Detection of atrial fibrillation in implanted devices by automatic scanning of sensing threshold and cycle length

A new approach is presented for computationally efficient detection of atrial fibrillation in which atrial rate is measured as the pacemaker sensing and cycle length threshold are independently varied. This extends the "threshold scanning" rate curve into a two dimensional integer function. A similarity metric is used to compare the shape of this function to one typical of atrial fibrillation. The algorithm correctly identified 26 of 29 episodes of atrial fibrillation and incorrectly identified as atrial fibrillation 3 of 83 episodes of other rhythms, for a sensitivity of 90% and specificity 96%. The integer character of the discriminating function as well as the simplicity of the implementation make this algorithm suitable for implantable devices.