OBJECTIVES: Recently, diffusion-weighted MR imaging (DWI) for the whole body has become available for clinical use, as has been previously used for the central nervous system. Favorable results have been reported using this imaging system to differentiate between benign and malignant lesions in some organs, and to correlate with the degree of cell differentiation in lung cancer. The purpose of this study was to assess the role of DWI for predicting tumor invasiveness of non-small cell lung cancers (NSCLC), especially for clinical stage IA patients. METHODS: From January 2006 to September 2007, preoperative DWI and 18F-FDG-PET/CT were performed on 41 patients with clinical stage IA NSCLC who had undergone curative operations. Lung cancers that exhibited nodal, lymphovascular or pleural invasion were defined as invasive lung cancers. Nodules with strong dark signal, as observed by DWI in spinal cords, were defined as DWI-positive. We analyzed the associations between the pathological findings and the following preoperative clinical factors: age, gender, smoking history, preoperative CEA levels (<5.0 or >/=5.0ng/ml), preoperative tumor size, SUV max on PET/CT (<5.0 or >/=5.0) and DWI (positive or negative). RESULTS: A total of 15 lesions (37%) were assessed as DWI-positive and 26 lesions (63%) were DWI-negative. Univariate analyses showed positive correlations for development of invasive cancer with the preoperative CEA level (p=0.049), SUV max (p=0.001) and DWI (p<0.001). Multivariate analysis showed that DWI (p=0.005) was an independent predictive factor for tumor invasiveness. CONCLUSION: Our results suggest that DWI might be a useful method for predicting tumor invasiveness for clinical stage IA NSCLC.