Créer une présentation
Télécharger la présentation

Télécharger la présentation
## Alberto Porta

- - - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - - -

**Evaluating complexity of short-term heart period**variability through predictability techniques Alberto Porta Department of Biomedical Sciences for Health Galeazzi Orthopedic Institute University of Milan Milan, Italy**Introduction**There is an increasing interest in evaluating short term complexity of heart period variability in humans mainly due to its relationship with cardiac neural regulation, pathology and aging Traditional approaches quantify complexity in terms of information carried by the samples (i.e. entropy-based approaches) However, complexity can be estimated in terms of predictability of future samples when a certain amount of previous values are given (the smaller predictability, the larger complexity)**Primary aims**To propose tools assessing complexity of heart period variability via predictability-based approaches To demonstrate that this approach is strongly linked to the methods based on conditional entropy**Secondary aims**To show that complexity analysis of heart period variability is helpful to distinguish healthy subjects from pathological patients To demonstrate that complexity analysis of heart period variability can be fully exploited under uncontrolled experimental conditions and during daily activities**Outline**1) Predictability approach based on conditional distribution and uniform quantization 2) Conditional entropy approach based on uniform quantization 3) Predictability approach based on conditional distribution and k nearest neighbors 4) Conditional entropy approach based on k nearest neighbors 5) Application to 24h Holter recordings of heart period variability obtained from healthy subjects and chronic heart failure population**Outline**1) Predictability approach based on conditional distribution and uniform quantization 2) Conditional entropy approach based on uniform quantization 3) Predictability approach based on conditional distribution and k nearest neighbors 4) Conditional entropy approach based on k nearest neighbors 5) Application to 24h Holter recordings of heart period variability obtained from healthy subjects and chronic heart failure population**max(RR)-min(RR)**q=6 with ε = q Uniform quantization RR(i), i=1,...,N with RR(i) R ε RRq(i), i=1,...,N with RRq(i) I 0RRq(i)q-1**Day**Night**Pattern construction**f f: RRq(i), i=1,...,N RRq,L(i), i=1,...,N-L+1 with RRq,L(i) = (RRq(i),RRq(i-t),...,RRq(i-(L-1)t)) 0RRq(i)q-1 When =1, RRq,L(i) is a feature extracted from the series**Example of pattern construction (L=3)**RRq(i)= 3, 3, 3, 3, 2, 2, 1, ... (3,3,3) (3,3,3) (3,3,2) (3,2,2) (2,2,1) … RRq,L(i)= (3,3,3),(3,3,3),(3,3,2),(3,2,2),(2,2,2),...**g**(2,0,5) 2.62+0.61+5.60 = 77 Transformation of a pattern into an integer g g: RRq,L(i)=(RRq(i),RRq(i-1),...,RRq(i-L+1)) IL hq,L(i) I hq,L(i) = RRq(i).qL-1 + RRq(i-1).qL-2 + .... + RRq(i-L+1).q0 0hq,L(i)qL-1 Example with L=3 and q=6**Uniform quantization in 3-dimensional embedding space**(2,1,0) 78**Example of pattern distribution in a 3-dimensional**embedding space Porta A et al, IEEE Trans Biomed Eng, 48:1282-1291, 2001**Toward the assessment of complexity based on**prediction • Uniform quantization (in general any type of coarse graining) • of the embedding space provides the basis for • 1) entropy-based approaches • 2) prediction techniques**Transforming any L-dimensional quantized pattern**into a 2-dimensional one RRq,L(i) = (RRq(i),RRq(i-1),...,RRq(i-L+1)) = (RRq(i),RRq,L-1(i-1)) L-dimensional pattern 2-dimensional pattern (RRq(i),RRq,L-1(i-1)) (RRq(i),hq,L-1(i-1))**RR(i)**hq,L-1(i-1) Conditional distribution of the current sample given L-1 previous values Given the transformation (RRq(i),RRq,L-1(i-1)) (RRq(i),hq,L-1(i-1)) the conditional distribution of the current sample given L-1 previous values can be drawn in the plane**Examples of conditional distribution of the current heart**period given three past RR intervals (L=3) Day Night Porta A et al, Chaos, 17, 015117, 2007****RR(i/L-1) = median(RR(j)/RRq,L-1(j-1) = RRq,L-1(i-1)) = median(RR/hq,L-1(i-1)) e(i) = RR(i) – RR(i) N 1 MSPEUQ(L) = S e2(i) with 0MSPEUQ(L)MSD N-L i=L N 1 MSDUQ = S (RR(i)-RRm)2 where and RRm= median(RR) N-1 i=1 Prediction based on conditional distribution: the uniform quantization (UQ) approach Predictor Defined the prediction error as the mean square prediction error (MSPE) is MSPEUQ(L) = 0 perfect prediction MSPEUQ(L) = MSDUQ null prediction Porta A et al, IEEE Trans Biomed Eng, 47, 1555-1564, 2000**(RR(i), hq,L-1(i-1))**(median(RR(i)/hq,L-1(i-1)), hq,L-1(i-1)) Examples of prediction based on conditional distribution with L=3 during daytime**(RR(i), hq,L-1(i-1))**(median(RR(i)/hq,L-1(i-1)), hq,L-1(i-1)) Examples of prediction based on conditional distribution with L=3 during nighttime**Overfitting** e(i) = RR(i) – RR(i) = 0 “Single” points do not contribute to MSPE**Fraction of “singles” 1 with L**Course of single patterns with pattern length Day Night**Mean square prediction error**Day Night MSPEUQ(L) 0 with L**Corrected mean square prediction error (CMSPEUQ)**and normalized CMSPEUQ (NCMSPEUQ) CMSPEUQ(L) = MSPEUQ(L) + MSD.fraction(L) with 0CMSPEUQ(L)MSD CMSPEUQ(L) NCMSPEUQ(L) = MSD 0NCMSPEUQ(L)1 Porta A et al, IEEE Trans Biomed Eng, 47, 1555-1564, 2000**Normalized unpredictability index (NUPIUQ)**Day Night NUPIUQ=min(NCMSPEUQ(L)) 0NUPIUQ1**Outline**1) Predictability approach based on conditional distribution and uniform quantization 2) Conditional entropy approach based on uniform quantization 3) Predictability approach based on conditional distribution and k nearest neighbors 4) Conditional entropy approach based on k nearest neighbors 5) Application to 24h Holter recordings of heart period variability obtained from healthy subjects and chronic heart failure population**Conditional entropy**CE(L) = -Sp(RRq,L-1(i-1)).SE(RRq/RRq,L-1(i-1)) where SE(RR/RRq,L-1(i-1)) = Sp(RRq(i)/RRq,L-1(i-1)).log(RRq(i)/RRq,L-1(i-1))) with 0CE(L)SE(RR) and SE(RR) = -Sp(RRq(i)).log(p(RRq(i)))**Conditional entropy**g Given the transformation g: RRq,L-1(i) hq,L-1(i) CE(L) = -Sp(hq,L-1(i-1)).SE(RRq/hq,L-1(i-1)) where SE(RR/hq,L-1(i-1)) = Sp(RRq(i)/hq,L-1(i-1)).log(RRq(i)/hq,L-1(i-1))) with 0CE(L)SE(RR) and SE(RR) = -Sp(RRq(i)).log(p(RRq(i))) Porta A et al , Biol Cybern, 78:71-78, 1998 Porta A et al , Med Biol Eng Comput, 38, 180-188, 2000**Example of calculation of conditional entropy (L=4)**Porta A et al, Chaos, 17, 015117 2007**Example of calculation of conditional entropy (L=4)**during daytime and nighttime Day Night CE(L=4) during daytime < CE(L=4) during nighttime**Bias of conditional entropy (L=4)**SE(RRq/hq,L-1(i-1)))=0 “Single” points do not contribute to CE**Fraction of “singles” 1 with L**Course of single patterns with pattern length Day Night**CE(L) 0 with L**Conditional entropy Day Night**Corrected conditional entropy (CCE)**and normalized CCE (NCCE) CCE(L) = CE(L) + SE(L=1).fraction(L) 0CCE(L)SE(L=1) CCE(L) 0NCCE(L)1 NCCE(L) = SE(L=1) Porta A et al , Biol Cybern, 78:71-78, 1998 Porta A et al , Med Biol Eng Comput, 38, 180-188, 2000**Normalized complexity index (NCIUQ)**Day Night NCIUQ=min(NCCE(L)) with 0NCI1**Outline**1) Predictability approach based on conditional distribution and uniform quantization 2) Conditional entropy approach based on uniform quantization 3) Predictability approach based on conditional distribution and k nearest neighbors 4) Conditional entropy approach based on k nearest neighbors 5) Application to 24h Holter recordings of heart period variability obtained from healthy subjects and chronic heart failure population****RR(i/L-1) = median(RR(j)/RRL-1(j-1) belongs to the set of the k nearest neighbors of RRL-1(i-1)) e(i) = RR(i) – RR(i) N 1 MSD = S (RR(i)-RRm)2 where and RRm= median(RR) N-1 i=1 Prediction based on conditional distribution: the k-nearest-neighbor (KNN) approach Predictor Defined the prediction error as the mean square prediction error (MSPEKNN) is N 1 MSPEKNN(L) = S with 0MSPEKNN(L)MSD e2(i) N-L i=L MSPEKNN(L) = 0 perfect prediction MSPEKNN(L) = MSD null prediction A. Porta et al, IEEE Trans Biomed Eng, 54:94-106 , 2007**Normalized k-nearest-neighbor mean square**prediction error (NKNNMSPE) MSPEKNN(L) NMSPEKNN(L) = MSD with 0NMSPEKNN(L)1 NMSPEKNN(L) = 0 perfect prediction NMSPEKNN(L) = 1 null prediction**Normalized unpredictability index based on**k-nearest-neighbor approach Day Night NUPIKNN=min(NMSPEKNN(L)) with 0NUPIKNN1 A. Porta et al, IEEE Trans Biomed Eng, 54:94-106, 2007**Outline**1) Predictability approach based on conditional distribution and uniform quantization 2) Conditional entropy approach based on uniform quantization 3) Predictability approach based on conditional distribution and k nearest neighbors 4) Conditional entropy approach based on k nearest neighbors 5) Application to 24h Holter recordings of heart period variability obtained from healthy subjects and chronic heart failure population**N**1 KNNCE(L) = SSE(RR/RRL-1(i-1)) N-L+1 i=L K-nearest-neighbor conditional entropy (KNNCE) where SE(RR/RRL-1(i-1)) is the Shannon entropy of conditional distribution of RR(j) given that RRL-1(j-1) belongs to the set of k-nearest-neighbors of RRL-1(i-1) with 0KNNCE(L)SE(RR) and SE(RR) = -Sp(RR(i)).log(p(RR(i))) Porta A et al, Physiol Meas, 34:17-33, 2013**Normalized k-nearest-neighbor**conditional entropy (KNNCE) KNNCE(L) NKNNCE(L) = SE(RR) with 0NKNNCE(L)1 NKNNCE(L) = 0 null information, perfect prediction NKNNCE(L) = 1 maximum information, null prediction**Normalized complexity index based on**k-nearest-neighbor approach Day Night NCIKNN=min(NKNNCE(L)) with 0NCIKNN1 Porta A et al, Physiol Meas, 34:17-33, 2013**Outline**1) Predictability approach based on conditional distribution and uniform quantization 2) Conditional entropy approach based on uniform quantization 3) Predictability approach based on conditional distribution and k nearest neighbors 4) Conditional entropy approach based on k nearest neighbors 5) Application to 24h Holter recordings of heart period variability obtained from healthy subjects and chronic heart failure population**Experimental protocol**12 normal (N) subjects (aged 34 to 55) 13 chronic heart failure (CHF) patients (aged 33 to 56) CHF patients are 2 in NYHA class I, 2 in NYHA class II, 9 in NYHA class III). Ejection fraction ranges from 13% to 30%, median=25% ECGs were recorded for 24h with a standard analogue Holter recorder. ECGs were sampled at 250 Hz and QRS detection was automatically performed by the software of the device**Analysis of 24h Holter recordings of heart period**variability Night Day Day: from 09:00 AM to 07:00 PM Night: from 00:00 AM to 05:00 AM NUPIUQ, NUPIKNN, NCIUQ, NCIKNN were calculated iteratively over sequences of 300 samples with 50% overlap The median of the distribution of NUPIUQ, NUPIKNN, NCIUQ, NCIKNN during daytime and nighttime was assessed**NUPIUQ in N subjects and CHF patients**Porta A et al, Chaos, 17, 015117 2007**NCIUQ in N subjects and CHF patients**Porta A et al, Chaos, 17, 015117 2007**Conclusions**Complexity of heart period variability can be assessed via predictability-based approaches These approaches lead to conclusions similar to those drawn using entropy-based methods Complexity analysis of heart period variability is helpful to distinguish healthy subjects from pathological patients Complexity analysis of heart period variability does not require controlled experimental conditions to provide meaningful results