In order to assess the accuracy of CT in predicting the resectability of lung cancer, a prospective study was performed on 96 patients undergoing thoracotomy. The tumors were classified preoperatively according to the TNM classification and the new international staging system for lung cancer, and scored as being resectable by lobectomy or pulmectomy, potentially resectable by lobectomy or pulmectomy, or nonresectable. Of the tumors predicted to be resectable or potentially resectable, 86.6% and 63% were radically resected, respectively, and the need for lobectomy versus pulmectomy was correctly estimated in 81.3% of them. The insufficiency of CT for defining lymph node metastases and infiltrative tumor growth was considered a marked disadvantage of the method.