For two years now, a study on intensity modulated radiotherapy (IMRT) has been in progress at the Antoine Lacassagne Hospital Center for Cancer Therapy (in Nice) in collaboration with the University of Nice-Sophia Antipolis. The kind of intensity modulation that was used is the "step and shoot" technique in which the modulated beam is created both by adding andjoining elementary fields. Before carrying out clinical tests, several problems regarding the production of modulated beams has to be mastered. The current developments of our study enable us to dosimetrically produce (in water phantom and in the PMMA phantom) complexmodulated whose segmentation was calculated by one commercial treatment planning system (TPS). Nevertheless, we showed and studied some critical discrepancies between standard clinical calculations and the calculations using field segmentation. We showed that with nonoptimal conditions of segmentation the discrepancies, which are due to the type of algorithm used, could bring about significant errors inside the field of up to 10% of maximum dose. Another point of our study is the quantification and resolution of differences between measurements and calculations due to the internal segmentation of calculated modulated fields and their realization on Linac. Once again, in none optimal conditions of segmentation and inside the field we obtained discrepancies up to 20% of maximum dose between calculations using field segmentation and measurements. That was mainly due to the tongue and groove effect and penumbra phenomena. This study allows us to show that the discrepancies between segmentation calculations and standard clinical calculations should be solved by the use of penumbra models during segmentation calculations. We will introduce both the study and its near-future perspectives.