Many intravascular procedures are prefaced by the placement of a slender wire called a guidewire. Steering these guidewires is met with challenges in controlling the distal end along with the possibility of damaging vessel walls, or even perforation, which can be fatal. To this end, utilizing robotic guidewires can improve steerability and enable force feedback through intrinsic force sensing. Enabling force sensing contains challenges such as discrete sensor placements in continuous structures and non-unique force distributions for a given deflection. In this work, we utilize image feedback and a Cosserat rod model to estimate and localize forces along the body of a micro-scale tendon-driven guidewire robot. This includes additional modeling of friction and hysteresis that is often neglected for force sensing. The model is tested on a variety of notched nitinol tubes under gravity loading with the shape predictions having an average RMSE of 0.46 mm. Utilization of friction and hysteresis models provide shape predictions with an RMSE of 1.22 mm compared to an uncompensated model (RMSE = 1.62 mm) for approximately 180° bends. The methods presented are able to localize forces with an average error of 4.79 mm (5.15% of the length) while force magnitudes are estimated with an average error of 13.03 mN.