Reproducibility of 3 mm-Slice-Thick Reconstruction of Paranasal Sinus Computed Tomography Scans

Background: After the failure of medical treatment, the surgery of chronic rhinosinusitis (CRS) is planned according to endoscopic and paranasal sinus computed tomography (CT) findings. Objective: The aim of this prospective study was to evaluate whether this study method might be eligible in studies aiming at radiation dose reduction. Sinus CT scans were chosen as a model because of the high variation of the radiological anatomy of surgically important sinonasal structures. We hypothesized that 3 mm-slice-thick reconstruction CT had poor reproducibility. Methods: 59 CRS patients underwent routine multi-detector sinus CT (CTMD). CT3mm was reconstructed from CTMD data-sets. Lund-Mackay (LM) scores and 43 other structural parameters were analyzed blinded. Agreement was studied between CTMD and CT3mm (intra-observer reproducibility), and between three observers (inter-observer reproducibility) by using Cohen’s kappa. Results: The inter-observer agreement was moderate (kappa 0.4 0.6, p < 0.01) in the majority of structures of CT3mm scans. The intra-observer reproducibility of CT3mm scans was very good in most structures, however, it was poor in important structures such as frontal and spheno-ethmoid recess, lamina paMatti Karjalainen and Anna Julkunen have equally contributed to this work. Jura Numminen and Sanna Toppila-Salmi are shared last authors. M. Karjalainen et al. 40 pyracae, and location of optic nerve or anterior ethmoidal artery. The grade of surgeon’s confidence of CT3mm in comparison to CTMD was lower (kappa 0.2 0.4, P < 0.05). Conclusion: This methodology might have some use in studies aiming at radiation dose reduction. As was expected, 3 mm-slice-thick reconstruction CT had poor reproducibility and surgeon’s confidence. More recent methods such as cone beam computed tomography scans have nowadays more relevant dose reduction potential.


Introduction
Chronic rhinosinusitis (CRS) is a common, multifactorial and variable disease with a prevalence of around 10% -16% [1].The diagnostics of CRS is based on typical symptoms and clinical findings [1].Computed tomography (CT) scans are the imaging modality of choice confirming the extent of pathology and the need for surgery.The main findings in CRS are mucosal changes within the osteomeatal complex and/or sinuses.Other characteristic findings are air-fluid levels, mucosal thickening and opacification of the normally aerated sinus lumen.The only CRS indicating change can be sclerotic, thickened bone of the sinus wall [1].The number of anatomical variants is very high in paranasal sinuses [2] [3].Several of them are located close to sinonasal surgical area and thus the intraoperative lesions of them may lead to severe illness or be life threatening.The operatively important structures are insertion of middle turbinate and the uncinated process, the location of anterior ethmoidal artery, and Keros class, Infraorbital cells, position of the Agger nasi cell, and the anatomical variants located in the operative area [4]- [6].Studies demonstrate that there is dose-reduction potential of CT scans [7]- [10].
The aim of this prospective study was to evaluate whether our method might be eligible in studies aiming at radiation dose reduction.Sinus CT scans were chosen as a model because of the high variation of the radiological anatomy of surgically important sinonasal structures.We hypothesized that 3 mm-slice-thick reconstruction of sinus CT scans had poor dose-reduction potential and inter-observer agreement.

Patients
This prospective cohort study was carried out in the Department of Otorhinolaryngology at Tampere University Hospital, Finland from 2006 to 2015.The study was approved by the ethical committee of the Pirkanmaa Hospital District (no 96032).Written informed consent was obtained of each participant.A random sample of 59CRS patients, who were evaluated to benefit from sinus CT scans during 2006-2007, was enrolled to this study.Patient data was collected from medical records and by a questionnaire at the time the sinus CT was performed.The follow-up data of sinonasal operations and time of surgery was collected form patient records of Tampere University Hospital or Tampere City Hospital in 2015 in average 9 years after the time of performing sinus CT scans.None of the subjects had used Aspirin desensitization, allergen immunotherapy, or anti IgE therapy prior to or during the sinus CT scans or during the follow-up.

CT Scans
The patients underwent routine sinus multiple detector computed tomography (CT MD ) examinations for clinical purposes.Two different CT MD machines were used: GE Light Speed 16 (GE Healthcare, Milwaukee, Wisconsin) and Philips Brilliance 64 (Philips, Best, the Netherlands).The patients were imaged in supine position with a kilovoltage of 120 kV and a milliampere second of 100 mAs.In the GE machine, the slice thickness was 0.625 mm with coronal reconstructions at 1.5 mm.In the Philips machine, the slice thickness was 0.9 mm with coronal reconstructions at 0.9 mm.Both were three dimensional (3D) in nature without any gap.In all cases, the imag-ing was performed using a bone filter technique.The imaging covered the entire sinonasal area both in the axial and coronal directions.The coronal reformations spanned through the entire area, starting from the anterior wall of the frontal sinuses and ending to the level of the posterior wall of the sphenoid sinuses.We performed 3 mm-slice-thick reconstructions (CT 3mm ) from coronal, sagittal and axial data sets of CT MD .The coronal reformations in the CT MD and CT 3mm spanned through the entire area, starting from the anterior wall of the frontal sinuses and ending at the level of the posterior wall of the sphenoid sinuses.

Evaluation of CT Scans
CT MD and CT 3mm scans were observed by three independent observers blinded to each other and to the patient history data.The observation of the same patient's scans occurred at least one week apart.The focus was to compare agreement between CT MD and CT 3mm scans and between three observers in evaluating CRS-related changes and radiological anatomy of the structures.The observers were an experienced head and neck radiologist (AM), an experienced ENT-and rhinosurgeon (JN), and a fifth year ENT resident (ST-S).They filled in a 49 -item form of sinonasal structures from both scans and from both sides of each patient (Table 3).Similar forms were filled from CT MD and CT 3mm scans.Evaluation of the same patient's scans occurred at least 7 days apart.All asked structures had 2 -5 different choices.Before starting the evaluation of the CT scans, all choices were carefully discussed by the observers.Observers also made a pilot of 15 CT scans in order to make sure that all observers understood how to fill in the forms.The Radiologist did not respond to the questions: "Need for septoplasty" and "Grade of surgeon's confidence based on images".

Data Analysis
Statistical analysis was carried out by the SPSS Base 15.0 Statistical Software Package (SPSS Inc., Chicago, IL, USA).Cohen's kappa was used to compare the degree of agreement of CT MD and CT 3mm scans (e.g.intraobserver agreement); and the inter-observer agreement of CT 3mm scans.The calculation is based on the difference between how much agreement is actually present compared to how much agreement would be expected to be present by chance alone.The established interpretation of Kappa-value is classified into 6 subgroups: Poor ≤ 0.2, Fair 0.21 -0.4,Moderate 0.41 -0.6, Good 0.61 -0.8, and Very Good 0.81 -1.0.A value under zero means that the agreement is worse than by chance, and the value range is from −1 to +1.Two-tailed P-values of <0.05 were considered statistically significant.

Patient Characteristics
The characteristics of patients are shown in Table 1.The mean (min-max) age was 43 (13 -77) years by the time of taking the CT scans.54.2% of patients underwent sinonasal operation within a year after the CT scans were performed (Table 1).Of these patients that were operated at the time of CT scans, 7 (21.9%)underwent revision sinonasal surgery in average 3 years later.47.5% of patients reported suffering from diseases (other than CRS) with regular need of medications.The most frequent diseases were (number of parients): heart and vascular diseases [11], hypothyreosis [8], migraine [3] and arthrosis/arthritis [4].

Inter-Observer Agreement of CT3mm Scans
Lund-MacKay scores (Table 2) and 43 other structures were evaluated by a radiologist, an ENT surgeon and an ENT resident from both CT MD and reconstructed CT 3mm scans (Table 2, Table 3).The inter-observer agreement for CT 3mm scans was moderate (kappa 0.4 -0.6) in the majority of structures (Table 3).In CT 3mm scans, the surgically important structures with greatest disagreement were found in Keros classification (radiologist vs. ENT resident), the location of anterior ethmoidal artery and optic nerve (all observers), prominent ethmoid bulla (ENT resident vs. ENT surgeon), and thickness and contact to middle turbinate of orbital lamina of ethmoidal bone (radiologist vs. ENT surgeon) (Figure 1, Table 3).

Intra-Observer Agreement
We compared the degree of agreement between CT MD and CT 3mm scans of the Lund-MacKay scores and 43 oth-   er structures evaluated by the radiologist.In general, the intra-observer agreement was very good (kappa 0.8 -1.0) in the majority of structures, such as Lund Mackay scores (Table 4).Yet, several surgically important structures did not show adequate agreement: OMC-area, frontal and spheno-ethmoid recess, lamina papyracae, and optic nerve (Table 4).Anterior ethmoidal artery was responded to be undetectable in 95% of the CT 3mm scans, but only in 13% of the CT MD scans (Figure 1).

Operative Confidence of CT3mm Scans
ENT surgeon and ENT resident estimated operative certainty of CT 3mm in comparison to CT MD .ENT surgeon responded that the operative confidence is good in 69.6% of CT MD scans, and 64.3% of CT 3mm scans.The intraobserver agreement in the grade of ENT surgeon's confidence of CT 3mm in comparison to CT MD was moderate (kappa 0.4, P = 0.005).ENT resident responded that the operative confidence is good in 73.2% of CT MD scans, but only in 37.5% of CT 3mm scans.The intra-observer agreement in the grade of ENT resident's confidence of CT 3mm in comparison to CT MD was thus poor (kappa 0.2, P = 0.035).

Discussion
This study was carried out to evaluate whether the used method would be eligible when studying radiation dose reduction potential.As an example, we compared two techniques of paranasal sinus CT scans.From the CT scans we evaluated important structures of CRS diagnostics and operation.By reformatting the CT 3mm images, we were able to evaluate the same patient exactly without additional imagining.Patient selection was performed randomly from a doctor's reception.A heterogeneous patient group made the extrapolation of the results possible for clinical practice We found that inter-observer agreement was only moderate in CT 3mm scans.This finding is in line with our unpublished data that inter-observer agreement is at similar level also in the conventional CT MD scans.This might be due to the fact that sinonasal anatomy is highly variable.Intra-observer agreement between CT 3mm and CT MD scans was very good in most structures.However, it was fair to poor in several surgically important structures.The grade of young surgeons' confidence was relatively good with CT MD scans whereas the grade of confidence was poor with CT 3mm scans.Taking together, CT 3mm scans seemed not to be clinically relevant imaging method.
Previously, it has been shown that CRS diagnostics is possible with other reduced radiation CT-techniques.However, the surgical aspects have not been dealt with in these studies [7]- [9].Our current study showed that it was possible to reduce the radiation dose and increase slice thickness without compromising the excellent bone and soft tissue contrast.However, successful endoscopic sinus surgery (ESS) requires detailed knowledge of the highly variable anatomy of the nasal cavities, the ostiomeatal unit, and the skull base.In addition, the vital structures especially the optic nerve and the anterior ethmoid artery and their relationship to operational areas need to be clear.
Multi-detector CT technology comprises multiple detectors that are equivalent to slices.Conventional CT MD comprises consecutive slices with a thickness of 0.6 -1.5 mm and causes a radiation dose of 0.5 mSv on average.Technological improvement has been rapid resulting in 4-to 8-, 16-, 32-, 40-and 64-detector machines.It is known that there is no safe radiation dose.In all exposure quantities, there is a corresponding mutational risk [11].The head and neck region, the eyes and thyroid gland are the most radiosensitive organs.In addition to the cancer risk, radiation-induced cataracts are possible after multiple exposures [3].Image quality is affected by image noise or quantum mottle and is almost always directly related to the radiation dose.Keeping this in mind, the cone beam computed tomography seems to be relevant recent implementation in CT imaging.Essentially, it could be the best compromise so far between low radiation and excellent image quality [12]- [15].
As a conclusion, our study demonstrated that this methodology was easy to use and might have some use in studies aiming at radiation dose reduction.As was expected, 3 mm-slice-thick reconstruction CT had poor reproducibility and surgeon's confidence.More recent methods such as cone beam computed tomography scans have nowadays more relevant dose reduction potential.

Table 1 .
Characteristics of the patients.

Table 2 .
Medians and interquartile ranges (Q1 -Q3) of the radiologic Lund McKay (LM) scores evaluated by the radiologist from multi-detector sinus computed tomography scans (CT MD ) from 59 patients suffering from chronic rhinosinusitis.