Fatigue fracture of small nitinol components commonly used in medical devices is currently dominated by the initiation and/or growth of small cracks at non-metallic inclusions preceding conventional fatigue crack growth. Therefore, an understanding of the threshold and growth of small cracks is critical to inform fatigue performance of devices. In this paper, we conduct rotary bend fatigue experiments of nitinol wire to 2 billion cycles, measure the inclusion from which the crack initiated, and calculate the stress intensity threshold. Inclusion size is compared to the size of a unique feature observable with a scanning electron microscope which appears as a smooth area surrounding the inclusion and only on specimens that fractured after > 10 million cycles. The hypothesis presented is that the smooth feature around the inclusion is the growth of a small crack which continues until it reaches a size large enough to cause conventional fatigue crack growth. Relating inclusion size to that of the smooth feature creates a damage curve that can be written as a function of cycles to fracture. This damage curve may be useful to estimate the critical size of the largest allowable defect based on the design life and applied loading of the nitinol component.