Deceleration Capacity of Heart Rate Methods & Applications PDF

Deceleration Capacity of Heart Rate: Methods and Applications George Manis Dept. of Computer Science and Engineering, University of Ioannina Ioannina, Greece  The Elec...

Deceleration Capacity of Heart Rate: Methods and Applications George Manis Dept. of Computer Science and Engineering, University of Ioannina Ioannina, Greece  The Electrocardiogram (ECG) The electrocardiogram (ECG) is an interpretation of the electrical activity of the heart over a period of time  Heart Rate Signals The peaks of the ECG correspond to heartbeats  Heart Rate Signals 887ms 852ms 831ms 830ms 902ms 852ms 812ms The heart rate signal consists of the time intervals between successive heartbeats Please note that the duration between successive intervals varies and that this is a physiological process  Heart Rate Signal 1000 RR intervals (in msec) 900 800 0 50 100 150 200 250 300 Beats An example of a heart rate signal (recorded just before the subject waked up)  Accelerations and Decelerations The heartbeat time series consists of accelerations and decelerations The heart accelerates the rhythm when the time interval between two successive beats is shortest than the preceding interval The heart decelerates the rhythm when the time interval between two successive beats is longer than the preceding interval 1000 RR intervals (in msec) 900 800 0 50 100 150 200 250 300 Beats  Sympathetic and Parasympathetic Influence The sympathetic and the parasympathetic (vagal) systems constitutes the autonomic nervous system The sympathetic nervous system is related with fast reactions and emerging situations (fight or flight) The parasympathetic nervous system is related with normal autonomic functions of body (rest and digest)  Sympathetic and Parasympathetic Influence The two systems interact with each other and help maintain the homeostasis of the organization The sympathetic system accelerates the heart rhythm The parasympathetic system decelerates the heart rhythm  Sympathetic and Parasympathetic Factors To distinguish between vagal and sympathetic factors that affect heart-rate variability, a signal processing algorithm to separately characterize deceleration and acceleration of heart rate has been proposed* It expresses the capability/capacity of the heart to decelerate/accelerate its rhythm The term proposed for this metric is Deceleration Capacity (DC) and Acceleration Capacity (AC) of heart rate respectively *Axel Bauer, Jan W Kantelhardt, Petra Barthel, Raphael Schneider, Timo Maekikallio, Kurt Ulm, Katerina Hnatkova, Albert Schoemig, Heikki Huikuri, Armin Bunde, Marek Malik, Georg Schmidt, “ Deceleration capacity of heart rate as a predictor of mortality after myocardial infarction: cohort study”, Lancet 2006; 367: 1674–81  Deceleration – Acceleration Capacity of Heart Rate for the computation of Deceleration/Acceleration Capacity a mathematical method based on Phase Rectification Signal Average (PRSA) has been proposed without loss of generality we will present the computation of Deceleration Capacity the algorithm is described in an excellent way in the paper given below* a description of the algorithm follows *Axel Bauer, Jan W Kantelhardt, Petra Barthel, Raphael Schneider, Timo Maekikallio, Kurt Ulm, Katerina Hnatkova, Albert Schoemig, Heikki Huikuri, Armin Bunde, Marek Malik, Georg Schmidt, “ Deceleration capacity of heart rate as a predictor of mortality after myocardial infarction: cohort study”, Lancet 2006; 367: 1674–81  Anchor Points for the computation of Deceleration Capacity (DC) the heartbeat intervals longer than the preceding interval are identified as anchors for the computation of Acceleration Capacity (AC) the heartbeat intervals shorter than the preceding interval are identified as anchors approximately 45000 of 100000 RR intervals become anchors in a typical 24-h Holter recording.  Anchor Points to suppress errors due to artifacts, RR interval prolongations (or shortenings for AC computation) of more than 5% are excluded  Definitions of Segments segments of interval data around the anchors are selected all segments have the same size segments that surround adjacent anchors can overlap.  Definitions of Segments X segments of size 4 are shown in this figure two RR intervals before and one after an anchor participate in the computation of DC  Phase Rectification Segments are aligned at the anchors  Phase Rectification anchors Segments are aligned at the anchors  Phase Rectification before the anchors we meet only decelerations Phase Rectification Signal Average The PRSA signal is obtained by averaging the signals within the aligned segments  Phase Rectification Signal Average x(0) is the average of the RR intervals at all anchors x(1) and x(–1) are the averages of the RR intervals immediatel y following and preceding the anchors etc x(-2) x(-1) x(0) x(1)  Quantification of Deceleration Capacity x(-2) x(-1) x(0) x(1) x(1) + x(0) - x(- the quantity characterizes 1) - x(-2) the average capacity of DC = the heart to decelerate -------------------------------- the cardiac rhythm from --- one beat to the next. 4  A Web Presentation an very illustrative presentation can be found in the following link from the Technische Universität München http://www.librasch.org/prsa/en/ I would like to thank Prof. Georg Schmidt and his research group for the informative presentation  Predictor of Mortality after Myocardial Infarction diminished deceleration-related modulation of heart rate is an important prognostic marker deceleration capacity is a better predictor of risk than left- ventricular ejection fraction (LVEF) and standard deviation of normal-to-normal intervals (SDNN).  Predictor of Mortality after Myocardial Infarction some background info from wikipedia  Predictor of Mortality after Myocardial Infarction some background info for completeness from wikipedia  Predictor of Mortality after Myocardial Infarction blindly validated the prognostic power of deceleration capacity in post-infarction populations – in Munich (n=1455) – in London, UK (n=656), – and Oulu, Finland (n=600) tested the hypotheses by assessment of the area under the receiver operator characteristics curve (AUC). Munich London Oulu DC 0.77 0.80 0.74 AC 0.61 LVEF 0.70 0.67 0.60 SDNN 0.68 0.69 0.64 Modifications on the Method  Maximum Derivative during Transient State the index is calculated by the maximum derivative during transient state  Maximum Derivative during Transient State From each ascending series of time intervals, the one with the maximum slope is selected  Maximum Derivative during Transient State the mean of these slopes gives the index it only takes into account the steepest anchor interval since it corresponds to the strongest vagal stimulus y Δy  x Δx  Maximum Derivative during Transient State the modified index was compared to DC subjects were classified as the control group and athlete group the area under receiver operating characteristic was used for comparison purposes using 10,000 bootstraps. DC modif. index Control 11.80 17.84 Athletes 25.98 45.62 P-value < 0.01 < 0.01 AUC 0.70±0.12 0.96±0.04 Conclusions: The modified index discriminates successfully athletes from control subjects  Beat to Beat Deceleration Capacity Deceleration Capacity computed from successive differences only (beat to beat)  Beat to Beat Deceleration Capacity x(-4) x(-2) x(-1) x(0) x(1) x(3) x(-3) x(2) … + x(1) + x(0) - x(-1) - x(-2) + … DC = -------------------------------------- ----- the general description of PRSA does not limit the computation N to four beats only four beats have been selected for the experiments since it is not a large, nor a small number  Beat to Beat Deceleration Capacity x(-4) x(-2) x(-1) x(0) x(1) x(3) x(-3) x(2) … + x(1) + x(0) - x(-1) - x(-2) + … DC = -------------------------------------- ----- if we cross out everything that has to do with non-successive N=2 beats we have the beat to beat Deceleration Capacity (BBDC)  Beat to Beat Deceleration Capacity x(-1) x(0) x(0) - x(-1) DC = --------------- since x(0) is always larger than x(-1), DC2is always positive we will call this quantity Beat to Beat Deceleration Capacity (BBDC)  The Sign of Deceleration Capacity A different approach on the computations which does not allow negative values for DC and applies more strict rules in filtering  The Sign of Deceleration Capacity DC can be Negative x(0)  x(1)  x( 1)  x( 2) DC  4  The Sign of Deceleration Capacity But what is the meaning of a negative DC? A negative DC means acceleration In other words we study the Deceleration Capacity in parts of the signal in which the heart mostly accelerates We tried to avoid this paradox in the computations by introducing a different definition of where the heart decelerates, based on four and not two beats  The Sign of Deceleration Capacity for the series of heartbeats: we define a series of vectors: and for each vector we compute: where xi(j) is the jth element of the vector xi  The Sign of Deceleration Capacity each vector is characterized as acceleration or deceleration according to the sign of acdci we compute the averaged vectors and and DC and AC (we will call is DCsgn and ACsgn) is given by:  Back to Filtering please remember that in order to suppress errors due to artifacts, RR interval prolongations (or shortenings for AC computation) of more than 5% are excluded  Back to Filtering However, segments considered as possible artifacts keep participating in computations  Excluding more in Filtering We want to totally exclude from computations every RR that does not pass the 5% filter we exclude from computations all vectors xi for which one of the following holds: and consider them as invalid:  Excluding more in Filtering for completeness we show how the averaged vectors are computed taking into account the filtering as well: and, of course, AC and DC are given again by the formulas:  Experimental Results – The Fantasia Dataset Twenty young (21 - 34 years old) and twenty elderly (68 - 85 years old) healthy subjects 120 minutes of recording approximately Each subgroup of subjects includes equal numbers of men and women. All subjects remained in a resting state in sinus rhythm while watching the movie Fantasia (Disney, 1940) to help maintain wakefulness. The continuous ECG, respiration, and (where available) blood pressure signals were digitized at 250 Hz. Each heartbeat was annotated using an automated arrhythmia detection algorithm and each beat annotation was verified by visual inspection.  The Fantasia Dataset available on the internet at http:www.physionet.org References and Acknowledgements Mean Values Statistical Significance Measures  Discussion on the Results BBAC presented better discrimination power for the specific dataset The two new methods seem to perform better than the original one However, each method reveals different information and all three methods must be used in a complementary way  All Methods reveal Complementary Information The original method examines how preceding and succeeding heartbeat intervals follow an acceleration/deceleration The Beat to Beat method examines what happens per beat basis The DCsgn method investigates deceleration/accelerations in a longer term basis, and given that a series of segments is deceleration/acceleration, how well the heart decelerates/accelerates in this period The Maximum Derivative method examines the rapid changes  DC is reduced before an NSVT episode Computing in Cardiology, Zaragoza, Spain, September 2013 study the differences between the behavior of Deceleration Capacity before the onset of NSVT episode and in the rest of the signal Deceleration Capacity is reduced before NSVT episodes  Non-Sustained Ventricular Tachycardia Ventricular tachycardia is a type of tachycardia, or a rapid heart beat that arises from improper electrical activity of the heart presenting as a rapid heart rhythm, that starts in the bottom chambers of the heart, called the ventricles. The ventricles are the main pumping chambers of the heart. If the rhythm lasts less than 30 seconds, it is known as a non- sustained ventricular tachycardia  DC is reduced before an NSVT episode Two groups of heart failure subjects that presented Non Sustained Ventricular Tachycardia Episodes were examined – low arrhythmia risk for sudden cardiac death – high arrhythmia risk for sudden cardiac death not statistically different in basic clinical characteristics 16 months of follow up we studied Deceleration Capacity before those episodes DC is reduced before an NSVT episode Dataset  DC is reduced before an NSVT episode we studied 3 hours before the episode we broke these 3 hours into 20 minutes overlapping windows the time distance between successive windows was 5 minutes we computed DC for each one of the windows we took the average values we plotted the results on a graph we compared: – high and low risk patients – 30 minutes before the episode and 2.5 hours before this 30 minutes period DC is reduced before an NSVT episode  DC is reduced before an NSVT episode 30min 30-180 before min p-value the before episode NSVT High Risk 3.11±0.46 3.53±0.33 0.00652 Low Risk 5.91±0.06 6.56±0.42 0.00065  Computation with DCsgn Deceleration Capacity before the Episode 9.5 9 8.5 8 p=0.00033 Deceleration Capacity 7.5 7 6.5 6 5.5 p=0.00035 5 4.5 4 1h 2h 3h Time before the Episode  DC is reduced before NSVT episode 30min 30-180 DC before min p-value (original) the before episode NSVT High Risk 3.11±0.46 3.53±0.33 0.00652 Low Risk 5.91±0.06 6.56±0.42 0.00065 DCsgn 30min 30-180 before min p-value the before episode NSVT High Risk 5.21±0.21 5.80±0.37 0.00035 Low Risk 7.36±0.15 8.08±0.47 0.00033  Other Related Activity of our Team Deceleration Capacity as an index to discriminate – low arrhythmia risk for sudden cardiac death – high arrhythmia risk for sudden cardiac death Circadian variability of based on Deceleration Capacity Deceleration Capacity before ventricular tachycardia episodes Deceleration Capacity before apnoea episodes  Summary Deceleration and Acceleration Capacity of heart rate were presented The original method was described Two variations on the original methods were also presented as well as one more method based on the maximum derivative Applications on specific datasets were also presented – as a predictor of mortality after myocardial infarction – in order to distinguish young and elderly subjects – to discriminate a control group and a group of athletes running long distances – to investigate the period before an NSVT episode Thank you