Background and purpose: The standard linear source breast implant of our institution was compared with alternative linear source implant geometries and a stepping source implant, to evaluate the possibility of minimizing the treated volume. Normalization to a higher isodose than the conventional 85% of the mean central dose (MCD) was investigated for the stepping source implant to reduce the thickness of the treated volume and to increase dose uniformity. The purpose of this study was to develop an implant geometry yielding a high conformity and a more uniform dose distribution over the target volume. Materials and methods: The dose distributions of four implant geometries were compared for a planning target volume (PTV) of 48 cm3. Implants #1 (standard) and #2 had linear sources arranged in a triangular pattern of equal lengths and lengths adapted to the shape of the PTV. Implants #3 and #4 were squared pattern arranged implants with linear sources and a stepping source with geometric optimized dwell times. The active lengths were adapted to the shape of the PTV. Using implant #4 for PTVs of different volumes, the reference dose (RD) was normalized to 85 and 91% of the MCD. Results: Comparing implants #2, #3, and #4 with #1, the treated volume (V(100)) encompassed by the reference isodose was reduced by 22, 35, and 37%, respectively. The volumes receiving a dose of at least 125% (V(125)) of the reference dose was reduced by 16, 30, and 30%, respectively. The conformation number increased being 0.30, 0.39, 0.47, and 0.48 for implants #1, #2, #3, and #4, respectively. The average reduction of V(125) when the dose was normalized to 91% compared with 85% of the MCD was 18%. Conclusions: A conformal treatment to a PTV could be best achieved with a geometrically optimized stepping source plan with needles arranged in a squared pattern. Reduction of high dose volumes within the implant was obtained by normalizing the RD to 91% instead of 85% of the MCD. © 2001 Elsevier Science Ireland Ltd.