1992, The British Journal of Radiology, 65, 852-856

Portable chest radiography in intensive care: a comparison of computed and conventional radiography By "Paul Jennings, MRCP, FRCR, *Simon P. G. Padley, MRCP, FRCR and David M. Hansell, MRCP, FRCR Department of Radiology, Royal Brompton National Heart and Lung Hospital, Sydney Street, London SW3 6NP, UK {Received 8 November 1991 and in revised form 9 March 1992, accepted 6 May 1992) Keywords: Digital radiography, Chest radiography

Abstract. A prospective study was performed comparing computed and conventional radiography for the detection and visibility of cardiovascular devices in intensive care unit patients. Computed images were obtained using a commercially available 2K x 2K, 12-bit storage phosphor plate system. Image sets from 50 patients were assessed independently by three observers. A significant difference between the types of image was found for the detection of mediastinal drainage tubes and prosthetic valves. Computed images allowed greater confidence in the identification of courses and tips of lines. This advantage was most marked with edgeenhanced computed images.

Consistent high-quality portable chest radiographs are difficult to obtain. Computed radiography employing reusable photostimulable phosphor plates has several advantages: the wide dynamic range of these imaging plates ensures that retakes due to incorrect exposure are rarely necessary. In addition, a reduction in radiation dose to the patient, whilst obtaining high-quality images, can be achieved (Cohen et al, 1989). Digital aquisition of the image enables post-processing such as contrast manipulation and edge enhancement. A theoretical disadvantage of computed radiographic systems is their reduced spatial resolution. Conventional screen/ film combinations have a resolution of greater than 10 line pairs/mm whilst computed systems achieve at best 5 line pairs/mm in adult chest radiography. In most centres, portable chest radiography is most frequently requested on the intensive care unit and in this context the evaluation of drains, catheters and other cardiovascular devices is of prime importance. These devices also lend themselves to detection studies as their presence or absence can be verified with certainty. We undertook a prospective study to compare objectively the performance of a computed radiographic system, the first of its kind in clinical operation in the UK, with conventionalfilm/screenradiography in this setting. The degree to which post-processing of the digital image improved the detection of cardiovascular devices was also investigated. Address correspondence to Dr D. M. Hansell, Department of Radiology, Royal Brompton National Heart and Lung Hospital, Sydney Street, London SW3 6NP, UK. *Present addresses: Dr P. Jennings, Department of Imaging, The Ipswich Hospital, Heath Road, Ipswich, Suffolk, and Dr S. Padley, Department of Radiology, Vancouver General Hospital, 10th Avenue and Heather Street, Vancouver, BC, Canada. 852

Materials and methods

A Fuji FCR AC-1 computed radiographic system (Fuji Photo Film UK Ltd) was in clinical operation on the adult intensive care unit for 3 months before the study was started. This system uses a reusable photostimulable phosphor plate (europium-doped barium fluoride bromide) housed in a cassette which is exposed in the same manner as a conventional film cassette. The imaging plate is scanned by a helium-neon laser which stimulates luminescence from the latent image; this signal is detected by a photomultiplier and subseqently digitized with a pixel matrix of 1760x2140 (pixel size 0.2 mm x 0.2 mm). The digital information is processed and recorded by laser printer on to film (Sonada et al, 1983). This system is configured to give the option of a single image on 24 cm x 30 cm film or a pair of smaller images, one of which is processed to resemble a conventional radiograph with a non-linear gradient curve and minimal spatial frequency enhancement; the second image of the pair has a wide latitude linear gradient and spatial frequency enhancement six times normal, generated by unsharp masking, which produces an edgeenhanced effect (Fig. 1). For this study all images were produced in the paired format. Approval for this study was given by the Hospital Ethics Committee. Consent for two radiographs to be taken, conventional film/screen and computed, was obtained from 50 consecutive adult patients undergoing major cardiopulmonary surgery. The sets of radiographs were obtained at the same time during thefirstor second post-operative day on the intensive care unit. The film/screen combination was Kodak XL with Regular calcium tungstate screens. Both computed and conventional images were obtained at 80 kVp and 3 mAs using Picker Explorer mobile units. The computed image pairs were separated and each of the three images (conventional, computed and The British Journal of Radiology, October 1992

Comparison of computed and conventional radiography in intensive care

visibility of the course and tip of all lines including central venous catheters and intercostal tubes were scored on a five point scale: 5, definitely identified; 4, probably identified; 3, indeterminate; 2, probably not identified; 1, definitely not identified. The significance of any difference in visibility of devices in the three types of image was measured using McNemars' test. Confidence levels for course and tip visibility were compared using the Wilcoxon matched pairs signed ranks test and binomial two-tailed /rvalues obtained (Armitage & Berry, 1971). Results

Figure 1. Format of laser printed hard copy pair. Lower image with wide latitude linear gradient and high spatial frequency enhancement.

computed enhanced) were assessed independently on separate occasions by three radiologists who were accustomed to interpreting computed radiographs. A questionnaire was completed for each image, in which the observers recorded their identification of individual cardiovascular devices. Each patient had at least one device in situ when the radiographs were obtained. The numbers of each type of device detected were summed for the three observers to produce detection scores. The

A total of 159 devices were present in the 50 patients, consisting of central venous catheters, Swan-Ganz catheters, endotracheal tubes, prosthetic valves, permanent pacing wires, intercostal and mediastinal drainage tubes. Detection scores for the different devices are shown in Table I. All Swan-Ganz catheters, endotracheal tubes and pacing wires were detected on each of the three types of radiograph. Other devices were most frequently detected on the computed radiography images with highest detection scores for enhanced images: the most marked differences in detection scores were found for prosthetic valves and mediastinal drains. For the latter, 44% were detected on conventional film and 76% on enhanced computed images (p < 0.025). No statistically significant differences were found in the detection scores for central catheters; however, significant differences were found in the confidence levels for the evaluation of the courses and tips of these devices, with the highest levels of confidence recorded for computed radiographs with enhancement (p < 0.0001). Similar results were obtained for mediastinal drains (p < 0.001). Subsequent review of the films alongside each other confirmed that the reason for missed or inadequately visualized devices was due to relative underexposure of the mediastinum, chest wall or upper abdomen. This was most obvious on conventional film/screen radiographs (Fig. 2). Discussion

In conventional radiography the film has three tasks, namely the capture, display and storage of the image.

Table I. Detection scores for each device summed for the three observers Device («)

Central venous catheters (102) Mediastinal drains (15) Endotracheal tubes (15) Intercostal tubes (13) Prosthetic valves (9) Pacing wires (3)

Vol. 65, No. 778

Detection scores (%) Conventional

Standard computed

Enhanced computed

290 20 45 38 14 9

302 25 45 38 21 9

295 34 45 39 25 9

(95) (44) (100) (97) (52) (100)

(99) (56) (100) (97) (78) (100)

(96) (76) (100) (100) (93) (100)

853

P. Jennings, S. P. G. Padley and D. M. Hansell

Figure 2. Examples of the three types of image to demonstrate the effect of image processing, (a) Conventional film/screen radiograph, (b) standard computed radiograph, and (c) enhanced computed image. Note the improved visibility of prosthetic valves and tips of central venous lines on the enhanced image.

For the function of image display and storage, film is excellent. As a receptor, however, film is far from ideal. The spatial resolution of film/screen combinations is more than adequate but the requirements of wide latitude and high speed tend to be mutually exclusive in terms of good contrast and low image noise (Merritt et al, 1985). A compromise is usually made between wide latitude to encompass a necessary wide exposure range {e.g. lung through to mediastinum) and sufficient contrast for the visibility of low-contrast detail. These drawbacks are compounded in portable bedside radiography by the difficulties of patient positioning, a short film-focus distance, poor control of scattered radiation and a lack of automatic exposure devices. As a result many chest radiographs obtained with portable equipment are of suboptimal diagnostic quality. 854

Computed radiographic systems offer the advantages of images that have been optimized for a wide range of exposure conditions (Fraser et al, 1989). The photoluminescence/dose response is almost two orders of magnitude greater than a conventionalfilm/screencombination and this may also allow a lower patient radiation dose to be used. The minimum dose is limited in practice by an increase in quantum noise in the image as the exposure is decreased. Comparative studies have addressed the potential problem of lower spatial resolution in digital systems, particularly in their application to chest imaging (Goodman et al, 1986, 1988; Lams & Cocklin, 1986). These investigators concluded that a pixel size less than 0.4 mm is required for the accurate detection of subtle abnormalities such as pneumothoraces and septal lines. The British Journal of Radiology, October 1992

Comparison of computed and conventional radiography in intensive care

Most of these studies used digitized images taken graphy for portable chest radiographs. Furthermore, from conventional chest radiographs but more recently this advantage is increased by post-processing of the comparisons of conventional and computed radio- digital image using a wide latitude, linear dose/density graphic systems using photostimulable phosphors have gradient and edge enhancement. The potential problems been published. These range from the subjective percep- of an unfamiliar appearance of the lungs and reduced tions of diagnostic quality, which proved favourable visibility of interstitial disease can be overcome by the (Fuhrman et al, 1988), to carefully designed observer simultaneous display of an unenhanced "analogue lookperformance studies of the detection of pneumothorax alike" image with the enhanced image. (Fajardo et al, 1989; Kehler et al, 1990), mediastinal lesions and pulmonary nodules (Morioka et al, 1988; References ARMITAGE, P. & BERRY, G., 1971. Statistical Methods in Schaefer et al, 1990). Medical Research (BlackweU Scientific Publications, There have been some discordant results but in Oxford). general, for these, tasks computed radiography has been COHEN, M. D., LONG, B., CORY, D. A , BRODERICK, N. J. & equivalent or superior to conventional radiography. SMITH, J. A., 1989. Digital imaging of the newborn chest. Clinical Radiology, 40, 365-368. Post-processing of the digital image using techniques such as spatial frequency "edge" enhancement by FAJARDO, L. L., HILLMAN, B. J., POND, G. D., CARMODY, R. F., JOHNSON, J. E. & FERRELL, W. R., 1989. Detection of unsharp masking theoretically offers the potential for pneumothorax: comparison of digital and conventional increased visibility of features such as pneumothoraces imaging. American Journal of Roentgenology, 152, 475-480. and low-contrast structures (Ishida, 1987). This was FOLEY, W. D., GOODMAN, L. R. & WILSON, C. R., 1987. The confirmed by Goodman et al (1988) but the authors effect of varying pixel size and image processing on the noted that for abnormalities other than pneumothorax, detection of interstitial lung disease. In Proceedings of the the edge enhanced images were significantly worse. Chest Imaging Conference, Madison, Wisconsin (Medical Several studies have raised doubts about the accuracy of Physics Publishing Corp), pp. 113-123. digital radiography in the detection of interstitial lung FRASER, R. G., SANDERS, C , BARNES, G. T., MCMAHON, H., disease with, under certain circumstances, inferior GIGER, M. L., Doi, K., TEMPLETON, A. W., Cox, G. C , DWYER, S. J., MERRITT, C. R. B. & JONES, J. P., 1989. performance compared with conventional radiographs Digital imaging of the chest. Radiology, 171, 297-307. (McMahon et al, 1985; Foley et al, 1987; Goodman et al, 1988; Schaefer et al, 1991). Post-processing inevitably FUHRMAN, C. R., G U R , D., GOOD, B., ROCKETTE, H., COOPERSTEIN, L. A. & FEIST, J. H., 1988. Storage phosphor changes the characteristics of normal lung features such radiographs versus conventional films: interpreters as vessels and may certainly cause difficulties in interperceptions of diagnostic quality. American Journal of pretation for those unfamiliar with the appearances. Roentgenology, 150, 1011-1014. The degree and type of image post-processing have a GOODMAN, L. R., FOLEY, W. D., WILSON, C. R., RIMM, A. A. & profound effect on the visibility of different structures LAWSON, T. L., 1986. Digital and conventional chest images: within the chest. One elegant study has demonstrated observer performance with film digital radiography system. Radiology, 158, 27-33. that with high-frequency edge enhancement, computed radiographs can be equal in performance to film/screen GOODMAN, L. R., FOLEY, W. D., WILSON, C. R., TIKOFSKY, R. S. & HOFFMAN, R. G., 1988. Pneumothorax and other radiography for the detection of interstitial lung disease lung diseases: effect of altered resolution and edge (Schaefer et al, 1991). Post-processing needs to be enhancement on diagnosis with digitised radiographs. tailored. As in other digital techniques such as Radiology, 167, 83-88. computed tomography it is likely that no single image ISHIDA, M., 1987. Image processing. In Computed Radiography, display will be sufficient for all tasks. ed. by Y. Tateno, T. Iinuma and M. Takano (SpringerMost clinical users of computed radiographic systems Verlag, Tokyo), pp. 25-30. routinely obtain a pair of images as described above KEHLER, M., ALBRECHTSSON, U., ANDRESDOTTIR, A., LARUSDOTTIR, H. & LUNDIN, A., 1990. Accuracy of digital (Fig. 1), one configured to match closely a conventional radiography using stimulable phosphor for diagnosis of chest radiograph, the other post-processed to enhance pneumothorax. Ada Radiologica, 31, 47-53. certain features. To date there has been only one study LAMS, P. M. & COCKLIN, M. L., 1986. Spatial resolution of the supposed advantage of this dual format and of requirements for digital chest radiographs: an ROC study of the relative value of the two images (Saito et al, 1988). observer performance in selected cases. Radiology, 158, This paper assessed a wide range of features including 11-19. unverified lung pathology. Images were assessed by only MCMAHON, H., VYBORNY, C. J., SABETI, V., METZ, C. E. & one observer and statistical evaluation was not Doi, K., 1985. The effect of digital unsharp masking on the detectability of interstitial infiltrates and pneumothoraces. attempted. Another study that looked specifically at the Proceedings of SPIE, 555, 246-252. detection of cardiovascular devices relied on images taken with the two techniques several hours apart and MERRITT, C. R. B., MATHEWS, M. D., SCHEINHORN, D. & BALTER, S., 1985. Digital imaging of the chest. Journal of assessed the standard and enhanced computed images Thoracic Imaging, 7(1), 1-13. concurrently and compared these with conventional MORIOKA, C , BROWN, K., BALTER, S., MILOS, M. J., HUANG, radiography (Thompson et al, 1989). H. K., KANGARLOO, H., BOECHAT, I. M. & BATRA, P., 1988. Our results confirm that there is an objective improveReceiver operator characteristic analysis of chest ment in the detection of cardiovascular devices using radiographs with computed radiography and conventional analogue films. Radiology, 169, 349. computed radiography rather than conventional radioVol. 65, No. 778

855

P. Jennings, S. P. G. Padley and D. M. Hansell SAITO, H., KURASHINA, T., ISHIBASHI, I., SHIMANUKI, Y. &

SAKAMOTO, K., 1988. Digital radiography in an intensive care unit. Clinical Radiology, 39, 127-130.

LINDEMANN, S. R., 1991. Interstitial lung disease: impact of

postprocessing in digital Radiology, 178, 733-738.

storage

phosphor

imaging.

SCHAEFER, C. M., GREENE, R., OESTMANN, J. W., HALL, D. A.,

SONADA, M., TAKANO, M., MIYAHARA, J. & KATO, H., 1983.

LLEWELLYN, H. J., MCCARTHY, K. A., PILE-SPELLMAN, E. R. & RUBENS, J. R., 1990. Digital storage phosphor imaging versus conventional film radiography in CT-documented chest disease. Radiology, 174, 207-210. SCHAEFER, C. M., GREENE, R., OESTMANN, J. W., LLEWELLYN,

Computed radiography utilizing scanner laser stimulated luminescence. Radiology, 148, 833-838. THOMPSON, M. J., KUBICKA, R. A. & SMITH, C , 1989. Evaluation of cardiovascular devices on chest radiographs: digital vs analogue radiographs. American Journal of

H. J., MROSE, H. E., PILE-SPELLMAN, E. R., RUBENS, J. R. &

856

Roentgenology, 153, 1165-1168.

The British Journal of Radiology, October 1992

Portable chest radiography in intensive care: a comparison of computed and conventional radiography.

A prospective study was performed comparing computed and conventional radiography for the detection and visibility of cardiovascular devices in intens...
919KB Sizes 0 Downloads 0 Views