Planar (99)Tc(m)-methoxyisobutylisonitrile ((99)Tc(m)-MIBI) scintimammography has been used for several years to detect breast cancer tumours, but with low sensitivity for small lesions. Results of tomoscintimammography studies have not been conclusive. We conducted a phantom study to compare the detection of small-sized tumours with planar versus tomoscintigraphic images. We used a data spectrum anthropomorphic fillable breast phantom with two 9.8 mm and 12.4 mm spheres superficially or deep in the breast compartment with sphere/breast activity ratios varying from 3 to 6. We acquired planar and 180 degrees tomoscintigraphic images in each configuration using a double head standard gamma camera. In certain cases we varied different parameters (64x64 matrix or 360 degrees rotation) in a second series of tomoscintigraphic acquisitions. We simultaneously used filtered back-projection reconstruction (FBP) and iterative reconstruction (IR). Planar images were shown by the sphere in 10 out of 25 cases. Tomoscintigraphic images were shown by the sphere in nine out of 25 cases with FBP and in 18 out of 25 with IR. There was a significant difference between IR and FBP (P<0.01) and between planar and IR images (P<0.01), but no significant difference between planar and IR images. The noise/signal ratio was lower with planar images than with the two types of reconstruction (P<0.05) but was not significantly different between the two types of reconstruction. Contrast was lower on planar images than on the two types of reconstruction (P<0.05) and was also better on IR than on FBP images (P<0.05). Granularity was lower for planar images than for reconstruction images (P<0.01) and also lower for IR than for FBP (P<0.01). The tomoscintigraphic reconstructions acquired with a 64x64 matrix were only positive in four out of 10 cases, while they were positive in nine out of 10 with a 128x128 matrix. We concluded that, in this phantom study, tomoscintimammography with IR provides a significant improvement in the detection of small-sized breast tumours compared with planar images. In addition, for tomoscintigraphic images, a 128x128 matrix is preferable to a 64x64 matrix. Those results have, of course, to be confirmed in vivo in a large population of patients with small-sized breast lesions.