[Differential diagnosis of solitary pulmonary nodules by computerized statistical analysis using CT number]

There have been previous reports on the differentiation of benign pulmonary nodules from malignant ones using CT number. However, most of the nodules were without calcification, and diagnosis could be made only in limited cases. In the present study we established new multiple variables such as relative high density over the 75th percentile, inner gradient number (inner irregular rate), edge gradient number (edge irregular rate) and increased area rate (IAR). They are automatically calculated from CT data by computer. These variables are transformed into relative values, enabling comparisons of nodules without a standard reference phantom. With IAR, pulmonary nodules were classified into three types: type 1 (IAR 1.0-1.4), a high-density solid mass with well-defined margin; type 2 (IAR 1.4-2.0) between type 1 and type 3; and type 3 (IAR > 2.0) a low-density infiltrated mass with ill-defined margin. We studied 52 cases with solitary pulmonary nodules, whose diagnoses were confirmed either histologically or on follow-up. There were 34 benign nodules and 18 primary lung cancers. Using step-wise multiple regression with three variables, 42 of 52 cases were correctly diagnosed. Twenty-four of 28 type 1 cases and all 14 of type 2 cases were correctly diagnosed. This computer analysis is considered to be useful in distinguishing between benign and malignant solitary pulmonary nodules.