The authors have developed magnetic resonance imaging-guided breast biopsy robot (MRSON), which is an automated magnetic resonance (MR) guided breast needle biopsy robot that works in an intragantry space with (3+1) degrees of freedom. This robot utilizes a bendable superelastic Nitinol biopsy needle. The MRSON’s needle path is normally parallel to theX-axis of the Cartesian coordinates. However, in some cases, parallel biopsy needle paths are not feasible based on obstacles that are present in front of the target tumor (i.e., compressor grid walls, visible blood vessels, or nontarget tumors). Therefore, a deflected needle path planning method for MRSON was developed in this article to avoid obstacles and accurately perform biopsies. The novelty of the proposed method lies in using a grid with 20×12 holes, through which the robot introduces the biopsy needle and actively deflects the needle against the grid to generate a deflected needle path leading to the target lesion without requiring additional actuators. To calculate the deflection amount and needle insertion depth, the deflected needle is modeled by an$\epsilon$-second-order polynomial, and feasible needle paths are determined through geometry-based analysis. A path-ranking algorithm is then applied to select the best path among all feasible needle paths. Furthermore, to handle with any target lesion movement and coordinate misalignment between the robot and MR image coordinates, a Jacobian-based needle path adjustment method is proposed and verified through numerical simulations. The proposed deflected needle path planning method was verified through 28 in vitro accuracy test and four in vitro validation tests with a linear obstacle, which revealed a root-mean-square targeting error of 2.3 mm.