was in the cranio-caudal direction for the epithelial/subepithelial layer; in the antero-posterior direction for the internal anal sphincter along the lateral aspect and in the medio-lateral direction along the anterior and posterior aspects of this structure reflecting a circular configuration; in a combined circular and cranio-caudal direction for the external anal sphincter reflecting the contribution of the longitudinal distribution of muscle fibers; and in a circular configuration for puborectalis (Fig. 4).

Figure 4.

Color coded tractography image of the anal canal demonstrating the directionality of diffusivity of the anal canal components (cranio-caudal = blue; antero-lateral = green; medio-lateral = red). Abbreviations: internal anal sphincter (IAS); external anal sphincter (EAS); intersphincteric space (IS); transverse perineal muscle (TP). [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Intra- and Inter-rater Agreement

Intra-rater Agreement

The mean difference and 95% limits of agreement for FA, RA, and ADC for anal canal components and control muscles are summarized in Table 2. The intra-rater agreement for the anal canal components was acceptable with a coefficient of variation of 6 to 9% (Table 2). Intra-rater agreement for FA and RA were not significantly different for the anal canal components or control muscles.

Table 2. Intra-rater agreement for fractional anisotropy (FA), relative anisotropy (RA), and apparent diffusion coefficient (ADC) for the anal canal and pelvic musculature*
Anal canal n=20Mean difference95% limits of agreementwCV (%)
  1. The mean difference, 95% limits of agreement and within-subject coefficient of variation (wCV) are shown.

FA   
Epithelium/subepithelium−0.006−0.068 to +0.0557.6
Internal anal sphincter+0.009−0.058 to +0.0756.9
External anal sphincter+0.006−0.072 to + 0.0846.9
Puborectalis−0.002−0.082 to +0.0776.8
RA   
Epithelium/subepithelium+0.004−0.098 to +0.1077.9
Internal anal sphincter+0.025+0.084 to + 0.1348.5
External anal sphincter+0.014−0.068 to +0.0958.4
Puborectalis−0.002−0.097 to +0.0948.9
ADC   
Epithelium/subepithelium+0.019−0.247 to +0.2866.4
Internal anal sphincter+0.014−0.322 to +0.3507.4
External anal sphincter−0.017−0.430 to +0.3969.4
Puborectalis−0.012−0.288 to +0.2636.7
Reference muscles   
FA   
Obturator internus+0.025−0.084 to +0.1349.1
Gluteus maximus−0.002−0.084 to +0.0807.7
RA   
Obturator internus+0.020−0.115 to +0.15611.6
Gluteus maximus+0.005−0.074 to +0.08348.5
ADC   
Obturator internus−0.015−0.624 to +0.59417.2
Gluteus maximus+0.076−0.231 to +0.3837.2

Inter-rater Agreement

The mean difference and 95% limits of agreement for FA, RA, and ADC for anal canal components and control muscles are summarized in Table 3. Mean difference was not significantly different to zero indicating that there was no overall bias for one or other of the raters. The inter-rater agreement for the anal canal components was acceptable with a coefficient of variation of 5 to 14% (Table 3). Inter-rater agreement for FA and RA were not significantly different for the anal canal components or control muscles.

Table 3. Inter-rater Agreement for Fractional Anisotropy (FA), Relative Anisotropy (RA), and Apparent Diffusion Coefficient (ADC) for the Anal Canal and Pelvic Musculaturea
Anal canal n=20Mean difference95% limits of agreementwCV (%)
  1. a

    The mean difference, 95% limits of agreement and within-subject coefficient of variation (wCV) are shown.

FA   
Epithelium/subepithelium−0.004−0.108 to +0.09912.8
Internal anal sphincter−0.015−0.067 to +0.0447.0
External anal sphincter+0.012−0.111 to + 0.13510.7
Puborectalis−0.006−0.096 to +0.0837.8
RA   
Epithelium/subepithelium−0.006−0.099 to +0.08613.5
Internal anal sphincter−0.017−0.084 to +0.0498.8
External anal sphincter+0.005−0.093 to +0.1029.4
Puborectalis−0.011−0.109 to +0.0869.6
ADC   
Epithelium/subepithelium+0.016−0.210 to +0.2425.5
Internal anal sphincter−0.018−0.228 to +0.1914.8
External anal sphincter−0.049−0.413 to +0.51211.5
Puborectalis+0.089−0.256 to +0.4269.2
Reference muscles   
FA   
Obturator internus+0.004−0.086 to +0.0926.5
Gluteus maximus−0.002−0.072 to + 0.0676.4
RA   
Obturator internus+0.005−0.092 to +0.0837.0
Gluteus maximus−0.006−0.056 to +0.0455.5
ADC   
Obturator internus+0.001−0.641 to +0.64518.7
Gluteus maximus−0.109−0.967 to +0.74919.4

Test–Retest Reproducibility

The mean difference and 95% limits of agreement for FA, RA, and ADC for anal canal components and control muscles are summarized in Table 4. Reproducibility was acceptable for the anal canal components with a coefficient of variation of 6 to 13% (Table 4).

Table 4. Test–retest agreement (reproducibility) for fractional anisotropy (FA), relative anisotropy (RA), and apparent diffusion coefficient (ADC) for the anal canal and pelvic musculaturea
Anal canal n=5Mean difference95% limits of agreementwCV (%)
  1. a

    The mean difference, 95% limits of agreement and within-subject coefficient of variation (wCV) are shown.

FA   
Epithelium/subepithelium−0.004−0.073 to +0.0647.0
Internal anal sphincter+0.013−0.065 to +0.0927.6
External anal sphincter−0.002−0.075 to +0.0726.0
Puborectalis−0.022−0.086 to +0.0427.0
RA   
Epithelium/subepithelium−0.015−0.117 to +0.08712.9
Internal anal sphincter+0.014−0.059 to + 0.0878.2
External anal sphincter−0.004−0.074 to + 0.0646.6
Puborectalis−0.020−0.089 to + 0.0498.4
ADC   
Epithelium/subepithelium+0.02−0.289 to +0.3286.2
Internal anal sphincter+0.08−0.223 to +0. 3726.9
External anal sphincter+0.02−0.251 to +0.2945.9
Puborectalis+0.03−0.268 to +0.3316.0
Reference muscles   
FA   
Obturator internus+0.022−0.085 to +0.1289.9
Gluteus maximus−0.029−0.006 to +0.0646.4
RA   
Obturator internus+0.023−0.080 to +0.12711.2
Gluteus maximus+0.028−0.006 to +0.0637.1
ADC   
Obturator internus+0.09−0.426 to +0.2479.6
Gluteus maximus+0.004−0.323 to +0.3336.3

DISCUSSION

Until the recent introduction of higher field MR systems and parallel imaging technology, artifacts from bowel and respiratory motion, vascular pulsation, and susceptibility artifacts from luminal gas have limited diffusion weighted and diffusion tensor studies of the bowel. MRI is the primary imaging modality for many anal canal diseases, including anal cancer (19) and fistulous disease (20). Diffusion weighted and tensor sequences can be integrated readily into standard clinical examinations and may potentially provide additional microstructural information.

Our study confirms that anal canal components demonstrate anisotropy. The anal canal consists of an inner epithelial/subepithelial layer, surrounded by the internal anal sphincter comprising smooth muscle, an inter-sphincteric fibro-muscular layer, and the external sphincter comprising striated muscle, divided anatomically into subcutaneous, superficial and deep components, merging superiorly with the puborectalis (pubovisceralis) muscle (27). In our study, the directionality of diffusion of the inner epithelial lining and vascular subepithelium was predominantly in the cranio-caudal direction reflecting the anal columns, the vertically orientated mucosal folds. The directionality of the internal sphincter was consistent with the known circular orientation of the smooth muscle fibers. The external anal sphincter also demonstrated anisotropy, with contribution from an antero-posterior and cranio-caudal direction reflecting the orientation of the underlying striated muscles.

The higher the FA and RA values the more ordered and directional a structure is. FA and RA differed for the anal canal components with lowest values for the internal epithelial/subepithelial lining, reflecting a less ordered/directional microstructure in comparison with muscle. FA and RA values were also different for the internal anal sphincter versus the external anal sphincter and puborectalis belying the difference in muscle type (smooth versus striated muscle, with higher values for striated muscle). There was no difference in ADC indicating no difference in overall diffusitivity. Overall FA and RA values were significantly lower for the anal canal than reference pelvic musculature, consisting of striated muscle, as might be anticipated from previous published musculoskeletal data (4).

The degree of inter-rater agreement was acceptable for clinical practice. Agreement was slightly narrower for the anal canal than reference muscles, ranging from 4.7% to 13.5% for FA, RA, and ADC for the anal canal, and for 5.5% to 19.5% for reference pelvic muscles. This greater variation between readers reflects variations in region of interest placements for pelvic muscle due to the greater available area. The inter-rater agreement was comparable or better than other studies, for example, cerebral white matter tracts where intraclass correlation coefficients were fair ranging from 0.31 to 0.88 depending on site of placement (28).

While we have shown that this technique is feasible with no additional patient preparation required, and that measures are reproducible between two different observers, there are limitations. First, imaging was performed at 3T to ensure signal to noise was higher than at 1.5T, however, this limited the number of evaluable patients as fewer patients were 3T-compatible: patients with prostate cancer are typically an older population, and therefore more likely to have undergone previous surgery, have a higher prevalence of metal prostheses, and cardiovascular disease precluding evaluation on our 3T scanner. The SNR was recorded from the trace images as the ratio of signal intensity of the anal canal to background signal. This was not high but acceptable at a mean of 6.55. This is a recognized issue for musculoskeletal DTI, reflecting the MR properties of muscle, and comparable to previous publications (29). Second, only male patients were evaluated, and these were from an older age group. It is well recognized that sex and age related anatomical sphincter variations are present (30). Furthermore, there are likely to be age and gender related differences in FA and RA measurements, as has been shown in the brain (28), which requires further prospective evaluation. However, it must be borne in mind that there is relatively high prevalence of occult sphincter damage in the postpartum female population related to vaginal delivery, which is difficult to exclude from retrospective case-note and imaging review alone. Third, artifacts related to rectal air (superiorly) and in the natal cleft (inferiorly) can result in EPI induced image distortion. For this reason, we undertook measurements at mid-canal level to demonstrate as proof of principle that DTI is a feasible technique. Fourth, there is no pathological correlation, however, previous ex vivo studies have been performed in other organs, e.g., uterus indicating that observed anisotropy does indeed reflect muscle fiber orientations (6).

To date, there have been no published DTI data exploring the clinical potential of this technique for evaluating the anal canal. Exploratory studies in extra-cranial cancers, for example, prostate (15–17) and breast (31) have shown a decrease in anisotropy in the presence of tumor. A recently published feasibility study of DTI of the female pelvic floor (32) has suggested that it may provide new insights into pelvic floor imaging. The clinical potential in anal canal disease has yet to be explored prospectively. However, measures of anisotropy may provide additional information of microstructural disruption secondary to trauma to the anal canal, pelvic radiotherapy, fistulous disease, or tumor. It is anticipated that these may decrease anisotropy values as a result of the loss of the normal fiber orientation.

In conclusion, we have shown that the anal canal demonstrates anisotropy on diffusion tensor imaging, and that anisotropy measures may be derived reproducibly. DTI has potential as a non-invasive means of investigating anal sphincter microstructural damage, and warrants clinical exploration.

NOTE ADDED IN PROOF

Corrections were made after initial online publication to page 1 (abstract, correspondence), 3 and 4 (formatting).

Ancillary