US Patent 9168004 Machine learning in determining catheter electrode contact

Cardiac catheterization is carried out by memorizing a designation of a contact state between an electrode of the probe and the heart wall as one of an in-contact state and an out-of-contact state, and making a series of determinations of an impedance phase angle of an electrical current passing through the electrode and another electrode, identifying maximum and minimum phase angles in the series, and defining a binary classifier adaptively as midway between the extremes. A test value is compared to the classifier as adjusted by a hysteresis factor, and a change in the contact state is reported when the test value exceeds or falls below the adjusted classifier.