The limited-angle acquisition geometry of digital tomosynthesis (DT) provides incomplete projection datasets which leads inevitably to artifacts in reconstructed images. This work presents both an algorithmic and a hardware approach to reduce these artifacts and to improve the image quality. The algorithmic approach introduces a nonlinear weighted backprojector in the simultaneous algebraic reconstruction technique (!SART). The hardware approach is based on an alternative dual-axis outside-the-arc geometry, where the acquisition is done by covering a spherical cap instead of an arc. The performance and artifactreduction ability of the proposed app-roaches are evaluated based on real and simulated data using a three-dimensional reconstruction framework. It is shown that the weighting prevents the formation of stripe artifacts produced by high-density tissues while it preserves the true structures, which belong to the object. While reducing artifacts, this approach is unable to reduce the triangle-like shape distortion in the “blind”-spots. In turn, the outside-the-arc geometry reduces the degree of data incompleteness and the size of the “blind”-spots by capturing more singularities of the object. This reduces artifacts, particularly the shape distortion and results in images with better axial resolution. In practice, such geometry can be implemented without major mechanical modifications of existing tomosynthesis devices e.g. by using an object-tilting platform.