Tumor delineation is a critical aspect in radiotherapy planning treatment and is usually performed on the anatomical images of a computed tomography (CT) scan. Recently, for non-small cell lung cancer, it has been recommended to use functional Positron Emission Tomography (PET) images to take into account the target biological characteristics. However, today, there is no satisfactory segmentation technique for PET images in clinical applications. In the present study, a solution of this problem is proposed. Methods: The optimizations of tumor delineation consisted primarily on the thresholds adjustment directly from patients, rather than phantoms. The development and the validation of this adjustment were done by comparing segmented lesions on PET images with two different gold standards: measurements performed on CT images of the selected lesions and histological measurements of surgically removed tumors. Results: For lesions greater than 20 mm, our segmentation technique showed very good estimation of histological measurements (mean difference diameter between measured and calculated data equal to +1.5 ± 8.4 %) and an acceptable estimation of CT measurements. For lesions smaller or equal to 20 mm, the method showed a large gap with the measurements derived from histological or CT data. Conclusion: This novel segmentation technique shows very high accuracy for the lesions of large axes between 2 and 4.5 cm. Nevertheless, it does not correctly evaluate smaller lesions probably because of the partial volume effect
