a b s t r a c t

Objective: To evaluate the performance of non-contrast computed tomography (CT) by reporting the difference in attenuation between normal and inflamed renal parenchyma in patients clinically diagnosed with acute pyelo- nephritis (APN).

Material and methods: This is a retrospective study concerned with non-contrast CT evaluation of 74 patients, ad- mitted with a clinical diagnosis of APN and failed to respond to 48 h antibiotics treatment. Mean attenuation values in Hounsfield units (HU) were measured in the upper, middle and lower segments of the inflamed and the normal kidney of the same patient. Independent t-test was performed for statistical analysis. Image evalua- tion included receiver operating characteristic (ROC), visual grading characteristic (VGC) and kappa analyses. Results: The mean attenuation in the upper, middle and lower segments of the inflamed renal cortex was 32%, 25%, and 29% lower than the mean attenuation of the corresponding cortical segments of the contralateral normal kidney, respectively (p b 0.01). The mean attenuation in the upper, middle, and lower segments of the inflamed renal medulla was 48%, 21%, and 30%, lower than the mean attenuation of the corresponding medullary segments of the contralateral normal kidney (p b 0.02). The mean attenuation between the inflamed and non-inflamed renal cortex and medulla was 29% and 30% lower respectively (p b 0.001). The AUCROC (p b 0.001) analysis dem- onstrated significantly higher scores for pathology detection, irrespective of Image quality, compared to clinical and laboratory results with an increased inter-reader agreement from poor to substantial.

Conclusion: Non-contrast CT showed a significant decrease in the parenchymal density of the kidney affected with APN in comparison to the contralateral normal kidney of the same patient. This can be incorporated in the diagnostic criteria of APN in NCCT in the emergency setting.

(C) 2017

Introduction

acute pyelonephritis APN is an infection of the renal parenchyma and pelvis [1]. APN can be caused by an ascending bacterial or fungal in- fection of the lower urinary tract or less commonly by hematogenous spread [2]. Current diagnosis is based on the classic clinical signs and symptoms of costovertebral angle tenderness, high-grade fever, nausea, vomiting, and dysuria, along with the laboratory signs of urinary tract infection pyuria, positive urine cultures, WBC cast [1,3]. Females,

* Corresponding author.

E-mail addresses: [email protected] (F. El-Merhi), [email protected] (M. Mohamad), [email protected] (A. Haydar), [email protected] (L. Naffaa), [email protected] (R. Nasr), [email protected] (I.A.-S. Deeb), [email protected] (N. Hamieh), [email protected] (Z. Tayara), [email protected] (C. Saade).

¹ Co-1st author.

immunocompromised, immunosuppressed, patients with recurrent urinary tract infections, and those diagnosed with vesicoureteral reflux, are more prone to develop APN [4]. APN could be a life threatening con- dition when diagnosis and treatment were delayed. Such a delay may lead to increased risk of renal complications, sepsis, and death. Radio- logical imaging is not a mandatory requirement for every patient with a clinical diagnosis of APN since it is usually saved for high-risk patients (Diabetics, Immunosuppressed, elderly), patients presenting with atyp- ical severe symptoms, and those with persisting symptoms despite 72 h of antibiotic treatment [5]. Moreover, imaging is crucial in cases with complicated pyelonephritis, and in those with a suspicion of ureteric stones or Renal infarction. Pyelonephritis is considered as complicated when the infection progresses into an abscess (nephric more common than perinephric), or leads to renal papillary necrosis, or forms emphy- sematous pyelonephritis [6]. Patients are suspected to have complicated pyelonephritis in cases whom their symptoms show no improvement

https://doi.org/10.1016/j.ajem.2017.09.026

0735-6757/(C) 2017

on antibiotic therapy or who display recurrent symptoms few weeks after treatment. Different radiological modalities play a role in the eval- uation of urinary tract infection. However, CT outperforms both renal Ultrasound and intravenous pyelogram and is considered as the key- stone in the assessment and diagnosis of pyelonephritis [7]. Renal Ultra- sound was considered as the primary imaging method used in pyelonephritis. Nonetheless, it has been proved to be a non-specific im- aging modality due to its limited ability in detecting the parenchymal le- sion in patients with pyelonephritis [8] . In a study done by Stunell et Al, the renal US sensitivity and specificity in APN was 74.3% and 56.7%, re- spectively, in comparison to a sensitivity and specificity of 86.8% and 87.5% respectively in CT [6]. Hence, CT is emerging as the leading radio- logic modality in evaluating patients with APN.

Despite the radiation hazard of CT, CT is considered as superior to ul- trasound in emergency settings as it is a faster tool, non-operator de- pendent, and not limited by increased body habitus, relatively. Though MRI is considered as another reliable imaging tool in pyelonephritis, but CT offers lower expenses [9] . CT scan of the kidneys helps in charac- terizing the nature of the detected mass, offers a better delineation of the extent and severity of the disease, and allows for the detection of renal parenchymal and collecting system abnormalities [1,7,9].With its ability to measure attenuation levels within and around organs it can serve to be a strong quantitative tool in the assessment of APN.

In a study done by Mellit et al., non-contrast abdominal CT in the emergency department was proven to yield a more precise diagnosis and a better management of acute abdominal pain in elderly in cases of Bowel obstruction (mechanical or by fecal impaction, pyelonephritis, and Biliary tract diseases) [10] Agarwal et al. conducted a retrospective analysis of 322 emergency department patients presenting with flank pain undergoing CT for a clinical diagnosis of nephroureterolithiasis [11]. The study found that while Contrast enhanced CT provided better delineation of an abnormality, it is unlikely to reveal additional acute findings not seen on Non-enhanced CT [11]. The study also revealed that use of Contrast enhanced CT after Non-enhanced CT resulted in Changes in management in only 2% of patients [11]. In another study conducted by Coursey et al. non-contrast CT was directly compared with IV urography in four series [12]. Non-contrast CT was found to be equally reliable for diagnosing flank pain due to causes other than ureterolithiasis, more rapid and safer since it uses no contrast media [12]. The study concluded that when non-contrast CT is available, it is the best first study in the nonpregnant adult presenting with flank pain likely to be due to stone disease [12]. These studies suggest that unenhanced CT alone may indeed be sufficient in the initial evaluation of patients coming to the emergency department with acute flank pain.

Despite the fact that non-contrast CT is considered as a prominent imaging tool for detecting the previously mentioned complications (perinephric fluid accumulation, inflammatory lesions, and emphyse- matous pyelonephritis or pyelitis) in addition to renal hemorrhage, renal calculi, and urinary obstruction, the clinical use of unenhanced CT has not been fully explored yet in patients with APN [13-16]. Using CT without contrast saves time in emergency settings, which could make it the ultimate diagnostic tool of pyelonephritis in the emergency department [17].

The aim of our study is to evaluate the performance of non-contrast CT by comparing the density measurements (in HU) of a normal kidney to that of the contralateral inflamed kidney of the same patient in con- junction with differentiation between upper, middle, and lower cortical and medullary segments.

Materials and methods

Study population

This study was approved by the institutional review board and a waiver of consent was granted for retrospective chart review. The study included 74 consecutive patients between November 2012 and

January 2015. Patients were diagnosed with APN and showed no im- provement after 48 h of antibiotics treatment (mean age 43.81 +- 16.56 years, 18 males, 56 females). We diagnosed patients with APN based on their clinical symptoms of fever (temperature N 38.3 ?C), nau- sea/vomiting, costovertebral angle tenderness, in addition to leukocyto- sis, pyuria ((N 20 WBCS/HPF), and positive urine culture (Table 1). We ruled out ureters stone obstruction by NCCT. This study excludes pa- tients diagnosed with bilateral APN.

Abdominal CT acquisition

Patients were instructed to lie flat in a supine position with arms above the head in a 256-slice MDCT scanner (Philips iCT, Philips Healthcare, The Netherlands). Anterior-posterior scout scan was done before scan acquirement. CT scan included the region from the liver to the symphysis pubis and the patients were asked to hold their breath with their tongue against their palate. The scanning parameters employed were 120 kVp, 120 mA with mA modulation, 0.4 s rotation time, pitch 0.984 mm/rot, and a craniocaudal scan. The image recon- struction parameters utilized iDose⁴ iterative reconstruction (level 4) with 256 x 0.625 mm slice thickness and reconstruction intervals of 0.5 mm, field of view 350 x 350 mm and a window width and win- dow level of 420 and 65 respectively.

Clinical parameters

We investigated age, sex, urinalysis, and urine culture. Positive urine cultures included E. coli, Enterobacter aerogenes, Enterococcus, and K. Pneumonia (Table 1). We classified Pyuria into 6 grades according to the number of WBC/HPF by HPF microscopic examination (grade 1: 0-1 WBC/HPF; grade 2: 1-4 WBC/HPF; grade 3: 5-9 WBC/HPF; grade

4: 10-29 WBC/HPF; grade 5: 30-50 WBC/HPF; grade 6: more than half the HPF filled with WBC).

Radiological assessment

Patients’ medical records and CT scans were collected by two radiol- ogists who reviewed the non-contrast (CT findings by consensus). We evaluated the CT scans using the following patterns: focal or global renal enlargement based on the difference in size between the right and left kidney, stranding of perinephric fat, thickening of Gerota’s fas- cia, obliteration of the renal sinus, pelvicalyceal wall thickening, urinary tract stones, pelvicalyceal air, abscess (well-defined hypo-attenuated mass with irregular wall), and hydronephrosis (Table 2).

Quantitative analysis

Mean renal cortex and medulla in HU values were obtained from axial images using a circular ROI with the greatest diameter that would fit within the area of interest. ROI measurements were per- formed at three anatomical segments in the kidney. Each measurement

Table 1

Clinical presentation and urinalysis in patients with acute pyelonephritis.

n (%)

Clinical symptoms

Fever 54 (73)

Pain 59 (80)

Leukocytosis 44 (59)

Urinalysis culture

E-coli 60 (81)

Enterobacter aerogenes 3 (b1)

K. Pneumonia 9 (12)

Normal 2 (b1)

Note: (%) is percentage.

Table 2

Radiological criteria in patients with acute pyelonephritis

Radiological appearance n (%)

Kidney enlargement 33 (45)

Perinephric stranding 69 (93)

Hydronephrosis 45 (60)

Presence of stones 30 (41)

Pelvicalyceal wall thickening 38 (51)

Pelvicalyceal air 0 (0)

Abscess 0 (0)

Note: (%) is percentage.

was taken in the coronal plane orthogonal cross-section from the renal cortex and medulla in Fig. 1:

Statistical analysis

We compared the measured attenuation values between the renal segments using the Wilcoxon Signed Rank Test and Cohen’s kappa of concordance. We considered the results as statistically significant ac- cording to a p-value <=0.05.

Radiation dose measurement

For each of the CT scans, individual effective dose (E_ff [mSv]) was cal- culated from the dose-length products (DLP [mGy x cm]), which were recorded from the patient protocol [10]. A normalized conversion factor (k [mSv/mGy x cm]) for the abdomen-0.015 mSv/mGy x cm-was used to calculate the E_ff [13]: E_ff = DLP x k.

Qualitative analysis

The qualitative analysis included a random sample of 50 cases (25 with pyelonephritis and 25 normal cases). Out of the total of 74 patients with pyelonephritis, we have randomly chosen 25 patients to be pre- sented to the observers. While the normal cases were randomly chosen from the radiology directory with no clinical symptoms or radiological evidence of pyelonephritis.

Trans-axial images showed pyelonephritis pathology. The clinical laboratory results of the recruited patients were concealed. To ensure that all images displayed an acceptable level of quality, two expert body radiologists not included in the study proper employed radiologic criteria to guarantee a correct scan range and anatomical inclusion.

Two abdominal radiologists (certified by the Royal College of Radiol- ogists and the American Board of Radiology respectively with a mean of

6.5 years) assessed the images. All reviewers were specialists in body imaging and each observer was allowed to manipulate the window and level of the images. Each image in the first sitting was presented and a score of 1-5 was assigned (5: pyelonephritis pathology was defi- nitely present and 1: pathology definitely not present).

We assessed the relative performance of the non-contrast imaging sequences using Visual Grading Characteristics (VGC) analysis [20]. VGC curve, the result of the nonparametric VGC analysis, measures the image quality and evaluates the performance of pathology detection. We classified the pathology presented by the images according to a 5-point-scale (5: pyelonephritis pathology is definitely present and 1: pyelonephritis pathology is definitely not present) (Table 3). We plot- ted the VGC curve and calculated the p-value and 95% confidence inter- val. If the area under the resulting VGC curve is 0.5 (diagonal), the two sequences were rated equally good, and if the area is larger than 0.5 (above the diagonal), image quality performed better than pathology detection.

We assessed the Interrater agreement using k statistics for the 2 raters, and the interrater variability by calculating the mean paired k values of both raters. We defined k as poor k b (0.00), slight (k = 0.00-0.20), fair (k = 0.21-0.40), moderate (k = 0.41-0.60), substantial (k = 0.61-0.80), and almost perfect (k = 0.81-1.00).

Results

Patient demographics

A total of 18 males and 56 females with a mean age of 43.81 +-

16.56 years were included in this study. The right kidney was affected more frequently than the left: the ratio between the right to the left kid- ney was 63/11 (5.72).

Fig. 1. Anatomical segments of the human kidney. Image (a) demonstrates a diagram of the kidney which is divided into upper (U), middle (M), and lower (L) segments. Image (b) shows a coronal reconstruction of the kidney segments with the measurements taken in the cortex (white dot) and medulla (black dot) during non-contrast abdominal CT.

Table 3

Five-point classification scale

	Perinephric space	Renal segments	Renal cortex	Renal Medulla	Pyeloureteric junction
1	1
2	1	1
3	1	1	1
4	1	1	1	1
5	1	1	1	1	1

Renal parenchyma attenuation measurements

The mean attenuation of the inflamed renal cortical segments (upper:32%, middle:25%, and lower:29%) was lower than that of the normal renal cortex in the same patient (p b 0.01). The attenuation of the renal medullary parenchymal segments was significantly lower in the inflamed kidney (Upper segment: 48%, Middle segment: 21%, and the lower segment: 30%) in the same patient (p = 0.02). The mean at- tenuation between the inflamed and non-inflamed renal cortex and medulla was 29% and 30% lower respectively (p = 0.001) (Table 4).

Radiation dose

The mean radiation dose for all patients was significantly lower than (3.07 +- 0.68 mSv) when compared to literature, with low Dose range from 1 to 2 mSv [15], while standard abdominopelvic CT ranged from

5.65 +- 3.45-8.40 +- 4.49 mSv [16].

Receiver operating characteristic

The six-point scale demonstrated a significant difference (p b 0.01) between normal and inflamed kidney with mean ROC values demon- strating increased reader confidence between each reader being 0.71 and 0.95 respectively (Fig. 2a and b).

Visual grading characteristic

The five-point scores individually graded by the two readers for image quality versus pathology detection protocol were expressed as a graph shown in Fig. 3. The area under the curve (AUC_VGC) expressing the ratio between image quality (x-axis) and pathology detection (y- axis) in patients with APN. Reader A (Fig. 3a) was 0.344 and reader B (Fig. 3b) was 0.371. This implies that both readers AUC_VGC demonstrat- ed that image quality had no significant effect on reader confidence when compared to ROC (p N 0.05). The straight diagonal line spanning the middle of the graph indicates an AUC of 0.5, the test effectiveness threshold.

Table 4

Mean cortex and mean medulla among inflamed kidney versus normal kidney.

Inflamed Normal p-Value

Fig. 2. ROC-curve obtained from the data. The boxes are the coordinates corresponding to the observer’s interpretation of the scale steps of the rating scale. The “Area under the curve” (AUG_ROC) in each reader (a and b) demonstrated a significant increase in reader confidence in pathology detection irrespective of image quality.

Reader agreement

We would expect the two radiologists to agree on 34.28% of the pa- tients assuming that each one of them had made his determination ran- domly. In fact, they agreed on 68% of the patients or 51.31% of the way between random agreement and perfect agreement. Each reader dem- onstrated substantial inter and intra-rater agreement (Reader A: inter 0.69, intra 0.72; Reader B: inter 0.73, intra 0.77) (p b 0.001).

Discussion

Due to the increased number of contrast-enhanced CT scans of the abdomen and the subsequent increase in the rate of extravasation

Cortex Upper	18.81 +- 6.42	27.72 +- 6.4	0.001	[21], iodinated contrast media allergy [22] and the risk of contrast- induced DNA damage [23], clinicians seem to prefer the use of non-
Middle	21.07 +- 6.23	28.07 +- 6.23	0.001	enhanced abdominal CT over enhanced scans for urologic related symp-
Lower	20.08 +- 6.04	28.08 +- 6.04	0.001	toms in the emergency setting [1,7].. Non-contrast CT scan has the abil-
Total Medulla Upper Middle	59.96 +- 18.39 10.86 +- 6.48 15.82 +- 6.09	83.96 +- 18.36 19.72 +- 6.4 19.07 +- 6.23	0.001 0.001 0.02	ity to detect APN complications (perinephric fluid accumulation, inflammatory lesions, and Emphysematous pyelonephritis or pyelitis, except for renal abscess which needs contrast enhanced CT), renal cal-
Lower	14.73 +- 5.89	20.61 +- 6.39	0.001	culi, and urinary obstruction. Hence, non-contrast CT is emerging as a
Total	41.42 +- 17.32	59.39 +- 18.12	0.001	promising diagnostic tool of pyelonephritis in the emergency

Note: (+-) is standard deviation.

department.

Fig. 3. VGC-curve obtained from the data. The boxes are the coordinates corresponding to the observer’s interpretation of the scale steps of the rating scale. There was no statistical significance when calculating the AUG_VGC for readerA (0.344)-Fig. 2a and reader B (0.371) Fig. 2b which demonstrated favor towards pathology detection over image quality (p N 0.05).

The results of our study highlight that direCT attenuation values can be compared between normal and pyelonephritic kidney, resulting in direct comparison as an indicator of APN.

The increased radiation exposure with CT scans has been linked with an increased risk of cancer [24]. Enhanced renal CT scans employing a four-phase protocol (pre-contrast, corticomedullary, nephrographic and excretory phases) have an effective radiation dose ranging between 25 mSv and 35 mSv [25], and studies with single CT protocols have their effective radiation dose range from 10-15 mSv [26]. On another hand, our study demonstrated, at a significantly lower radiation dose, a com- bination of quantitative and qualitative radiological analysis is a strong indicator in patients with APN.

To our knowledge, this is the first study that employed both quanti- tative renal parenchymal measurements (in each kidney segment) coupled with qualitative image interpretation analysis (including confi- dence intervals, image quality, and radiologic signs). Interestingly, the inflamed renal cortex and medulla had a significantly lower mean at- tenuation compared to the normal renal cortex and medulla by up to 32% (p b 0.001) at lower radiation dose levels.

Confirmation of the qualitative data was supplied by the statistical analysis that led to the creation of the ROC and VGC curves. The area under the ROC curve of each reader values of patients with APN was (0.82 and 0.85) showed that when comparing APN detection with clin- ical laboratory results, each reader had N 82% accuracy in detecting APN in non-enhanced abdominal CT scans. These results confirmed the

reliability of non-enhanced qualitative image interpretation in identify- ing inflammatory kidney diseases.

There were shortcomings in our study. First, this study is a retrospec- tive review of a limited number of examinations. Second, only two radi- ologists interpreted the images, limiting the reproducibility of the results. Larger scale studies are needed to confirm these results and to validate the accuracy with contrast-enhanced abdominal CT and mag- netic resonance imaging.

Conclusion

In summary, both Qualitative and quantitative radiological analyses are strong indicators in patients with acute pyelonephritis during non- contrast abdominal CT.

References

1. Stunell H, Buckley O, Feeney J, Geoghegan T, Browne RF, Torreggiani WC. Imaging of acute pyelonephritis in the adult. Eur Radiol 2007;17(7):1820-8.
2. Craig WD, Wagner BJ, Travis MD. Pyelonephritis: radiologic-pathologic review. Ra- diographics 2008;28(1):255-77 [quiz 327-8].
3. Kim SH, Kim YW, Lee HJ. Serious acute pyelonephritis: a predictive score for evalu- ation of deterioration of treatment based on clinical and radiologic findings using CT. Acta Radiol 2012;53(2):233-8.
4. Alshamari M, Norrman E, Geijer M, Jansson K, Geijer H. Diagnostic accuracy of low- dose CT compared with abdominal radiography in non-traumatic acute abdominal pain: prospective study and systematic review. Eur Radiol 2016;26(6):1766-74.
5. Wie SH, Ki M, Kim J, Cho YK, Lim SK, Lee JS, et al. Clinical characteristics predicting early clinical failure after 72 h of antibiotic treatment in women with community- onset acute pyelonephritis: a prospective multicentre study. Clin Microbiol Infect 2014;20(10):O721-9.
6. Shields J, Maxwell AP. Acute pyelonephritis can have serious complications. Practi- tioner 2010;254(1728):19 [21, 3-4, 2].
7. Demertzis J, Menias CO. State of the art: imaging of renal infections. Emerg Radiol 2007;14(1):13-22.
8. Piccoli GB, Picciotto G, Rossetti M, Burdese M, Consiglio V, Magnano A, et al. Imaging data suggesting acute pyelonephritis in the kidney graft: report of five cases with Atypical clinical presentation. Int J Antimicrob Agents 2006;28(Suppl. 1):S64-71.
9. Kawashima A, Sandler CM, Goldman SM. Imaging in acute renal infection. BJU Int 2000;86(Suppl. 1):70-9.
10. Millet I, Sebbane M, Molinari N, Pages-Bouic E, Curros-Doyon F, Riou B, et al. System- atic unenhanced CT for acute abdominal symptoms in the elderly patients improves both emergency department diagnosis and prompt clinical management. Eur Radiol 2017;27(2):868-77.
11. Taniguchi LS, et al. Are the unenhanced and excretory CT phases necessary for the evaluation of acute pyelonephritis? Acta Radiol 2017;58(5):634-40.
12. Agarwal MD, et al. Limited added utility of performing follow-up contrast-enhanced CT in patients undergoing initial non-enhanced CT for evaluation of flank pain in the emergency department. Emerg Radiol 2015;22(2):109-15.
13. Talner LB, Davidson AJ, Lebowitz RL, Dalla Palma L, Goldman SM. Acute pyelonephri- tis: can we agree on terminology? Radiology 1994;192(2):297-305.
14. Huang JJ, Sung JM, Chen KW, Ruaan MK, Shu GH, Chuang YC. Acute bacterial nephri- tis: a clinicoradiologic correlation based on computed tomography. Am J Med 1992; 93(3):289-98.
15. Goldman SM. Acute and chronic urinary infection: present concepts and controver- sies. Urol Radiol 1988;10(1):17-24.
16. Thornbury JR. Acute renal infections. Urol Radiol 1991;12(4):209-13.
17. Singh A. Emergency Radiology: Imaging of Acute Pathologies. New York: Springer; 2013.
18. Bath M, Mansson LG. Visual grading characteristics (VGC) analysis: a non- parametric rank-invariant statistical method for image quality evaluation. Br J Radiol 2007;80(951):169-76.
19. Taylor S, Van Muylem A, Howarth N, Gevenois PA, Tack D. CT dose survey in adults: what sample size for what precision? Eur Radiol 2017;27(1):365-73.
20. Wang CL, Cohan RH, Ellis JH, Adusumilli S, Dunnick NR. Frequency, management, and outcome of extravasation of nonionic iodinated contrast medium in 69,657 in- travenous injections. Radiology 2007;243(1):80-7.
21. Yoon SH, Lee SY, Kang HR, Kim JY, Hahn S, Park CM, et al. Skin tests in patients with hypersensitivity reaction to iodinated contrast media: a meta-analysis. Allergy 2015; 70(6):625-37.
22. Piechowiak EI, Peter JF, Kleb B, Klose KJ, Heverhagen JT. Intravenous iodinated con- trast agents amplify DNA radiation damage at CT. Radiology 2015;275(3):692-7.
23. Levesque VM, Shyn PB, Tuncali K, Tatli S, Nawfel RD, Olubiyi O, et al. Radiation dose during CT-guided percutaneous cryoablation of renal tumors: effect of a dose reduc- tion protocol. Eur J Radiol 2015;84(11):2218-21.
24. Noroozian M, Cohan RH, Caoili EM, Cowan NC, Ellis JH. Multislice CT urography: state of the art. Br J Radiol 2004;77(Spec 1):S74-86.