Detection of Monitoring Materials on Bedside Chest Radiographs with ...

Radiology

Thoracic Imaging Edith Eisenhuber, MD Alfred Stadler, MD, MSc Mathias Prokop, MD Michael Fuchsja¨ger, MD Michael Weber, MSc Cornelia Schaefer-Prokop, MD

Index terms: Radiography, bedside, 68.11 Radiography, comparative studies, 68.11, 68.1215 Radiography, storage phosphor, 68.1215 Receiver operating characteristic (ROC) curve Thorax, radiography, 68.11, 68.1215 Published online 10.1148/radiol.2271020045 Radiology 2003; 227:216 –221 Abbreviations: Az ⫽ area under ROC curve ICU ⫽ intensive care unit OD ⫽ optical density ROC ⫽ receiver operating characteristic 1

From the Department of Radiology and Ludwig Boltzmann-Institute for Clinical and Experimental Radiologic Research, University of Vienna, Wa¨hringer Gu¨rtel 18-20, A-1090 Vienna, Austria. Received February 6, 2002; revision requested April 15; revision received June 17; accepted August 8. Address correspondence to E.E. (e-mail: [email protected]).

Detection of Monitoring Materials on Bedside Chest Radiographs with the Most Recent Generation of Storage Phosphor Plates: Dose Increase Does Not Improve Detection Performance1 PURPOSE: To evaluate the performance of the most recent generation of storage phosphor plates for the detection of low-contrast catheter material on bedside chest radiographs. MATERIALS AND METHODS: In 10 patients in the intensive care unit, bedside chest radiographs were obtained with a 400-speed conventional screen-film system and with storage phosphor plates with exposure levels comparable to a 200-, 400-, or 800-speed conventional system. The chest radiograph was divided into 20 regions, 60% of which were superimposed with low-contrast catheter fragments. Six observers independently assessed the presence of catheter fragments by using a receiver operating characteristic (ROC) methodology. RESULTS: Detection performance (mean area under the ROC curve [Az]) with the storage phosphor plates was significantly superior to that with the screen-film system (Az ⫽ 0.76) at all three dose levels (Az ⫽ 0.88, 0.87, and 0.83 for 200-, 400-, and 800-speed doses, respectively; P ⬍ .05). Increasing the dose to a 200-speed system did not significantly increase detection performance compared with that with the 400-speed digital radiographs (Az ⫽ 0.88 vs 0.87). Dose reduction to 800 speed significantly deteriorated the detection performance (Az ⫽ 0.83) compared with that with the 400- and 200-speed digital radiographs, respectively. CONCLUSION: The most recent generation of storage phosphor plates is superior to a 400-speed screen-film system for the detection of catheter material, even at an exposure level of 800 speed. ©

Author contributions: Guarantors of integrity of entire study, E.E., C.S.P.; study concepts, E.E., C.S.P.; study design, E.E., C.S.P., M.P.; literature research, E.E.; clinical studies, E.E., M.F.; data acquisition, E.E., A.S.; data analysis/interpretation, all authors; statistical analysis, M.W., A.S., E.E.; manuscript preparation, E.E., C.S.P., M.P.; manuscript definition of intellectual content, E.E., C.S.P.; manuscript editing, E.E.; manuscript revision/ review, all authors; manuscript final version approval, C.S.P., M.P. ©

RSNA, 2003

216

RSNA, 2003

Bedside chest radiography is one of the most frequently performed radiologic examinations and is still of high diagnostic importance despite continuously changing technical developments (1). The effectiveness and limitations of bedside chest radiography in patients in the intensive care unit (ICU) have been well established (1–3). Owing to the lack of automatic exposure control and to rapidly changing patient conditions, conventional bedside chest radiography frequently has inconsistent image quality. Because of their properties of signal normalization, digital storage phosphor systems have been shown to provide more consistent image quality in bedside chest radiography regardless of acquired exposure dose (4,5). Previous generations of digital storage phosphor systems, however, had relatively lower spatial resolution and increased dose requirements compared with those of conventional systems (4,6,7). For older digital storage phosphor systems (ST II; Fuji, Tokyo, Japan), a

Radiology

75%–100% increase in exposure dose relative to that of conventional screen-film systems was found to be necessary to achieve comparable image quality (6). Although it has been well documented that more consistent and improved image quality yields an improved detection rate for monitor devices, a phantom study (7) involving the use of ST II plates revealed that a dose increase of only 150% was able to compensate for the increased noise in the high-attenuation areas of the mediastinum. The most recent storage phosphor plates, ST-V and ST-Vn (Fuji), are characterized by a substantially improved quantum efficiency. For upright chest radiographs, an equivalent detection performance, especially for pulmonary lesions, was found at the same exposure level as required by a 400-speed conventional system (8,9). A reevaluation of the dose requirements of digital storage phosphor radiography at the bedside, therefore, appears to be warranted. Assessment of correct positioning of various monitoring and support devices (eg, tubes and lines) is one of the most frequent indications for routine bedside chest radiography. Thus, the purpose of our study was to evaluate the performance of the most recent generation of storage phosphor plates (ST-Vn) for the detection of low-contrast catheter material on bedside chest radiographs.

MATERIALS AND METHODS Study Group and Image Acquisition From January to May 2001, we selected 10 patients (six women, four men; age range, 44 – 80 years; mean age, 56 years) from the University Hospital of Vienna ICU. Only patients in stable clinical condition in whom no major cardiopulmonary changes were expected during their participation in the study were selected for the study. Only patients who were scheduled to undergo clinically indicated daily bedside chest radiography and who were not expected to leave the ICU during the study period were included. The patients’ mean weight was 82 kg ⫾ 12 (SD) (range, 57–95 kg). Visual inspection of the radiographs showed that three of the 10 study patients were markedly overweight. Seven patients were mechanically ventilated. Five patients had cardiac failure, six patients had pleural effusions, and three patients had intrapulmonary opacities owing to either atelectasis or pneumonic infiltrates. Institutional review board approval was not required, and inVolume 227

䡠

Number 1

formed consent was not obtained. All radiographs were obtained for accepted clinical indications, and all studies were considered acceptable for patient care. Only those patients who were scheduled to undergo daily chest radiographic examinations were selected for the study. Beyond these clinically indicated radiographs, no extra bedside chest radiographs were obtained. On each of 4 consecutive days, we obtained bedside chest radiographs by using one of the following four techniques on each occasion: a conventional screenfilm system (400 speed, Insight VHC; Eastman-Kodak, Rochester, NY) and storage phosphor plates that were exposed with the same, with half, and with double the dose as compared with that of the conventional system (corresponding to the dose of a 200-, 400-, or 800-speed screen-film system). All radiographs were obtained with the patient in the supine position, with a focus-film distance of 100 cm without automatic exposure control. Exposure parameters were kept constant for all images, except for the milliampere second values (tube voltage of 125 kVp, antiscatter grid with 40 lines per centimeter, and a grid ratio of 10:1). We adjusted the x-ray dose by changing only the milliampere second settings. The images were obtained by using a portable x-ray generator (Mobilett II; Siemens, Erlangen, Germany). For conventional chest radiography, we used an asymmetric screen-film combination ([35 ⫻ 43 cm] Insight VHC). Digital images were obtained with a storage phosphor radiography unit (Fuji film, FCR 5000; Fuji, Tokyo, Japan). We used ST-Vn imaging plates with a matrix of 4,280 ⫻ 3,520, 10 bits per pixel, a pixel size of 0.1 mm, and a format of 35 ⫻ 43 cm. Processing was performed with the parameters defined at our institution for acquisition of bedside chest images. In the terminology of the manufacturer (Fuji), this comprised a sigmoid gradation curve, a slight unsharp mask filtering with an enhancement factor of 0.5, and a frequency rank of RN ⫽ 0. We also used an additional dynamic range compression, as offered by the manufacturer. All images were printed on hard copies of 34.5 ⫻ 43 cm by using a laser imager (Filmprinter, Fm-DP 3543; Fuji). In each patient, we obtained the conventional radiograph first. As in a routine clinical situation, the exposure values for this conventional radiograph were determined by a radiographer experienced in manually choosing the exposure values, depending on the constitutional charac-

teristics of the patient. The acquisition radiation dose ranged from 1.25 to 2.2 mAs (mean, 1.5 mAs). All chest radiographs used in the study were obtained by the same radiographer. Acquisition parameters of the digital radiographs were accordingly adjusted to the identical level (400 speed), 50% (800 speed), and 200% (200 speed) of the dose level of the corresponding conventional radiograph. For each of the 10 patients, we created a different template, which was then kept constant for each of the four images per patient. The catheter fragments were taped onto the hard copy of a prior chest radiograph to facilitate placement of the catheters. The radiograph used as the template had no influence on the subsequent images. The fragments were equally distributed over the areas of the mediastinum and the lung. Each template consisted of different types of catheters organized in a randomly chosen order. Since the radiographs were also used for clinical purposes, a grid was drawn onto the radiographs retrospectively so that it would not interfere with the diagnostic review. One hard copy of all digital chest radiographs was made for diagnostic purposes; the grid was drawn on a second hard copy. A copy of the conventional chest radiograph was used for diagnostic review and remained in the patient file. The original version of the conventional radiograph was included in our study. The grid divided the chest radiograph into 20 fields. Sixty percent of the fields contained catheter fragments; 40% were empty. Each field could be empty or contain one type of catheter fragment. The following three types of monitoring materials were used: (a) a 7-F central venous catheter (Certofix; Braun, Melsungen, Germany) with two lumina; (b) an 8-F chest tube (Plastimed; Laboratoire Pharmaceutique, Saint-Leu-La-Foret, France); and (c) a nasogastric feeding tube (Wiruthan; Ru ¨ sch, Kernen, Germany). The length of the catheter fragments varied between 1.5 and 2.5 cm. To superimpose the catheter fragments onto the anteroposterior chest radiograph, the template was placed between the patient and the radiographic cassette. Care was taken to place the template in a position as identical as possible for the different radiographs. To minimize effects of potential clinical changes, the three digital images were obtained in different orders. Retrospective review of the study images with respect to clinical changes revealed minor changes in three patients. In two patients, a minor increase in pleural effusion occurred during

Most Recent Generation of Storage Phosphor Plates

䡠

217

Detection Performance with Conventional Screen-Film and Digital Radiography Radiography Technique

Mean Az

95% CI

Radiology

Overall Detection Conventional Digital 400-Speed dose 800-Speed dose 200-Speed dose

0.76

0.74, 0.79

0.87* 0.83* 0.88*

0.85, 0.89 0.80, 0.84 0.86, 0.90

Conventional Digital 400-Speed dose 800-Speed dose 200-Speed dose

0.72

0.67, 0.76

0.83* 0.78 0.90*

0.79, 0.87 0.74, 0.82 0.86, 0.92

Mediastinum

Retrocardiac Area and Chest Wall Conventional Digital 400-Speed dose 800-Speed dose 200-Speed dose

0.76

0.72, 0.80

0.90* 0.86* 0.87*

0.87, 0.93 0.82, 0.89 0.84, 0.90 Lung

Conventional Digital 400-Speed dose 800-Speed dose 200-Speed dose

0.79

0.74, 0.84

0.86 0.84 0.87

0.82, 0.89 0.80, 0.88 0.83, 0.91

* Significantly superior to conventional radiography (P ⬍ .05).

acquisition of the high-dose digital image. In one patient, minor improvement of cardiac insufficiency was seen when acquiring the 400-speed digital image. The remaining seven patients had comparable radiographic findings on each of the four images.

Image Evaluation All digital and conventional images were divided into two reading subsets and presented in random order. The radiographs were analyzed independently by six radiologists (including E.E., A.S., M.F., C.S.P.) with varying experience in chest radiography (range, 2–16 years; mean, 7 years). To avoid learning-curve effects, the interval between the two reading sessions was at least 2 weeks. To minimize reading order bias, all images were rerandomized for each reader. For each test field, the readers were asked to note the presence or absence of a catheter fragment by using a five-score confidence scale: 1, definitely not present; 2, probably not present; 3, equivocal; 4, probably present; and 5, definitely present. The readers were not informed about the numbers and distribution of the catheter fragments. The readers were informed that each field could be empty or contain one single type of catheter fragment. Each of 218

䡠

Radiology

䡠

April 2003

the 20 fields per image was separately evaluated for the presence or absence of a catheter fragment. An attempt was made to evaluate the effect of the catheter localization on the detection performance: Catheter background was classified according to its optical density (OD) as representing low-OD (ie, high-attenuation) areas (central mediastinum along thoracic spine), intermediate-OD (ie, intermediate-attenuation) areas (retrocardiac and retrodiaphragmatic areas and chest wall), and high-OD (ie, low-attenuation) areas (lungs). Low-OD areas had an OD of less than 0.5, intermediate-OD areas had an OD between 0.5 and 1.1, and high-OD areas had an OD of more than 1.1. Forty-eight (40%) of 120 catheter fragments were placed in low-OD areas, 40 (33%) were placed in intermediate-OD areas, and 32 (27%) were placed in high-OD areas. Data analysis was based on 800 observations (10 patients ⫻ four images each ⫻ 20 fields) by each reader, resulting in a total of 4,800 observations (800 observations ⫻ six radiologists) for the entire study. There was no limit placed on reading time. The hard-copy images were mounted sequentially on a single light box in a room with no other ambient illumination. The standard viewing dis-

tance, as measured from the light-box surface to the eye of the observer, was approximately 50 cm. The readers, however, were allowed to alter the distance according to their individual preferences. All images were interpreted under identical conditions with use of the same viewing box in the same room and low ambient light conditions. The conventional and digital radiographs were not identified as such, but the observers readily ascertained which were the digital and which were the conventional images on the basis of the specific properties of the laser film. There were no signs indicating the acquisition dose. Data for 1,200 observations (10 patients ⫻ 20 fields ⫻ six radiologists) per image modality were analyzed according to receiver operating characteristic (ROC) analysis. The ROC analysis was performed by using a computer program (MedCalc, version 6.0; Medcalc Software, Mariakerke, Belgium). Observer performance was expressed in mean areas under the ROC curve. The 95% CIs for the differences in areas under the ROC curve (Az) among the various imaging modalities provided evidence of whether the difference was significant.

RESULTS Detection Performance Mean areas under the ROC curve and 95% CIs are given in the Table to illustrate detection performance with the conventional screen-file system versus that with the digital system at the various levels of acquisition dose. The detection performance of the storage phosphor plates was significantly superior to that of the screen-film system at all three dose levels (Fig 1). Increasing the dose to a 200-speed system did not significantly increase detection performance compared with that with the 400-speed digital radiographs. Dose reduction to 800 speed significantly deteriorated the detection performance compared with that with the 400- and 200-speed digital radiographs.

Effect of Catheter Localization High-attenuation area (low OD).—For the central portion of the mediastinum (thoracic spine), catheter localization performance with the 200- and 400speed storage phosphor radiographs was significantly superior to that with the screen-film system. We observed no statistically significant difference between the 800-speed storage phosphor radioEisenhuber et al

Radiology Figure 1. Images of a focal area with all four techniques in one patient: (a) conventional frontal radiograph and (b) 400-speed, (c) 800-speed, and (d) 200-speed digital radiographs. The catheter (arrow) in the high-attenuation area of the mediastinum is better seen on all three digital radiographs than on the conventional radiograph.

graphs and the screen-film radiographs in this region. Intermediate-attenuation area (intermediate OD).—For the retrocardiac and retrodiaphragmatic spaces and the chest wall, catheter localization performance with the storage phosphor radiographs was significantly superior to that with the screen-film radiographs, regardless of the acquisition dose. No statistically significant difference was observed among the various digital radiographs. Volume 227

䡠

Number 1

Low-attenuation area (high OD).—For the lungs, we found no statistically significant difference between the screenfilm and storage phosphor radiographs, regardless of the acquisition dose (Fig 2).

DISCUSSION Digital radiography based on storage phosphor systems is a well-accepted and already approved system, especially for

acquisition of bedside radiographs in the ICU (3,4). Storage phosphor radiography has been shown to provide a more consistent and uniform image quality, even under difficult conditions (5). Spatial resolution, which is limited by the pixel size, poses no problem in bedside radiography. However, storage phosphor radiographs have been reported (6,7) to have higher image noise and thus lower contrast resolution compared with those of an ideally exposed conventional radio-


䡠

219

Radiology Figure 2. Images of a focal area with all four techniques in one patient: (a) conventional frontal radiograph and (b) 400-speed, (c) 800-speed, and (d) 200-speed digital radiographs. The catheter (arrow) in the low-attenuation area of the lung is seen equally well with all techniques.

graph, particularly in the high-attenuation areas of the mediastinum and the retrocardiac space. In fact, authors of previous studies (6,7) recommend the use of a higher acquisition dose to acquire storage phosphor radiographs if depiction that is equivalent to that on conventional radiographs is the goal. This was shown for catheter material as well as geometric patterns in a phantom study (6,7). However, all of these studies involved the use of older-generation storage phosphor plates that are known to have substantially lower detection quantum efficiency and therefore higher image noise at a given acquisition dose. The most recent generation of storage phosphor plates is characterized by substantially increased quantum efficiency owing to a different chemical composition of the detector material, as well as a smaller crystal size associated with an increased packing density (8). For upright 220

䡠

Radiology

䡠

April 2003

chest radiography, storage phosphor plates have been shown to have at least equivalent performance compared with conventional radiographs obtained with a 400-speed acquisition dose (9,10). In addition, in clinical practice there is a variable tendency to increase the acquisition dose because this procedure yields lower image noise and the perception of increased image quality (11). Thus, a reevaluation of the performance of the most recent storage phosphor plates for bedside radiography appeared to be warranted. We decided to evaluate the detection of catheter material because (a) it is one of the most frequent clinical indications for bedside chest radiography; (b) most catheters appear as a low-contrast linear opacity, which makes them a suitable test object for evaluating the effect of image noise as a function of acquisition dose; and (c) we wanted to directly com-

pare our results with previously reported results obtained with older detector systems. We found that the overall performance of the digital radiographs, as described by the area under the ROC curve, was significantly superior to that with the conventional radiographs at all three exposure levels. Acquisition of digital radiographs with the dose of an 800-speed conventional system was still superior to the acquisition of conventional radiographs but significantly poorer compared with the acquisition of 200-speed and 400speed digital images. Even in the highattenuation region of the central mediastinum along the thoracic spine (low-OD area), the 800-speed digital radiographs still performed equivalently to the conventional radiographs. This finding represents a substantial improvement to previously reported results obtained with Eisenhuber et al

Radiology

older generations of storage phosphor plates (6,7). The better signal-to-noise ratio on the 200-speed radiographs did not translate into a significantly superior visualization of monitoring materials. All radiographs were obtained by using an antiscatter grid. This improves detection of low-contrast structures whose visualization would otherwise be deteriorated by scatter. Thus, it could be argued that the effective scatter reduction, especially in the high-attenuation areas, contributed to the fact that the decreased noise on the 200-speed images did not result in further improvement in detection performance. The effect of the obesity of patients on image quality represents a problem, especially in bedside chest radiography with the digital and conventional techniques. Storage phosphor technology is characterized by a higher structural contrast when appropriate processing is used. As a consequence, there is increased depiction of monitoring devices on digital bedside radiographs compared with that on conventional radiographs (4,5). This effect is probably even more apparent with softcopy reading when interactive adjustment of window width and window level is available. Although the number of obese patients in our study group was small, we suspect that an intrinsic advantage of the digital technique (contrastenhancing processing) was its capability to compensate for the decreased signalto-noise ratio on images obtained in obese patients when they were acquired with a lower acquisition dose. The technique with high kilovolt peak and high grid ratio was chosen because it is the normally applied technique in patients in the ICU at our institution. We still review our bedside radiographs on hard copies without the availability of interactive windowing. The goal with this technique therefore is to obtain images of possibly high transparency of the high-absorption areas of the mediastinum for optimum visualization of catheter material. The high kilovolt peak setting is chosen to decrease acquisition time and to increase penetration of highabsorption areas. The latter effect is further emphasized with the use of an antiscatter grid. The increased susceptibility of high ratio grids to misalignment does not pose a serious problem in our institution because our technicians are experienced in this technique. None of the study images had a misaligned grid. On hard-copy images with lower contrast resolution within the mediastinum (when Volume 227

䡠

Number 1

obtained without a grid), reduction of the acquisition dose is likely to substantially decrease the visualization of catheter material. Our study had the following limitations. The study group of 10 patients was relatively small. This may have affected the performance of conventional radiography since it is known that the image quality of conventional radiographs is much more variable under ICU conditions owing to the lack of automatic exposure control and to rapidly changing patient conditions. In a larger study group, the varying image quality may have a smaller effect on performance. However, we took care to ensure that the 10 conventional radiographs included in this study were all of standard to good image quality and were considered to be representative of ICU chest radiographs. Although only patients in stable clinical condition were included in the study, minor changes during the study period could not be fully excluded. To minimize the effects of clinical changes, the three digital images were obtained in different orders. To minimize learning bias, the images were presented in a different random order in two reading sessions. Two versions of one template were reviewed within one session. Learning bias could not be excluded; however, we think that it was small since the detection of a catheter fragment in a predefined chest area is a very specific task with a lower learning effect as compared with the diagnostic review of the whole chest radiograph. To further decrease learning effects, the random order was altered for each reader. Another limitation is that a single radiograph was used to test the detection of several catheter fragments. We think that this was acceptable because we intended to assess the performance of a very specific task—namely, the detection of catheter fragments—rather than a diagnostic capability. The effect of acquisition dose on the detectability of catheters was further evaluated for different subareas of the chest with respect to their different attenuation characteristics. In summary, we found that the improved dose efficiency of the most recent storage detector systems resulted in a significantly better performance in the localization of catheter material on ICU bedside chest radiographs than has been previously reported. The most recent generation of storage phosphor plates is superior to a 400-speed screen-film system for the detection of catheter fragments on bedside chest radiographs, even at an ex-

posure level of 800 speed. Image acquisition with a dose comparable to that with a 400-speed conventional system yields superior results compared with those achieved with conventional reference radiography performed with the same dose. An increased dose does not result in further improvement in the detection performance of a digital ICU chest radiograph when the image is obtained with an antiscatter grid. Acknowledgments: The authors acknowledge and appreciate the participation of Martin Uffmann, MD, and Patrick Wunderbaldinger, MD, in the panel of observers. References 1. Wandtke JC. Bedside chest radiography. Radiology 1994; 190:1–10. 2. Hall JB, White SR, Karrison T. Efficacy of daily routine chest radiographs in intubated, mechanically ventilated patients. Crit Care Med 1991; 19:689 – 693. 3. Henschke CI, Yankelevitz DF, Wand A, Davis SD, Shiau M. Accuracy and efficacy of chest radiography in the intensive care unit. Radiol Clin North Am 1996; 34:21– 31. 4. Niklason LT, Chan HP, Cascade PN, Chang CL, Chee PW, Mathews JF. Portable chest imaging: comparison of storage phosphor digital, asymmetric screen-film, and conventional screen-film systems. Radiology 1993; 186:387–393. 5. Schaefer CM, Greene RE, Oestmann JW, et al. Improved control of image optical density with low-dose digital and conventional radiography in bedside imaging. Radiology 1989; 173:713–716. 6. Dobbins JT III, Rice JJ, Beam CA, Ravin CE. Threshold perception performance with computed and screen-film radiography: implications for chest radiography. Radiology 1992; 183:179 –187. 7. Galanski M, Prokop M, Thorns E, et al. The visibility of a central venous catheter using digital luminescence radiography in intensive care radiology. Rofo Fortschr Geb Rontgenstr Neuen Bildgeb Verfahr 1992; 156:68 –72. [German] 8. Dobbins JT III, Ergun DL, Rutz L, Hinshaw DA, Blume H, Clark DC. DQE (f) of four generations of computed radiography acquisition devices. Med Phys 1995; 22:1581–1593. 9. Bernhardt TM, Otto D, Reichel G, et al. Detection of simulated interstitial lung disease and catheters with selenium, storage phosphor, and film-based radiography. Radiology 1999; 213:445– 454. 10. Mansson LG, Kheddache S, Lanhede B, Tylen U. Image quality for five modern chest radiography techniques: a modified FROC study with an anthropomorphic chest phantom. Eur Radiol 1999; 9:1826 – 1834. 11. Weatherburn GC, Davies JG. Comparison of film, hard copy computed radiography (CR) and soft copy picture archiving and communication (PACS) systems using a contrast detail test object. Br J Radiol 1999; 72:856 – 863.


䡠

221