Purpose: Increase detection sensitivity of breast carcinoma is greatly needed for digital mammography. We show that the tissue‐lesion contrast‐noise ratio (CNR) at a given dose level can be improved by increasing the Rh‐filter thickness used for the Rh‐target in digital mammography. Method and materials: Currently a 25μm‐Rh filtration is the standard in digital mammography for all the modes with a Rh‐target. To test if this standard filter‐thickness is optimal, we studied how the contrast‐noise ratio between breast and infiltrating ductal carcinoma (IDC) of 5mm‐size embedded a 6cm‐thick breast changes with Rh‐filter thickness for Rh‐target. We performed imaging experiments by modifying the filter wheel of a GE Senographe 2000D unit with Rh‐filters of 25μm, 37μm, 50μm, 62μm and 75μm. Before imaging a phantom the x‐ray HVL values at 29 kVp and radiation outputs were measured for each Rh‐filter thickness ranging from 25μm to 75μm. A 50%‐glandular and 50%‐adipose breast phantom of 6cm with an infiltrating ductal carcinoma (IDC)‐simulating insert of 5mm in size was used as the phantom for all the cases. The CNR's between the breast phantom and the IDC‐insert were measured, and average glandular doses were calculated by using a filtration‐dependent x‐ray spectra model and a breast‐dosimetry model based on a validated Monte Carlo simulation. The exposure times were recorded as well. Results: In contrast to conventional wisdom, the tissue‐lesion CNR at a given dose level increases with increasing Rh‐filter thickness from 25μm to 75μm. The measured squared CNR per dose were increased by 7%, 13%, 18% and 21% for 37μm, 50μm, 62μm and 75&μm Rh‐filters compared to that for the standard 25μm Rh‐filter, respectively. Conclusions: Increasing Rh‐filter thickness for Rh‐target from 25μm to 50μm can increase tissue‐lesion squared CNR per dose by 13% with a tolerable increase of exposure duration. © 2005, American Association of Physicists in Medicine. All rights reserved.