An evaluation of different imaging chains in clinical chest radiography

1984, The British Journal of Radiology, 57, 991-995

NOVEMBER

1984

An evaluation of different imaging chains in clinical chest radiography By H. Manninen, M.D., M.Sc, *E. 0. Terho, M.D., M. Wiljasalo, M.D., S. Wiljasalo, M.D., and S. Soimakallio, M.D. Department of Diagnostic Radiology and 'Department of Pulmonary Diseases, Kuopio University Central Hospital, SF-70210 Kuopio, Finland (Received October 1983 and in revised form February 1984)

ABSTRACT

Six imaging techniques in clinical chest radiography have been evaluated: four film-screen combinations in the conventional grid technique and two combinations in the air gap technique. Five parameters characterising the quality of a chest radiograph were evaluated by three radiologists and one chest physician by using a nominal grading scale from —2 to + 2 compared with the standard technique. The quality parameters judged were: the visibility of peripheral lung vessels, lung parenchyme, the pulmonary hilum, and lung structure behind the heart shadow, as well as the visibility of miscellaneous findings of clinical interest. The air gap technique was shown to be superior to the ordinary grid technique. The diagnostic quality of chest radiography does not necessarily deteriorate with the screen speed. However, statistically significant differences were noticed, even between techniques which had equal speed and physical resolution.

Several imaging techniques have been used in chest radiography. Ordinary film-screen techniques with or without a grid, photofluorography and air gap techniques are perhaps the best known. During the last decade high-speed rare-earth screens have been available for chest radiography. At the same time a low radiation dose has increasingly become as important in imaging chains as an image of good diagnostic quality. Physical parameters, such as resolution, contrast sensitivity and radiographic mottle are useful for characterising the quality of radiographic systems, but their relationship to the practical use of the image for clinical diagnosis is not well known. Therefore, when several techniques are compared in a particular task, e.g., in chest radiography, observer performance should be directly measured with clinical radiographs. A wellknown approach for assessing observer performance is to superimpose artificial lesions on clinical radiographs and to use statistical decision theory and receiver operating characteristic (ROC) curves for analysis. Another attempt to measure observer performance of clinical radiographs has used a minified Snellen type Eplate introduced by Soila and Tahti (1965) and further developed by Laasonen (1968). The test plate consists of three contrast zones, and, in the analysis, the smallest 'E' visible in each zone is registered. The test device is attached to the patient during radiography, giving realistic scattering conditions.

These techniques, however, measure observer performance only in one particular radiographic task in each setting: the visibility of an artificial coin lesion or a letter 'E' shape in this case. De Smet et al (1981) used a nominal grading scale in the comparison of film-screen combinations for skeletal radiography. In this technique, the diagnostic quality of the image was evaluated by a quality scale from —3 to +3 compared with a standard technique. This technique evaluates the overall diagnostic quality of the techniques by determining several parameters essential for a clinical radiograph. The present study was undertaken to evaluate different film-screen combinations and especially to compare grid versus air gap techniques in clinical chest radiography by applying a visual grading scale analysis. TECHNIQUES AND METHODS

Imaging chains

The systems examined and their characteristic features are listed in Table I. The techniques for the study were selected from twenty systems previously physically assessed using water phantom and Alderson chest phantom experiments (Manninen et al, 1982, 1984). The relative patient exposure (the system speed) is expressed relative to the 3M Trimax T4/Trimax XD film-screen combination as used in the grid technique. The equivalent pass-band (Ne) as a measure of physical resolution, the inherent subject contrast (Cs) as a measure of contrast sensitivity, and the visual detectability (Z)va) as a measure of observer performance in the Alderson chest phantom investigations have been given as a percentage of the best technique. Rare-earth screens were used instead of calcium tungstate screens, since their higher X-ray absorption and better conversion efficiency, X rays to light, allow a higher speed without loss in image quality. The three types of rareearth screens evaluated were the mainly blue-emitting Cawo Se-series, the mainly green-emitting Trimax-Tseries (earlier Alpha-series) and the Kodak Lanex Regular with a Par-speed film, all recommended for chest radiography by the manufacturers. Special note was taken of the effect of screen speed on observer performance, the phosphor material being unchanged. The conventional grid technique was compared with the

991

VOL.

57, No. 683 H. Manninen, E. Terho, M. Wiljasalo, S. Wiljasalo and S. Soimakallio TABLE I

THE CHEST IMAGING CHAINS AND THEIR CHARACTERISTIC FEATURES: SYSTEM SPEED, EQUIVALENT PASSBAND (Nt), INHERENT SUBJECT CONTRAST ( C s ) AND VISUAL DETECTABILITY (Z) v a ) IN ALDERSON CHEST PHANTOM INVESTIGATIONS. Nc, Cs AND Z>va HAVE BEEN NORMALISED AS A PERCENTAGE OF THE BEST TECHNIQUE (MANNINEN ET AL, 1 9 8 2 , 1 9 8 4 )

System

System characteristic

Grid technique

Focus-to-film distance = 200 cm, 10:1 stationary grid, 40 lines/cm

Technique symbol Film

Focus-to-film distance = 350 cm, 18 cm air gap

System speed

%

%

2)va %

Se2g

Fuji RX NIF

Cawo Se2

1.6

85

92

94

Se6g

Fuji RX NIF

Cawo Se6

3.5

79

95

90

T8 g

g

3M Trimax XD 3M Trimax XD

Trimax T4 Trimax T8

1.0 1.6

92 86

100 99

99 99

T4a

3M Trimax XD

Trimax T4

2.6

100

70

100

LR a

Kodak OrthoG

Lanex Regular

7.2

85

66

86

T 4

Air gap technique

Screen

air gap technique using an 18 cm air space between the patient and the film-screen cassette. A three-phase, six-pulse generator (Philips DA 1000) with Rotalix SRO 100 tube and focal spot of 1.64 x 2.32 mm was used. The tube voltage was 120 kV. The films were developed in an automated processor at 33°C and developing conditions were controlled with test films (Agfa Gevaert System Control RP 79001 A). Patients Chest radiographs, in PA projection, upright position, were taken of 40 volunteers using each of the six techniques on each volunteer. The image field was strictly blended to the thorax size of each patient. The volunteers were mainly patients with malign neoplasms or chronic heart diseases (19 men, 21 women, with mean age of 69 years) (range 54-80 years). Patient exposures were adjusted so as to obtain approximately equal film densities from one technique to another. The average patient exposures were measured with an Alderson chest phantom by using a dosemeter (Ionex 2500/3) and a 0.6 cm 3 ionisation chamber. Average exposure to the skin from the six radiographs taken was 95 mR and did not exceed 170 mR, even for the biggest persons, assuming two extra radiographs taken to obtain images of adequate quality. Image analysis Before analysis, films of inadequate quality were excluded from the study. The most common reasons for exclusion were inappropriate phase of inspiration, incorrect projection, artefacts due to film development, and differences in film density greater than 0.2 between the techniques in the film series of each patient. The film density was measured at a standard location (third intercostal space of the left lateral lung field) using the 1 mm diameter aperture of a Repromaster RM 21 densitometer. The film series for the analysis consisted of 3-6 images/patient. The combination of Se2 screen with Fuji RX NIF film used with a grid (Se2g) was

included in all 40 of the image series. Other techniques were included in 31 to 38 of the series. In the visual analysis, the films of each patient were graded by three radiologists and one chest physician on a conventional viewing box. The room illumination was dim and kept constant. The films were free of identifying marks and assigned by code numbers. The reading time and distance were not restricted. For each film series a standard technique was selected with which the other techniques were compared. All techniques in turn were used as the standard, but for the data analysis the scores were calculated with reference to the Se2/Fuji RX NIF combination of the grid technique (Se2g). For each film series, five rectangular regions of interest (side length 4-8 cm) were masked with a black film. In each region one anatomical parameter was assessed. The five parameters evaluated were: visibility of peripheral pulmonary vessels of the upper lobe, visibility of parenchymal structure of the lower lobe of the left lung, visibility of the lung structure behind the heart shadow, visibility of pulmonary hilum, and finally the visibility of miscellaneous findings of clinical interest, one on each film. These findings included 13 pleural adhesions and calcifications, 8 metastatic lung nodules, 8 prominent pulmonary arteries, 5 infectious infiltrations, 3 osteophytes, 2 atelectases, and in one case a pericardial cyst. In the visual assessment, the observers graded the quality of each film compared with the standard technique by using an arbitrary scale from —2 to +2, the standard system being graded as 0. In the scale + 2 — clearly better, + 1 = slightly better, 0 = equal, — 1 = slightly worse and — 2 = clearly worse compared with the standard. A preliminary analysis using the films of ten patients had revealed that a scale from — 3 to 4-3 used by De Smet et al (1981) was too wide for this kind of clinical radiograph. To test the reproducibility of the grading, each observer graded the films of five patients twice and the statistic K (Cohen, 1960; Fleiss et al, 1969) for inter-observer agreement for

992

NOVEMBER

1984

Different imaging chains in clinical chest radiography TABLE II

TABLE III

COMBINED MEAN FILM SCORES FOR EACH TECHNIQUE, AVERAGED OVER ALL OBSERVERS, THE NUMBER OF OBSERVATIONS FOR ALL OBSERVERS AND THE SIGNIFICANCE OF THE DIFFERENCES

THE COMBINED MEAN FILM SCORES OF EACH TECHNIQUE GIVEN BY

Technique Combined mean film score Number of observations

T4_

T4,

T8O

LRa

Se6o

0.42J 0.11§

0.00

620

640

800

700

620

Combined mean film score of each technique Observer T4a T4O T8B LRa Se6o Se2o

Se2o

0.76* 0.55| 0.48 760

INDIVIDUAL OBSERVERS

M.W. S.W. S.S. E.T.

1.07 0.91 0.63 0.43

0.84 0.68 0.43 0.25

0.73 0.69 0.32 0.20

0.46 0.41 0.48 0.33

0.09 0.17 0.10 0.07

0.00 0.00 0.00 0.00

For the meaning of the symbols refer to Table I.

*T4 a T4 g :/>< 0.001 fT4g LRa: p < 0.05 JLR a Se6 g :/7< 0.001 §Se6g Se2g: p