Session P7D.4

From Knowledge to Understanding: Usefulness of Decision Support and Visualization Techniques in Patients with Atrial Fibrillation

A Drzewiecka*, R Mlynarski, G Ilczuk, A Wakulicz-Deja, W Kargul

Medical University of Silesia
Katowice, Poland

Atrial fibrillation (AF) is a huge medical and economic problem in highly developed societies. Treatment of AF is very complicated due to many disagreements and the number of existing therapeutic methods. Pacemaker implantation can be useful especially if the AF is accompanied by sinus node dysfunction. Visual techniques and decision support can have a special place in support of the treatment of patients with AF.
Purpose: Developing a visual method for presenting decision rules by rough set algorithms generated for support of treatment of patients with AF.
Methods: Decision rules were generated by our modified MLEM2 method based on the data of 1745 patients suffering from paroxysmal AF (934) as well as permanent AF (811). Afterwards post-processing methods were used to join and generalize induced decision rules, which were the input for our visualization method. Decision rules and original data were displayed as multiple translucent layers on a 2D matrix. Additionally, gradient LUT tables were used for the encoding of data values. Finally rules in the form of decision trees (AQDT-2 algorithm) were validated by experts from the Electrocardiology Department.
Results: The results are presented as the number of decision rules generated by the method / accuracy in % compared to the experts. The overall accuracy of the system for a decision about a pacemaker implantation was 188 / 83, 3%. For various types of pacemakers - DDD: 101 / 86, 5; DDDRP: 56 / 78, 2%; VVI: 122 / 83, 1% and VDD: 43 / 91, 2%. Good usefulness of the system was also confirmed for decision: rate control vs. rhythm control (196 / 76, 5). Decisions about cardioversion (121 / 75, 1%) or ablation (36 / 65, 2%) have lower accuracy at this time – additional research is necessary to improve accuracy. Our method shows a relation between data in a way that allows easy and fast validation and decision supporting treatment by experts which was confirmed in all cases.
Conclusions: The presented method assures good accuracy of the results in combination with an understandable graphical interpretation of the achieved results which simplifies the verification by experts. The practical application of this method in support / analysis systems for patients with AF was confirmed.

(Abstract Control Number: 262)