Partial volume effects (PVEs) are well-known consequences of the limited spatial resolution in emission tomography which can affect the quality of PET oncology images both qualitatively and quantitatively and thus impact the diagnostic task. There exists a wide variety of PVE correction methods but they are not all well suited to oncology images. In an attempt to obviate the need for drawing or segmenting ROIs as well as to produce PVE-corrected images, Boussion et al have proposed a novel voxel-wise PVE correction using the iterative Lucy-Richardson deconvolution methodology in combination with a dedicated wavelet-based denoising algorithm (LR-W). The purpose of this study is to evaluate the impact of this PVE correction method on clinical detection performances. This study is based on a series of realistic simulated whole-body FDG images including spherical lesions of calibrated diameters and uptakes. Detection performance is evaluated based on a computer-aided detection system that we are developing for whole-body PET/CT images based either on the linear LDA or the non linear SVM classifiers. Comparison of the detection performances achieved with this PVE correction method is performed visually based on the analysis of the detection maps derived from the CAD system and quantitatively based on a Free-Response Receiver Operating Characteristic (FROC) analysis. The SVM and LDA classifiers lead to contradictory results regarding the impact of PVC on detection performances. According to the SVM classifier, PVC significantly improved detection performances of small and low contrast lesions. Further investigation is required to analyze the results achieved with the LDA which is not sensitive to the LR-W PVE correction.