Complexity Diagnostics in Cardiology: Methods GEORG SGHMIDT and GREGOR E. MORFILL* From the I. Medizinischo Klinik der Technischen Universitat Miinchen, and the *Max-Planck- Institut fur Extraterrestriche Physik, Munich, Germany Introduction In living organisms, there is a general princi- ple that governs behaviors and reactions; they are designed to optimize "survival." In higher organ- isms this self-preservation aspect is maintained by a plethora of specialized sensors, which monitor essential functions, note potentially troublesome deviations from the normal state, transmit this in- formation, and initiate suitahle control mecha- nisms to compensate for this, if necessary. In the cardiovascular system, relevant proper- ties that are monitored continnously are the arte- rial hlood pressure, the oxygen content, the pH value of the blood, the atrial wall tension, etc. These properties are continuous variahles, the cor- responding "signals" can be regarded as ampli- tude modulated analog signals. The sensor respon- sible for monitoring these properties passes the in- formation on to the central nervous system (CNS). It is ahle to do so mainly in a "frequency modu- lated" form via afferent nerve fibers of the auto- nomic nervous system. (The integrals of the neural action potentials carry some information, too.) This implies an analog-to-digital (AD) conversion. In the CNS the incoming information from many such sensors is processed and some "reaction" is arrived at. The details of this particular process are not well understood yet. For our purposes it is sufficient to regard the CNS as a "hlack hox," whose function is to assemble incoming informa- tion, process it. and produce new instructions, which are then transmitted in digital form through efferent fibers of the autonomic nervous system to their destination—the "effectors" in the pe- riphery. Examples of such effectors are the pacemaker Address for reprints: Georg Schmidl. I. Medizinische Klinik der Technischen Universitat Miinchen, Isrnaninger Str, 22, Miinchen 8000. Germany, Fax: 011-49-89-41805162, Received October 6, 1994; accepted October 6. 1994. cells in the sinus node, "ordinary" myocardial cells, or vascular smooth muscle cells. These "translate" the incoming digital instructions back into analog signals (DA conversion) causing the desired reaction. This type of information transfer may be described as analog-digital-analog (ADA) coding (modified, following 1). Properties, such as the peripheral resistance of the hlood vessels and the myocardial contractility, are thus continu- ously adjusted to the momentary requirements of the system. A somewhat more complicated situa- tion is the pacing of the heart frequency. Here, too, we have an ADA coding of the signal transfer. However, the second analog signal, the action po- tential of the cells in the sinus node, is digitized once more. When during phase 4 of the action po- tential a certain threshold potential is reached, the conductivity of the cell membrane to calcium ions is changed drastically. Phase 0 of the next action potential is initiated and the cell begins to depolar- ize. For the neighboring cells the detailed form of the action potential of the pacemaker cells is im- material. The important point is that the threshold potential is reached, w^hich is the essential prereq- uisite for its own depolarization. (The situation is not quite as simple as described here. Some "ana- log properties" are preserved even in the digitized signal.) This second digitization reduces an analog signal to an all-or-nothing type of information. From this point of view, a 24-hour RR interval rfata set is not a fragmentary ECG signal subset, hut pro- vides, in principle, a complete "tracer" of the heart's control mechanisms. Figure 1 shows the 24-hour tachogram of a pa- tient with coronary artery disease (GAD). Clearly, there appear to be complex patterns, some of which are the result of complex control mecha- nisms. The simplest analysis methods (averages and standard deviations) assume that the fluctua- tion patterns are essentially random, which clearly they are not. The more sophisticated Fourier anal- 2336 December 1994, Part I PACE, Vol. 17