A comparative study on image quality of two digital intraoral sensors

Abstract

Objectives:

The aims of this study were to evaluate the subjective image quality and reliability of two digital sensors. In addition, the image quality of the two sensors evaluated by specialists and general dentists were compared.

Methods:

30 intraoral bitewings from five patients were included in the study, 15 were exposed with a Dixi sensor (CCD-based) and 15 with a ProSensor (CMOS-based) using modified parallel technique. Three radiologists and three general dentists evaluated the images in pair. A five-point scale was used to register the image quality. Visual grading characteristics (VGC) analysis was performed to compare the image quality and the observer agreement was assessed in terms of intra class correlation co-efficient.

Results:

No statistically significant difference was found on image quality between the sensors. The average scores of the observer agreement were moderate with an average of 0.66 and an interval of 0.30 to 0.87, suggesting that there was a large variation on preference of image quality. However, there was a statistically significant difference in terms of the area under the VGC- curves between the specialist group and the general dentist group ( p = 0.043), in which the specialist group tended to favor the ProSensor.

Conclusions:

Subjective image quality of the two intraoral sensors were comparable when evaluated by both general and oral radiologists. However, the radiologists seemed to prefer the ProSensor to the Dixi as compared to general dentists. Inter- observer conformance showed a large variation on the preference of the image quality.

Introduction

During the past two decades, digital radiographs have gradually replaced analogue radiographs in Sweden. The digital technique has several advantages compared to the analogue technique, such as lower radiation dose to patients, possibility of image processing, fast image acquisition, no chemical processing and being environmentally friendly.^1–3

Many digital intraoral systems are currently available on the market, which can be classified into direct and indirect digital techniques depending on mode of acquisition. For the direct digital technique, the charged couple device (CCD) and the complementary metal oxide semi-conductor (CMOS) are the most commonly used intraoral image receptors. One advantage of CCD and CMOS, when compared with PSP, is the immediate display on the screen after exposure whereas the PSP needs to be scanned after exposure.³

CCD and CMOS techniques were invented around 1970, and it was not until 1987 that the CCD sensor was introduced to dentistry which was the first digital image receptor to be adapted for intraoral imaging.^3,4

CMOS technique does not differ much from CCD from a technical point of view. The two types of sensors work in a similar fashion except for how the electronic signals are transferred from each pixel of the detector. The CCD sensor has one output node per line or column in the pixel matrix, whereas for the CMOS sensor, the electric charge stored in each pixel is transferred to the output individually. Compared with the CMOS, the CCD sensor has some disadvantages, such as relative high cost of production and higher requirement for electronic energy supply. Previous studies on some earlier sensors have shown that the image quality of CCD- and CMOS based sensors are presently comparable.^4–6 Subjective image quality of a radiograph is not only dependent on technical properties of the X-ray unit and sensor, but may also depend on factors like the observers clinical experience, applied image processing, external factors such as the ambient light in the viewing environment, and possibly the monitor used for viewing.^7–12

Earlier studies comparing diagnostic accuracy of digital imaging by using CCD and CMOS sensors with E-speed film in the detection of periapical bony lesions demonstrated that no statistically significant differences were found among film, CCD, and CMOS-APS systems.^13,14 One study compared Ektaspeed plus film (EPF) with a Phosphor plate and a CCD detector for detection of periapical lesions and the results showed that EPF demonstrated the highest sensitivity and specificity, followed by PSP and CCD images.¹⁵

Since there is a constant development of digital techniques, new sensors should be evaluated and confirmed clinically with the existed systems with respect to image quality and diagnostic accuracy.³ Kitagawa compared the subjective image quality of CMOS and CCD sensors in an in vitro cadaver study. The results showed that the CMOS and CCD sensors were perceived to produce radiographic images of similar overall quality.⁵

A study by Shi et al investigated the psychophysical properties of the CCD-based Dixi sensor and the CMOS-based ProSensor in terms of dose response function and perceptibility curve test. It was concluded that by applying the new ProSensor, it might be beneficial to the patients by reducing dose and increasing perception for low contrast details.¹⁶ However, the conclusion may not be directly implemented in clinical routine since it is an in-vitro study performed on a test object with simple pattern and no image processing was applied to the test images. Therefore, the aim of the present study is to compare the subjective image quality of the two sensors by evaluating bitewing radiographs obtained from patients.

The hypothesis of the present study was that the ProSensor and Dixi sensor demonstrated comparable subjective image quality.

Aims

The aims of this study were to compare the subjective image quality of two digital sensors and to evaluate the reliability of two sensors in terms of inter observer agreement. In addition, the image quality of the two sensors evaluated by specialists and general dentists was compared.

Methods and materials

Ethical consideration

The study was approved by the regional ethics committee at Karolinska Institutet (Registration number 2011/515-31/2)

Assessment of radiographs

Patients with indication for bitewing radiographic examination were referred from adult clinic to Section of Oral Diagnostics and Surgery, Department of Dental Medicine, Karolinska Institutet, Sweden. The patients referred were informed about the project and possible radiation risk associated with an intraoral radiographic examination. Patient consent was received from those willing to participate before the X-ray examination. Two sets of bitewing radiographs were obtained for each patient, using both the Dixi and the ProSensor, (Planmeca OY, Helsingfors, Finland) by two final year dental students supervised by an experienced dental nurse. Modified paralleling technique was employed for image acquisition where proximal surfaces were free without overlapping. Sensor placement had to be similar when image areas were exposed with both sensors. The distance between X-ray focus to object was around 23 cm. The exposure parameters were set at 66 kV and 8 mA for both CCD and CMOS, and the images were exposed with the recommended exposure time, i.e. for the Dixi, the exposure times were 0.16 s for premolar bitewings and 0.20 s for molar bitewings. The corresponding exposure times for the ProSensor were, 0.12 s and 0.16 s, respectively.

A total number of five patients were examined with both sensors and a total number of 54 images were taken. Radiograph pairs of identical or very similar projection (discrete faults in projection was acceptable) and sensor placement were included in the study. Exclusion criteria were pairs of radiographs that did not have same field of view (FOV) as well as pairs of radiographs that had obvious difference in projection. Out of 54 images 15 pairs of the images with acceptable projection exposed with Dixi sensor and ProSensor were included in the present study.

One author who did not participate in the viewing session viewed all the images and performed necessary image enhancements in terms of histogram equalization and gamma function to ensure optimal light intensity and image contrast. Since the shape of the two sensors is different, to avoid recognizing the type of sensor used, the edges of all radiographs were cut out. The 15 pairs of radiographs were arranged in a PowerPoint presentation. On each slide, there was a pair of images, CCD or CMOS was randomly arranged on the left and right hand side (Figure 1).

Figure 1.

An example of the power point slide. The observers were instructed to compare the right hand image with the left hand image using a given criteria.

Evaluation of radiographs

Six dentists, three specialists in oral and maxillofacial radiology and three general dentists evaluated the images. All observers were familiar with digital technique. The radiologists’ working experience ranged from 3 to 35 years and for the general dentists it varied from 5 to 15 years. No calibration was performed in order to mimic the clinical situation. Written and verbal instructions for image evaluation of overall image quality were given to each observer. The evaluations were performed under identical conditions, pre-calibrated monitor according to the DICOM settings in a room with dimmed light.¹⁷ The observations were performed independently for each observer with a viewing distance of 50 cm.

The observers were informed to neglect possible projection-related difference and focus on the overall image quality between the pairs of images acquired by the two types of sensors taking into consideration of the following criteria:

1. Contrast resolution = Clear difference in gray shades between enamel and dentin, marginal bone levels, air, and the trabecular bone pattern.

2. Spatial resolution = Sharpness, how well fine details such as the lamina dura/pulp canal is visible in the image.

3. Subjectively experienced noise level.

The six observers were asked to evaluate the overall image quality of right hand image as compared to left hand image using a 1–5 grading scale: 1 = definitely better image quality, 2 = slightly better image quality, 3 = equally good image quality, 4 = slightly worse image quality 5 = definitely worse image quality.

Data analyses and statistics

The registered ordinal data were evaluated using the so called visual grading characteristics (VGC) analysis, which is a non-parametric rank-invariant statistical method for evaluating image quality.¹⁸ In VGC analysis, ratings for two image types are used to create a VGC curve, similar to the receiver operating characteristics (ROC) analysis in which ratings for signal (for example caries) and no-signal images (caries free) are used to create an ROC curve. The area under the VGC curve (AUC_VGC) ranges from 0 to 1 and was used as a measure of the difference in image quality between the two image types being compared. The AUC_VGC was proposed as a single measure of the difference in image quality between the two compared modalities. The area described the relationship between the proportions of fulfilled image criteria. In our set up, i.e. ProSensor on X-axis and Dixi on Y-axis, a VGC value of 0.5 meant that two sensors were equally good, a value higher than 0.5 meant that the observer assessed that CMOS had better image quality and a value lower than 0.5 meant the opposite. When conducting the VGC curve, for those slides where CCD images were positioned on the left hand, the registered data from 1 to 5 were kept; whereas for slides with CMOS positioned on the left hands the observers’ ratings were mathematically reversed to 5–1. In this way, the results represented the subjective image quality of CMOS as compared to that of CCD. Mann Whitney U test was used to study the difference between the specialist and general dentist group in terms of the areas under the VGC curves. The reliability of the two sensors was studied in terms of inter observer agreement applying intra class correlation coefficient (ICC).

Results

The comparison of subjective image quality based on the 15 pairs of images by the six observers is shown by a box-plot in Figure 2. Figure 3 presents the resulting VGC curve along with the operating points used by observer and the corresponding AUC_VGC values, with a mean value of 0.55 listed in Table 1. No statistically significant difference was found among observers, indicating that the subjective image quality between the Dixi sensor and the ProSensor was similar. However there was a statistically significant difference in terms of area under the VGC- curves between the specialist group and the general dentists group using Mann Whitney U test (p = 0.043) with a mean value of 0.65 vs 0.45, in which the specialist group tended to be more in favor of the ProSensor.

Figure 2.

Compared image quality by six observers. Y axis represents the registered scores of Dixi as compared to Prosensor: 1 = definitely better image quality, 2 = slightly better image quality, 3 = equally good image quality, 4 = slightly worse image quality 5 =definitely worse image quality. On the x axis, 3, 4 and 5 are specialists.

Figure 3.

VGC curves based on each observer. ProSensor on X-axis and Dixi on y axis, a VGC value of 0.5 meant that two sensors were equally good, a value higher than 0.5 meant that the observer assessed that CMOS had somehow better image quality, and a value lower than 0.5 meant the opposite.

Table 1.

Shows the six observers assessments in comparison of image quality taken with Dixi and ProSensor. The area was proposed as a single measure of the difference in image quality between the two compared modalities

Test Result Variable(s)	Area	Std. Errora	Asymptotic 95% Confidence Interval
			Lower Bound	Upper Bound
observer1	0.426	0.168	0.097	0.755
observer2	0.500	0.165	0.176	0.824
observer3a	0.685	0.142	0.406	0.964
observer4a	0.593	0.162	0.275	0.910
observer5a	0.685	0.143	0.406	0.965
observer6	0.426	0.163	0.106	0.746

^aspecialist in dentomaxillofacial radiology

Intra class Correlation Coefficient using the two-way random effects model with a consistency definition, showed that the average of the scores of the six observers was moderate reliable, average 0.66 with an interval of 0.30 to 0.87, suggesting that there was a large variation on preference of image quality.

Discussion

There are studies comparing caries detection, periapical lesions with different sensors, in which no statistically significant differences were found between the CCD sensor and the CMOS sensor.^13,14,19 According to the results of a literature search, the clinical research in comparing image quality between the CCD sensor and the CMOS sensor is scarce. A similar study to this current study is the one by Kitagawa H. et al who compared the subjective image quality of CMOS and CCD sensors in an in vitro cadaver study. The results showed that the CMOS and CCD sensors were perceived to produce radiographic images of similar overall quality.⁵ The purpose of this study was to compare the subjective image quality between the two sensors in a clinical set up. Shi et al reported that over the whole exposure range, the observers could detect a lower contrast difference on a simple test object for CMOS sensor.¹⁶ However, the difference observed from in vitro study seemed to be no longer obvious when involving six observers and clinical images with much more complicated anatomical structures. The specialist dentists tended to give higher values for CMOS sensor (VGC >0. 5). This may imply that the better contrast resolution may only be detected by specialists who have more experience in viewing radiographs and it is a known fact that the ability of the eye to evaluate different numbers of gray levels increases with practice. We know that a significant improvement in technical properties does not mean significantly better subjective image quality since there are many other factors involved in image viewing, including complicated anatomy, observer’s clinical experience, viewing environment and limitation of the human faculty of vision. Another important argument for a new sensor is the amount of radiation dose to the patient. In the present study, the exposure was 20–25% less for the ProSensor compared to the Dixi sensor. Therefore, the ProSensor may be recommended since it is provided with at least same image quality as the Dixi with reduced radiation dose to the patient.

In this study, the observers did not have the possibility to adjust the light intensity and contrast according to their own preference in viewing images. This is due to a technique barrier after the edges of the original images were cut, in order to avoid recognizing the CMOS images. Instead, the image quality was optimized for each image, taking into consideration the utilization of the whole range of histogram by histogram equalization and the optimal contrast by gamma function.^20,21 All the images were treated in the same manner.

Båth and Månsson introduced a method of comparing image quality obtained using different imaging systems or settings and termed it “visual grading characteristics” (VGC) analysis.¹⁸ It is a hybrid of receiver operating characteristics (ROC) and image criteria (IC). The strengths are that it is a valid statistical method and gives freedom for the observers. VGC enables the use of multiple scale steps and it can directly be used on the image quality criteria defined by the European Commission. In the present study the VGC method was applied to compare image quality with pre-optimized exposure parameters since the VGC is a relative method for comparing image quality between two image modalities without consideration of radiation exposure levels to the patients. One should be cautious in generalized application of the VGC method having in mind the importance of acquiring the diagnostically acceptable images instead of radiographic image with best image quality.

To our knowledge this study is the only clinical study on image quality, in which same projections were taken with two types of digital sensors. The results have therefore clinical relevance and provides important evidence for dentists in daily clinical practice. Future studies will focus on how image quality affects the diagnostic accuracy of various diagnostic tasks.

Conclusion

Subjective image quality of the two intraoral sensors were comparable when evaluated by both general and oral radiologists. However, the radiologists seemed to prefer the ProSensor to the Dixi as compared to general dentists. Inter- observer conformance showed a large variation on the preference of the image quality.