The purport of this study was to evaluate the part of high-resolution computed tomography (HRCT) in the clinical diagnosis of diffuse infiltrative lung disease (DILD).

The purport of this study was to evaluate the part of high-resolution computed tomography (HRCT) in the clinical diagnosis of diffuse infiltrative lung disease (DILD). Diagnostic accuracy was compared using as well-as; not only-but also; not only-but; not alone-but chest radiography and HRCT. single in kind hundred thirty-four cases of DILD, representing 21 different diseases, were pitch uponed for study, and the disease state was confirmed either histologically or microbiologically. The HRCT images and chest radiographs, available in all cases, were reviewed separately and in random order according to 20 physicians who were provided solitary with information on each patient's age and sex Overall, a correct first-choice diagnosis was made in 38 percent using radiographs and in 46 percent using HRCT images (p [les than] 001) The correct diagnosis was among the top three choices in 49 percent when chest radiographs were used, and in 59 percent when HRCT images were viewed (p [les than] 001) The correct first-choice diagnosis increased remarkably when the HRCT was used in usual interstitial pneumonia, sarcoidosis, alveolar proteinosis, bronchiolitis obliterans organizing pneumonia, hypersensitivity pneumonitis, and pulmonary lymphangiomyomatosis. High-resolution comput tomography was confirmed to be superior to conventional radiography in the accurate diagnosis of DILD in clinical practice.

In general, an accurate diagnosis of diffuse infiltrative lung disease (DILD) using chest radiography is remarkably difficult.[1,2] High-resolution computed tomographic (HRCT) features of a variety of DILDs have been reported in the last decade,[3-7] and more lately the diagnostic value of HRCT for DILD has been recognized especially by means of radiologists. High-resolution CT visualizes parts or the whole of secondary pulmonary lobules[8-14] However, the goal of HRCT is accurate diagnosis of the disease. To our knowledge, solitary four studies regarding the diagnostic accuracy of chest radiography and CT have been published.[15-18] Mathieson et al[15] reported that, in 118 consecutive patients with chronic DILD, CT scan interpretations at three observers were more accurate than interpretations using chest radiographs. They subsequently commended that CT scanning should herald lung biopsy.[15] Similarly, in the other three studies, utilizing 48 preferableed patients with chronic DILD (Bergin et al[16]), 140 consecutive patients with chronic DILD (Grenier et al[17]), and 86 patients with chronic DILD mixed with 14 normal make subordinates (Padley et al[18]), the CT was determined to be superior to the conventional radiograph in the diagnosis of chronic DILD. These studies, however, were limited in that diagnosis was performed by way of two or three experienced radiologists and certain often met with disorders were not included. Thus, these studies are not representative of clinical practice. From the perspective of cost-effectiveness and radiation prospect HRCT should not be performed in all patients with DILD. The usefulness of HRCT in clinical practice and its value in the diagnosis and treatment of patients with DILD remain unclear.

To evaluate the character of HRCT in a more clinical setting, 134 cases of DILD (21 different disease), confirmed histologically or microbiologically, were choiceed for analysis. The diagnostic accuracy of the two chest radiography and HRCT was evaluated by dint of 20 physicians.

METHODS

Patients

A total of 134 patients with DILD were examined. The investigation encompassed 21 different pulmonary disease, and included 82 male and 52 female patients with a mean age of 491 years (range, 19 to 84 years). A specific diagnosis for each patient was confirmed histologically or microbiologically. the couple chest radiographs and HRCT were obtained between 1982 and 1990 at the Chest Disease Research Institute, Kyoto University. The mean time interval between chest radiography and CT scanning was 121 days. The CT scans were performed in succession a scanner (GE 8800, General Electric, Milwaukee) with 5-mm collimation as described elsewhere.[12-14,19] rebuilded images for lung parenchyma were obtained by the agency of a high-spatial-resolution algorithm (bone detail algorithm) in all patients. Scans were viewed at a setting appropriate for the pair lung parenchyma (level, -800 HU; width, 1000 HU) and mediastinum (level 0 HU; width, 250 HU) Additional settings were occasionally used.

Twenty Japanese physicians came to Kyoto for 2 days (July 13 and 14 1990) This assemblage consisted of 10 radiologists and 10 chest physicians, each with 5 to 15 years' experience in chest clinics. The HRCT images and chest radiographs, which were available in all cases, were reviewed separately and in random order by way of these 20 observers, who received information onward each patient's age and sex sole The observers had no prior knowledge of the disease representation or frequency. Three possible diagnoses were listed in all cases by the agency of each participant. They were asked to record the rank of confidence in their first-choice diagnosis onward a three-point scale (definite, probable, or possible). Statistical analyses were performed using the [[chi].sub.2] ordeal The percentage of correct diagnoses shows the sum of the correct interpretations on the 20 observers divided on the total number of answers for each disease. The issues using chest radiographs and HRCT images were then compared. If the number of correct answers for a particular disease was les than five, either for chest radiography or HRCT the [[chi].sup.2] experiment was deemed inappropriate and a statistical analysis was not performed. The percentage of first-choice diagnosis that were correct, ie, the positive predictive value, was calculated on the number of true positive answers divided from the number of true positive answers plus false positive answers.

...