We analyzed 7 years of filtregrams data (150000 photograms and magnetograms) of the SOHO/MDI experiment. We used the maximum possible sampling compatible with full frame recording, carefully avoiding any suspicious filtregram. Going further than the previous analysis of Emilio et al. (Ap. J. 543, 2000) and Kuhn et al. (Ap. J. 613, 2004) we better corrected for changes in optical aberrations and, along Turmon et al. (Ap. J., 568, 396, 2002), we reduced radius measurement errors by identifying active regions from magnetograms and by avoiding radius measurements herein. We found that, within the limit of our noise level uncertainties (8 to 9 mas), the solar diameter is compatible with constancy over the half cycle investigated. Our results confirm the reanalysis of the 7 years of MDI data of Antia (Ap. J. 590, 2003), with a completely different method since using the ultra-precise frequency variation of the f-modes (fundamental modes linked to the diameter), who found (carefully removing the yearly Earth induced variations and avoiding the SOHO data gap of 1999) that the diameter is constant over the half solar cycle (radius variation are less than 0.6 km, 0.8 mas -nothing over noise level). We can conclude, along Antia, that: "If a careful analysis is performed, then it turns out that there is no evidence for any variation in the solar radius." There were no theoretical reasons for large solar radius variations and there is no observational evidence for them with consistent space observations made with 3 different approaches.