Diffusion tensor imaging of the healthy feline kidney

Abstract

Objectives

The aim of this study was to describe diffusion tensor imaging (DTI) and its derived parameters, such as apparent diffusion coefficient (ADC), fractional anisotropy (FA), radial diffusivity (RD) and axial diffusivity (AD) of the feline kidneys.

Methods

A total of 17 cats with clinically normal urinary tracts underwent 3 T MRI with DTI sequence (b-values of 0 and 800 s/mm², 15 diffusion directions). DTI images were analysed and sampled in manually drawn regions of interest (ROIs) on three transverse slices, with two different sampling methods (small and large ROIs), on the cortex and on the medulla separately, and on the entire parenchyma.

Results

DTI images of the kidneys of 17 cats were analysed. ADC, FA, RD and AD values differed significantly between the cortex and medulla with both sampling methods.

Conclusions and relevance

The results of the present study indicate microstructural similarities between feline and human kidneys. This study provides baseline data for healthy feline kidneys for future potential application in the early diagnosis of renal pathologies in cats.

Introduction

Chronic kidney disease (CKD) is very common in cats, with prevalence ranging from 37.5% in younger cats (aged 0–4.9 years) to up 80% in cats aged over 15 years. CKD is characterised by irreversible and progressive loss of renal functionality.^1,2 Irreversibly damaged nephrons are replaced by fibrosis, irrespective of the original cause of injury, ³ leading to loss of functionality. ⁴ In the majority of cats, clinical signs and laboratory values compatible with chronic renal failure are only present when damage affects 80–85% of all nephrons, because of the large functional renal reserve and the compensatory hypertrophy of remaining viable nephrons.^3,5 Currently, kidney biopsy is the only method that provides a definitive diagnosis of renal fibrosis; however, it is an invasive procedure, subject to sampling error ⁶ and unsuitable for disease monitoring.

Ultrasound is routinely indicated in cases of suspected CKD but the findings are non-specific, do not always correlate with kidney functionality ⁷ and provide no data about renal function. ⁸

Renal MRI, particularly when combined with functional images, can provide information not only about renal anatomy, but also about physiological and functional parameters. ⁹

Diffusion-weighted imaging (DWI) MRI investigates the Brownian random motion of water molecules in biological tissues. ¹⁰ Diffusion tensor imaging (DTI) is a development from DWI, which quantifies diffusion in different directions. ⁵ DTI analyses diffusion along multiple directions through multiple diffusion acquisitions (directions) ⁶ and demonstrates the anisotropic nature of the kidney. ¹¹ DTI characterises the analysed tissue through multiple parameters: the apparent diffusion coefficient (ADC, or mean diffusivity) is the average diffusivity in all directions; axial diffusivity (AD) represents the first eigenvalue (along the longitudinal direction of the tensor in a cylindrical model); and radial diffusivity (RD) is the average of the second and third eigenvalues (radial direction, in a cylindrical model). ¹² Fractional anisotropy (FA) quantifies the difference of directionally dependent diffusion within a voxel of interest.

The use of DTI for renal evaluation has never been investigated in cats.

The aims of the present study were to assess the feasibility of DTI sequences applied to feline subjects, to assess the ADC, FA, RD and AD values in the kidneys of clinically healthy cats, and to compare ADC,FA, RD and AD values obtained from different sampling strategies during image analysis.

Materials and methods

Animals

A total of 18 healthy cats were prospectively enrolled for an independent, concurrent main dietary study, approved by the cantonal authority (number ZH162/2022, to be published). The origins of the cats were as follows: 15 were from the approved breeding permissionof the Vetsuisse Facility of the University of Zurich (Stiegenhof, Oberembrach, Zurich, Switzerland); and three were from the experimental animals of animal nutrition unit at Ludwig-Maximilians-University (Munchen, Hackerstrs 27, Badersfeld, Oberschleissheim (Germany).

Before enrolment, the cats underwent a clinical examination, blood (serum creatinine, blood urea nitrogen, symmetric dimethylarginine) and urine (pH, specific gravity, creatinine) analysis, blood pressure measurement (mean Doppler) and ultrasound of the kidneys. Cats were included in the main study if blood work, blood pressure measurements and ultrasonographic examination were normal. Cats showing any kind of abnormalities referable to nephropathy would have been excluded from the study.

Clinical examination, blood and urine analysis, and blood pressure measurement were performed by a veterinarian (DS) specifically trained by a specialist in internal medicine (AM, DipACVIM). Mean serum creatinine was 106.5 µmol/l (range 81–136), urea nitrogen was 8.2 mmol/l (range 6.4–10), mean symmetric dimethylarginine was 11 µg/dl (range 9–14), urine pH was 7.4–9, mean specific gravity was 1.028–1.030 g/ml, creatinine was 9056–11,757 µmol/l and mean blood pressure was 142 mmHg (range 115–200).

The ultrasonographic examination was performed by a veterinarian specialised in diagnostic imaging (FD, DipECVDI). The urinary tract was examined ultrasonographically following a standardised procedure, after clipping the ventral abdomen and applying coupling gel to the skin. Cats were typically awake, and were gently manually restrained during the examination. They were scanned in right lateral and/or dorsal recumbency. For the ultrasonographic examination a LogiqE10 machine (GE HealthCare) was used with different probes (curvilinear array, linear array and hockey stick) depending on the animal, with a frequency in the range of 5–18 MHz. The cortex of the left kidney was subjectively isoechogenic to the spleen in 8/18 cats, mildly hypoechogenic in 7/18 cats and mildly hyperechogenic in 3/18 cats. The cortex of the right kidney was always subjectively mildly hyperechogenic to the liver parenchyma. The medulla was always hypoechogenic to the cortex and both were homogeneous. The capsule was smooth in 16/18 cats and mildly irregular in two cats. Median kidney length:aorta ratio was 11.6 for the left kidney, 11.8 for the right; both were considered normal. ¹³

The enrolled cats underwent MRI examinations before the beginning of the dietary study, which included morphological images (T1- and T2-weighted) and a DTI sequence as an add-on to the morphological MRI examination required by the dietary study.

Anaesthesia

The MRI examination was performed under general anaesthesia and cats were fasted for 12 h before the induction of anaesthesia. Cats were premedicated with 0.2 mg/kg butorphanol tartrate, and 16/18 cats with medetomidine (10/16 cats: 0.5 µg/kg; 6/16 cats: 1 µg/kg). All cats were induced with ketamine hydrochloride (1 mg/kg) and alfaxalone (10/18 cats: 1 mg/kg; 4/18 cats: 1.5 mg/kg; 2/18 cats: 0.5 mg/kg; one cat each 2 or 3 mg/kg). Anaesthesia was maintained with isofluorane, with pressure-controlled ventilation and with oxygen and air (0.1 l/min in 13/18 cats and 0.2 l/min in 5/18). Peripheral oxygen saturation (SpO₂) was maintained typically at 100%, and infusion volume was in the range of 12–20 ml/kg.

MRI examination

The cats were scanned in dorsal recumbency in the MRI unit at the Vetsuisse Faculty, University of Zurich teaching hospital, using a 3 T MRI scanner (Ingenia 3.0T; Philips AG Healthcare) and a small extremity coil (dStream Small Extremity 8ch coil; Philips) covering the cranial abdomen to the inguinal area.

Morphological images were acquired with the reported setting: T2-weighted images in the transverse plane (SE; rotation time [TR] 3332 ms; echo time [TE] 109 ms; slice thickness 3 mm; flip angle 90°; field of view [FOV]: adapted to animal); T2-weighted images in the dorsal plane (spin echo [SE]; TR 3500 ms; TE 119 ms; slice thickness 3 mm; flip angle 90°; FOV: adapted to animal); and a three-dimensional T1 sequence in the transverse plane (gradient echo; TR 4.6 ms; TE 2285 ms; slice thickness 3 mm; flip angle 10°; FOV: adapted to animal). DTI diffusion-weighted images were acquired on a transverse plane with the following setting: SE; TR 2762 ms; TE 84.5 ms; slice thickness 2.5 mm; flip angle 90°; FOV: adapted to animal, two b-values: 0 and 800; and 15 diffusion directions.

Data analysis

All images were transferred and assessed with dedicated software and workstation (IntelliSpace Portal; Philips AG). Registration of the DTI data set was performed, if necessary, to correct distortion by Eddy current motions. Different maps were created, and subjectively and qualitatively assessed: diffusion images at b0 and b800, grayscale FA map, colour FA map, axial and radial diffusivity map, and tractography (Figure 1).

Figure 1

Example of the different maps qualitatively assessed on the transverse view crossing the renal pelvis of the right kidney in one representative cat of the study. (a) Diffusion image at b0; (b) diffusion image at b800; (c) apparent diffusion coefficient map; (d) axial and (e) radial diffusivity map; (f) grayscale fractional anisotropy (FA) map; (g) colour FA map; and (h) tractography

Qualitative assessment of the images included assessment of image quality, presence of obvious artefacts and subjective comparison of the signal intensities of the cortex and the medulla.

Quantitative assessment of the images was performed at b0, with recording of ADC, AD, RD and FA in different regions of interest (ROIs). ROIs were manually drawn on the transverse view in the mid (image crossing the pelvis), cranial and caudal half of the kidney parenchyma (typically one or two slices cranially or caudally to the mid of the kidney, respectively, as recommended ¹⁴ ). Two methods of sampling were tested, with small and large ROIs. Small ROIs (approximately 10–15 mm² in size) were drawn with an elliptic cursor centred on the cortex and on the medulla on the dorsolateral region of the kidney (Figure 2a).

Figure 2

Example of the regions of interest (ROIs) drawn for the calculation of the tensor parameters, on the transverse image crossing the renal pelvis of the right kidney in one representative cat of the study at b = 0. (a) Small ROI method: elliptic ROI centred on the cortex (solid line) and on the medulla (dotted line). (b) Large ROI method: ROIs drawn freehand on the entire cortex (solid line), medulla (dotted line). (c) Entire kidney parenchyma. The top of the image is ventral in the cat

Large ROIs were drawn freehand on the entire cortex, medulla (Figure 2b) and entire kidney parenchyma (excluding the collecting system, if imaged, Figure 2c).

Slices including large renal vessels or visible artefacts were excluded from the analysis.

ADC, AD and RD (expressed in 10^–3mm²/s), FA (in absolute number) and ROI size (expressed in mm²) were recorded for each ROI.

FA colour map and tractography images were generated. The following parameters were used for the tractography images: FA threshold 0.2; angle threshold 55°; and colour coding direction.^{15 –17} In conventional colour-coded FA and tractography images, red represents a left–right direction, blue represents a craniocaudal direction and green represents a dorsoventral direction.

Statistical analysis

Data were analysed using a statistical software package (SPSS Version 27.0; IBM Corp). Descriptive statistics included mean ± SD, median and range. ADC, AD, RD, FA and ROI size values were averaged on the three localisations (cranial, mid and caudal renal parenchyma). Comparisons were made between the cortex, medulla and entire parenchyma, between the left and right kidneys, and between the small and large ROI methods.

Non-parametric tests were applied. The Mann–Whitney U-test was applied for two-group comparisons and Kruskal–Wallis test for multiple comparisons.

Statistical significance was set at P <0.05.

Results

A total of 18 cats underwent MRI examinations with DTI sequences. One DTI sequence was not available for review, so 17 cats were enrolled in the study. Of these, nine were female and eight were male, with a median body weight of 4.1 kg (range 3.5–5.3) and a median age of 67 months (range 13–111).

MRI examination

The mean acquisition time for the DTI sequence was 6.8 ± 0.7 mins.

MRI scans were subjectively assessed as good quality, with mild motion blurring in the dorsal T2 sequence in three cats. Morphological MR images were considered normal in all the cats.

No visible artefact was present on the DTI images; however, a close relationship between the ventral border of the left kidneys and the colon was often present, causing mild blurring of the edges.

Before measuring, DWI and morphological images were co-registered in overlay and underlay, and a mask was optimised to reduce noise outside of the kidneys.

Across the different maps, the cortex had higher signal intensity than the medulla in the diffusion images, ADC maps and radial diffusivity maps. The medulla had higher signal intensity in the axial diffusivity and FA grayscale maps. The map with the optically less well-defined differentiation between cortex and medulla was the AD map. On tractography, a consistent, well-organised radial orientation of tracts from the hilar region to the periphery was visible throughout the entire kidney. Tractography can be displayed in three-dimensional images and within a selected ROI, indicating the predilection direction of water molecule movement (Figure 3).

Figure 3

Three-dimensional tractography reconstructed from one region of interest drawn in the medulla of the right kidney shows a regular arrangement of the lines with a predominant radial orientation from right to left

In all cats, using both the small and large ROI methods, no statistically significant difference was observed between the left and right kidneys. Accordingly, the values for the left and right kidneys were averaged for further analysis.

Quantitative results (ROI size, ADC, FA, RD and AD) are summarised in Table 1.

Table 1

Region of interest (ROI) size, apparent diffusion coefficient (ADC), fractional anisotropy (FA), radial diffusivity (RD) and axial diffusivity (AD) for the different analysed ROI

	Cortex	Medulla	P values
Small ROI size (mm2)	12.1 ± 1.39	12.1 ± 1.39	–
ADC (10–3mm2/s)	1.56 ± 0.16	1.42 ± 0.13	<0.001
FA	0.25 ± 0.12	0.49 ± 0.15	<0.001
RD (10–3mm2/s)	1.34 ± 0.12	1.14 ± 0.74	<0.001
AD (10–3mm2/s)	2.09 ± 0.12	2.11 ± 0.20	0.022
Large ROI size (mm2)	130.46 ± 27.61	186.18 ± 37.99	–
ADC (10–3mm2/s)	1.55 ± 0.9	1.50 ± 0.11	0.038
FA	0.24 ± 0.09	0.35 ± 0.10	<0.001
RD (10–3mm2/s)	1.32 ± 0.08	1.27 ± 0.09	0.02
AD (10–3mm2/s)	2.00 ± 0.15	2.09 ± 0.13	0.006
Entire kidney ROI size (mm2)	513.96 ± 93.58	–	–
ADC (10–3mm2/s)	1.56 ± 0.74	–	–
FA	0.32 ± 0.93	–	–

Data are mean ± SD

No statistically significant differences between the large vs small ROI methods were observed for any parameters, with the exception of RD in the medulla. In this region, mean RD measured using the small ROI was 1.14 ± 0.74 and using the large ROI was 1.27 ± 0.09 (P <0.001).

All parameters (ADC, FA, RD and AD) differed significantly between the cortex and the medulla using both sampling methods. FA and AD were higher in the medulla, whereas ADC and RD were higher in the cortex (Figure 4).

Figure 4

Box plot of mean fractional anisotropy (FA) (in blue), apparent diffusion coefficient (ADC) (in red), radial diffusivity (RD) (in green) and axial diffusivity (AD) (in purple) in the cortex and medulla. Values of the right and left kidneys and small and large regions of interest are averaged. For each plot, the box represents the 25th–75th percentiles, and the dark line represents the median. Whiskers represent the highest case within 1.5 times the interquartile range (IQR) and the lowest case within 1.5 times the IQR. Circles represent the outliers, stars extreme outliers

Discussion

This is the first study to apply DTI to the feline kidney and show its feasibility by providing reference data for DTI parameters in healthy feline kidneys. Historically, DTI has been applied to the study of the central nervous system; however, its application to the kidneys has become a growing field of interest in human medicine. ¹⁸

The ADC from DWI sequences quantifies the average diffusivity of water molecules averaged in all three directions, ¹⁰ with no information about possible preferential direction of the water movement. Conversely, DTI depicts the anisotropic properties of the tissues by appreciating the preferential water diffusion direction, quantitatively represented by FA. ⁵ FA indicates whether water molecules are free to diffuse equally in all directions (isotropic diffusion, with FA = 0 if the tensor is spherical) or whether water molecule movement is restricted in some directions (anisotropic diffusion, with FA = 1 corresponding to diffusion only along one orientation ¹⁰ in a cylindrical model). DTI is particularly suitable for kidney evaluation because of the anatomical structure, consisting of tubules orientated in a radial fashion and resulting in strongly anisotropic diffusion. ¹⁹ DTI has never been applied to feline patients, despite the high incidence of chronic renal pathologies in this species, the need for a means of early diagnosis and the low specificity of routinely performed ultrasonographic examinations.

In human medicine, the use of DTI has been described. In normal kidneys, ADC values of approximately 2.4 and 2 (10^–3mm²/s) and FA values of approximately 0.28 and 0.4 have been reported for the cortex and medulla, respectively. ¹⁶ Similar, but lower, values were reported in another study’s control group, with ADC values of 2.08 and 1.99 (10^–3mm²/s) and FA values of 0.19 and 0.22 in the cortex and medulla, respectively. ⁶ In the present study, ADC values were lower than those reported in humans, whereas FA values were similar. Beyond the reported values, which are strongly dependent on several factors (including the MRI system used, b values and number of diffusion directions^19,20), FA in the medulla is consistently higher than in the cortex.^21,22 The radial organisation of the tubules and collecting ducts, as well as the radially orientated vessels draining into the renal pelvis, is the most widely accepted explanation for the high medullary anisotropy. ²³ RD has been reported to be higher in the cortex of healthy human kidneys,^12,24 whereas AD is more variable, having been reported as similar in the cortex and medulla, ²⁴ lower ²⁵ or higher in the cortex. ¹² The present study supports the similar architecture between healthy feline and human kidneys, with FA and AD significantly higher in the medulla than in the cortex, and ADC and RD significantly higher in the cortex.

These values in normal kidneys are important in light of the variations observed in renal pathologies. In human medicine, medullary FA decreases in patients with diabetic nephropathy who still have a normal estimated glomerular filtration rate, ²⁶ and in sickle cell kidney disease. ²⁷ FA also decreases in a variety of renal diseases ²⁸ (including detection of allograft dysfunction, ²⁹ Bardet–Biedl syndrome ²⁵ and chronic renal disease of different aetiology ²⁸ ). Medullary FA is the most sensitive parameter for detecting renal damage in different diseases, followed by ADC. In diabetic nephropathy, medullary FA correlates with estimated glomerular filtration rate ²⁶ and is a reliable indicator of disease progression. ³⁰ In a mouse model, medullary FA and AD strongly correlated with histological fibrosis score. ³¹

The high sensitivity of FA and the reduction in anisotropy are likely due to structural parenchymal alterations, such as interstitial fibrosis, tubular atrophy and cellular infiltration, especially involving the renal medulla ²⁸ in CKD. It has been suggested that most chronic parenchymal diseases alter directed diffusion, represented by FA, before free diffusion, represented by ADC and mainly influenced by perfusion. ²⁹ In the human kidney, the renal interstitium is more abundant in the medulla than in the cortex, ³² meaning that parenchymal changes affecting the interstitium can cause an early and measurable change in medullary FA. If the same pathophysiology is demonstrated in feline patients, this sequence may offer the same advantage of early detection of microstructural abnormalities. Moreover, it could represent a valid monitoring method, capable of detecting subtle disease progression.

Abdominal MRI studies are not yet widely used in veterinary medicine, likely for a variety of reasons, including high costs, availability, indications and the need for general anaesthesia. Only one previous study has investigated renal ADC from DWI scans in healthy cats. ³³ ADC was higher in the cortex than in the medulla, based on a monoexponential model derived from DW images. ADC values in the present study behaved differently; however, they were obtained from DTI sequences and were based on a different algorithm, making direct comparison impossible. The variability of ADC values and the inconsistent finding of higher ADC values in the cortex than in the medulla from DW imaging ³⁴ highlight the limitations in interpretating the ADC values derived from a monoexponetial model in the kidney. ADC and FA values obtained from DTI sequences may overcome these limitations.

The substantial agreement between the two tested sample strategies supports the use of small, single ROIs as an appropriate and less time-consuming alternative to the large ROI method. This is expected in normal organs, where the architecture is assumed to be homogeneous. In chronic nephropathy, the entire organ is usually affected, but it is unknown whether the pathology could be unevenly distributed, potentially affecting FA or other parameters differently across regions. The only difference observed between the large and small ROI methods was in medullary RD. Considering the agreement of the other parameters, this difference is more likely due to sampling inaccuracy than a true difference in tissue architecture. Drawing the large ROI freehand may inadvertently include boundaries such as the renal crest. Notably, even in human medicine, no standardised sampling technique has been recommended.

The acquisition of the DTI sequence has reasonable duration – less than 7 mins – and its clinical use can be hypothesised. The shallow respiration of cats and their dorsal recumbency, combined with the use of a surface coil positioned close to the ROI, produced images of good quality with only sporadic motion artefacts. Another noticeable advantage of DTI, compared with other MRI-based sequences like perfusion imaging, is that it avoids the need for contrast medium. In human patients, administration of gadolinium-based contrast can lead to systemic fibrosis. ³⁵ Although a similar pathology has not been described in cats, it could represent a risk factor in some patients. Clinical application would also benefit from analysing the two kidneys separately, allowing precise assessment of the right and left kidneys.

Tractography can illustrate the typically radial tract orientation, reflecting the radial organisation of the anatomical structures, but it lacks quantification. It is unknown whether tractography can detect abnormalities in fibre number or orientation, not only in advanced stages of nephropathy but also at early stages.

The present study has several limitations. It included only healthy cats and had a relatively small sample size. Biopsies were not performed, so subclinical pathology cannot be excluded. Clinical patients were not included; therefore, the diagnostic value of the tested sequence has yet to be established. The study nonetheless hypothesises a potential clinical use; for this reason, other factors that could potentially influence the results – such as signal:noise ratio in each examination, the exact temperature and hydration status of the animals (both maintained within physiological ranges) or the effects of different anaesthetic drugs – were not further investigated.

Imaging parameters (b values, number of encoding directions) were adapted from human protocols and may require further optimisations for feline patients as they are known to influence image quality and quantitative values.^19,20

Inter- and intrareader reproducibility was not assessed and a single operator drawing the ROIs was chosen for consistency.

Conclusions

DTI is feasible in cats and substantial similarities in feline and human healthy kidneys have been proved.

Despite current limitations (high costs, equipment requirements and the need for general anaesthesia), DTI has the potential to improve feline nephrology diagnostics and management.