The incidence and prevalence of atrial diseases, particularly atrial fibrillation (AF), are today reaching pandemic proportions. Despite the considerable research efforts and progress regarding the understanding of mechanisms driving AF, the ability to treat it remains to be problematic. This thesis aims to contribute to reducing the burden of atrial arrhythmia from the technical point of view by processing the electrical signals coming from the heart. Specifically, the main objective of this thesis is to investigate signal processing techniques of invasive atrial signals that provide useful tools for helping electrophysiologists and technicians in the decision process during ablation treatment procedures or arrhythmia studies. First, we proposed a new algorithm, for ventricular activity cancellation in atrial signals during AF. The methodology combines two common techniques, average beat subtraction (ABS) and interpolation, in a unified framework. Such framework was able to refine and improve the ventricular activity estimate, under the stationary assumption of the atrial activity in very short time windows. Briefly, the local atrial activity is first modeled with an autoregressive (AR) process, then the estimate is refined by maximizing the log maximum a posteriori of the atrial residual activity according to the fitted AR model. The new algorithm was tested on both synthetic and real atrial signals, and the performance was compared with the other five algorithms. The proposed algorithm outperformed all the others in terms of average root mean square error (0.038 vs 0.045 for interpolation; p<0.05) on synthetic data. On real data, it outperformed two variants of ABS (p<0.05) and performed similarly to interpolation when considering the high-power residues left (both <3%), and the log-likelihood with the fitted AR model. Second, we investigated a novel approach to atrial propagation pattern analysis based on directed networks (graphs). The networks are generated by processing signals collected during electrophysiologic studies. Network vertices represent the locations of the recordings and edges are determined by computing measures between recorded signals. The algorithm automatically identifies potential targets for treatment, such as rotational activity, spreading from electrode to electrode creating a closed loop, or focal activity, manifesting as a divergence of excitation from a given region. The method was tested on two subjects in sinus rhythm, seven in an experimental model of in-silico simulations, and ten subjects diagnosed with complex atrial tachycardia who underwent catheter ablation. The algorithm correctly detected the electrical propagation of both sinus rhythm cases and all in-silico simulations. Regarding the clinical cases, arrhythmia mechanisms were identified in most of the cases (9 out of 10), i.e., cycles around the mitral valve, tricuspid valve, and figure-of-eight macroreentries. Third, we proposed a recommender system, built as a solution to an optimization problem, able to suggest the optimal ablation strategy for the treatment of complex atrial tachycardia. The problem was designed on top of directed network mapping. The optimization problem modeled the optimal ablation strategy as that one interrupting all reentrant mechanisms while minimizing the ablated atrial surface. Considering the exponential complexity of finding the optimal solution to the problem, we introduced a heuristic algorithm with polynomial complexity. The proposed algorithm was applied to the data of i) 6 simulated scenarios including both left and right atrial flutter; and ii) 10 subjects that underwent a clinical routine. The recommender system suggested the optimal strategy in 4 out of 6 simulated scenarios. On clinical data, the recommended ablation lines were found satisfactory on 67% of the cases according to the clinician's opinion comparing to the actual treatment performed, while they were correctly located in 89% considering the mechanism in place. The algorithm made use of only data collected during mapping and was able to process them nearly real-time.
L'incidenza e la prevalenza delle malattie atriali, in particolare della fibrillazione atriale (FA), stanno raggiungendo oggi proporzioni pandemiche. Nonostante i notevoli sforzi di ricerca e i progressi compiuti nella comprensione dei meccanismi alla base della FA, la capacità di trattarla rimane problematica. Questa tesi si propone di contribuire a ridurre il peso dell'aritmia atriale dal punto di vista tecnico, elaborando i segnali elettrici provenienti dal cuore. In particolare, l'obiettivo principale di questa tesi è studiare tecniche di elaborazione dei segnali atriali invasivi che forniscano strumenti utili per aiutare gli elettrofisiologi e tecnici di laboratorio nel processo decisionale durante le procedure di ablazione delle aritmie. In primo luogo, abbiamo proposto un nuovo algoritmo per la cancellazione dell'attività ventricolare nei segnali atriali durante la FA. La metodologia combina due tecniche comuni: la sottrazione del battito medio (ABS) e l'interpolazione, in una struttura unificata. Tale struttura è stata in grado di affinare e migliorare la stima dell'attività ventricolare, sotto l'ipotesi di stazionarietà dell'attività atriale in finestre temporali molto brevi. L'attività atriale locale viene prima modellata con un processo autoregressivo (AR), quindi la stima viene raffinata tramite la tecnica di massimo a posteriori tra l'attività residua atriale e il modello AR adattato. Il nuovo algoritmo è stato testato su segnali atriali sintetici e reali e le sue prestazioni sono state confrontate con quelle degli altri cinque algoritmi. L'algoritmo proposto ha superato tutti gli altri in termini di errore quadratico medio (0,038 vs 0,045 per l'interpolazione; p<0.05) su dati sintetici. Sui dati reali ha superato due varianti di ABS (p<0.05) e si è comportato in modo simile all'interpolazione. In secondo luogo, abbiamo studiato un nuovo approccio all'analisi dei modelli di propagazione atriale basato su reti (grafi) dirette. Le reti sono generate dall'elaborazione dei segnali raccolti durante gli studi elettrofisiologici. I vertici della rete rappresentano le posizioni delle registrazioni e i collegamenti sono determinati dal calcolo delle misure tra i segnali registrati. Dopo la modellizzazione della rete, è possibile identificare automaticamente i potenziali bersagli del trattamento, come l'attività rotazionale, che si propaga da un elettrodo all'altro creando un ciclo chiuso, o l'attività focale, che si manifesta come una divergenza di eccitazione da una determinata regione. Il metodo è stato testato su due soggetti in ritmo sinusale, sette in un modello sperimentale di simulazioni in-silico e dieci soggetti con diagnosi di tachicardia atriale complessa sottoposti ad ablazione. L'algoritmo ha rilevato correttamente la propagazione elettrica di entrambi i casi di ritmo sinusale e di tutte le simulazioni in-silico. Per quanto riguarda i casi clinici, i meccanismi dell'aritmia sono stati identificati nella maggior parte dei casi (9 su 10). In terzo luogo, abbiamo proposto un sistema di raccomandazione, costruito come soluzione a un problema di ottimizzazione, in grado di suggerire la strategia di ablazione ottimale per il trattamento della tachicardia atriale complessa. Il problema è stato progettato basandosi sulla rete precedentemente proposta. Il problema di ottimizzazione modellava la strategia di ablazione ottimale come quella che interrompe tutti i meccanismi rotazionali, riducendo al minimo la superficie atriale ablata. Considerando la complessità esponenziale di trovare la soluzione ottimale al problema, abbiamo introdotto un algoritmo euristico con complessità polinomiale. L'algoritmo proposto è stato applicato ai dati di i) 6 scenari simulati che includevano sia il flutter atriale destro che sinistro; e ii) 10 soggetti sottoposti a routine clinica. Il sistema di raccomandazione ha suggerito la strategia ottimale in 4 dei 6 scenari simulati. Per quanto riguarda i dati clinici, le linee di ablazione raccomandate sono risultate soddisfacenti nel 67% dei casi, secondo l'opinione del medico, rispetto al trattamento effettivamente eseguito, mentre sono state correttamente localizzate nell'89% dei casi, considerando il meccanismo in uso. L'algoritmo ha utilizzato solo i dati raccolti durante la mappatura avvenuta durante la procedura clinica ed è stato in grado di elaborarli quasi in tempo reale.
ATRIAL COMPLEX NETWORKS IN ENDOCAVITARY RECORDINGS DURING ATRIAL FIBRILLATION / M. Vila ; tutor: S. Roberto, M. W. Rivolta ; coordinatore: P. Boldi. Dipartimento di Informatica Giovanni Degli Antoni, 2022 Jul 18. 34. ciclo, Anno Accademico 2021.
ATRIAL COMPLEX NETWORKS IN ENDOCAVITARY RECORDINGS DURING ATRIAL FIBRILLATION
M. Vila
2022
Abstract
The incidence and prevalence of atrial diseases, particularly atrial fibrillation (AF), are today reaching pandemic proportions. Despite the considerable research efforts and progress regarding the understanding of mechanisms driving AF, the ability to treat it remains to be problematic. This thesis aims to contribute to reducing the burden of atrial arrhythmia from the technical point of view by processing the electrical signals coming from the heart. Specifically, the main objective of this thesis is to investigate signal processing techniques of invasive atrial signals that provide useful tools for helping electrophysiologists and technicians in the decision process during ablation treatment procedures or arrhythmia studies. First, we proposed a new algorithm, for ventricular activity cancellation in atrial signals during AF. The methodology combines two common techniques, average beat subtraction (ABS) and interpolation, in a unified framework. Such framework was able to refine and improve the ventricular activity estimate, under the stationary assumption of the atrial activity in very short time windows. Briefly, the local atrial activity is first modeled with an autoregressive (AR) process, then the estimate is refined by maximizing the log maximum a posteriori of the atrial residual activity according to the fitted AR model. The new algorithm was tested on both synthetic and real atrial signals, and the performance was compared with the other five algorithms. The proposed algorithm outperformed all the others in terms of average root mean square error (0.038 vs 0.045 for interpolation; p<0.05) on synthetic data. On real data, it outperformed two variants of ABS (p<0.05) and performed similarly to interpolation when considering the high-power residues left (both <3%), and the log-likelihood with the fitted AR model. Second, we investigated a novel approach to atrial propagation pattern analysis based on directed networks (graphs). The networks are generated by processing signals collected during electrophysiologic studies. Network vertices represent the locations of the recordings and edges are determined by computing measures between recorded signals. The algorithm automatically identifies potential targets for treatment, such as rotational activity, spreading from electrode to electrode creating a closed loop, or focal activity, manifesting as a divergence of excitation from a given region. The method was tested on two subjects in sinus rhythm, seven in an experimental model of in-silico simulations, and ten subjects diagnosed with complex atrial tachycardia who underwent catheter ablation. The algorithm correctly detected the electrical propagation of both sinus rhythm cases and all in-silico simulations. Regarding the clinical cases, arrhythmia mechanisms were identified in most of the cases (9 out of 10), i.e., cycles around the mitral valve, tricuspid valve, and figure-of-eight macroreentries. Third, we proposed a recommender system, built as a solution to an optimization problem, able to suggest the optimal ablation strategy for the treatment of complex atrial tachycardia. The problem was designed on top of directed network mapping. The optimization problem modeled the optimal ablation strategy as that one interrupting all reentrant mechanisms while minimizing the ablated atrial surface. Considering the exponential complexity of finding the optimal solution to the problem, we introduced a heuristic algorithm with polynomial complexity. The proposed algorithm was applied to the data of i) 6 simulated scenarios including both left and right atrial flutter; and ii) 10 subjects that underwent a clinical routine. The recommender system suggested the optimal strategy in 4 out of 6 simulated scenarios. On clinical data, the recommended ablation lines were found satisfactory on 67% of the cases according to the clinician's opinion comparing to the actual treatment performed, while they were correctly located in 89% considering the mechanism in place. The algorithm made use of only data collected during mapping and was able to process them nearly real-time.File | Dimensione | Formato | |
---|---|---|---|
phd_unimi_R12143.pdf
accesso aperto
Descrizione: full text
Tipologia:
Post-print, accepted manuscript ecc. (versione accettata dall'editore)
Dimensione
4.68 MB
Formato
Adobe PDF
|
4.68 MB | Adobe PDF | Visualizza/Apri |
Pubblicazioni consigliate
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.