Computed tomography and magnetic resonance imaging have respectively decreased the incidence of exploratory thoracotomy by 11.6% to 5.7% then 5.3% after their successive inclusion in the assessment of the operability of lung cancers. Based on a systematic comparison of CT and MRI with exploratory thoracotomy in 111 patients, the yield, sensitivity and positive predictive value of these examinations was assessed in relation to tumour extension to various sites: to the chest wall, for which the sensitivity was poor (38% for CT, 54% for MRI) with a moderate predictive value (71% for CT, 77% for MRI); to the mediastinum with improved sensitivity (69% and 72%) but an uncertain predictive value (61% and 72%). The sensitivity and predictive value were then measured for lymph node involvement: N1: moderate sensitivity (69% and 76%) but a good predictive value (95% and 92%); N2 and N3: good sensitivity (79% and 93%) but a poor predictive value (70% and 66%). Apart from a few particular indications specific to MRI, especially in the apex, left hilum and in contact with the atrium, the efficacy of these two investigations is very similar. However, their lack of specificity means that certain exploratory thoracotomies are still justified in order to assess the operability of the tumour. A surgical or mediastinoscopic anatomical assessment is still necessary for a good classification of lymph node extension of lung tumours.