Portal de Periódicos da CAPES

Solitary Pulmonary Lesions

1967; Radiological Society of North America; Volume: 89; Issue: 4 Linguagem: Inglês

10.1148/89.4.605

1527-1315

Arch W. Templeton, Carl Jansen, James L. Lehr, Robert Hufft,

Superconducting Materials and Applications

The Roentgenographic differential diagnosis of solitary lung nodules is a difficult problem (3). In a significant group of patients the roentgen findings are virtually nonspecific, while in other cases the radiologist feels confident in making a specific diagnosis. Thus, the radiologist faces a twofold dilemma. Not only must he decide what disease process is most likely responsible for the roentgenographic findings, but he must also determine how confident he can be about his diagnosis in each case. To resolve the type of diagnostic problem discussed above, Ledley and Lusted (4) proposed a mathematical approach based on Bayes' theorem of prior probabilities. With the aid of the digital computer, this method has been applied to the differential diagnosis of many disease groups, including congenital heart lesions (9), primary bone tumors (6), thyroid dysfunctions (2), and gastric ulcers (10). All of these studies have employed not only Bayes' theorem but also the assumption that the signs and symptoms used to make a diagnosis may be treated as independent variables. This assumption has been shown to be poorly justified and to result in unreliable diagnostic probabilities, at least in certain circumstances (9). Therefore, a modified approach, which allows for linear relationships between variables, has been developed (1). In this paper roentgenographic data collected from 242 histologically proved cases of solitary lung lesions are presented. These data are then analyzed both by the"standard" method assuming independence of roentgenographic findings and by the"new" linear method. The diagnostic results obtained by these two methods are discussed and compared with respect to their accuracy and reliability. The applicability of statistical methods to roentgenographic diagnosis of solitary lung nodules is discussed in the light of these findings. Material and Methods The following nine solitary lung lesions have been evaluated: nonspecific granuloma, primary adenocarcinoma, epidermoid carcinoma, undifferentiated carcinoma, abscess, infarction, hamartoma, chronic pneumonitis, and histoplasmoma. The patient's age, sex, and smoking history were recorded on a work sheet, using information taken from the patient's chart. Pos-tero-anterior and lateral chest roentgenograms exposed within forty-eight hours of definitive biopsy, surgery, or autopsy were evaluated in all patients. Eighteen roentgenographic findings were initially evaluated for each of the 242 cases. After the roentgenographic information had been recorded on the work sheet, the information was transferred to data punch cards, from which a frequency distribution table and a prior probability matrix were compiled by the computer.