Endovascular aneurysm repair (EVAR) is currently the treatment of choice for most patients with an abdominal aortic aneurysm (AAA). Real-time guidewire tracking in 2D X-ray fluoroscopy can greatly assist physicians in EVAR treatment. Nevertheless, this task is often accompanied by the challenges of the noisy background of X-ray fluoroscopy and the elongated deformable structure of the guidewire. In this work, a novel lightweight network architecture called Fast Attention Segmentation Network is proposed for real-time guidewire tracking. The novel network combines the advantages of the attention mechanism, the pre-trained lightweight components, and reinforced focal loss to effectively address the problem of extreme foreground-background class imbalance and misclassified examples. Experiment results on clinical 2D X-ray image sequences of 30 patients demonstrate that the proposed approach can achieve state-of-the-art performance. To the best of our knowledge, this is the first approach capable of real-time segmenting and tracking guidewire in EVAR treatment.