Cardiac Positioning System (CPS) - An automated navigation system to guide catheter ablation therapy

Why is it important to improve treatments used after a heart attack?
Each year in the UK, approximately 150,000 people have a heart attack when the blood supply to their heart is compromised. As a result, affected regions of the heart can become diseased and scarred. In a healthy person, electrical waves propagate across the heart in a regulated pattern which triggers contraction to pump blood around the body. The scar tissue that forms as a result of a heart attack can disrupt the propagation of the electrical waves. If significant disruptions occur, blood cannot be pumped out of the body effectively, leading to sudden death.

Ablation therapy aims to eliminate areas of diseased tissue that cause disruption to the heart rhythm, by applying radiofrequency using catheters inserted into the heart. The most accurate techniques used to locate the region to ablate require the induction of dangerous heart rhythms, which are only inducible in about 65% of people.

Pace mapping is a technique used to locate regions to ablate, which can be performed during normal heart rhythm. ECG data, which records electrical signals from the heart, is collected when the patient has an abnormal heart rhythm. From this template ECG, a clinician can tell the approximate location of the diseased tissue. A catheter is directed to that location, the heart stimulated, and another ECG, called the paced ECG is recorded. If the paced ECG matches the template ECG, it is assumed that the heart was paced in the location that requires ablation.

Current ablation techniques are difficult, time consuming, and inaccurate. As a result, the procedure may work in only half of all patients, and result in unnecessary damage to healthy tissue, leading to later impairment of heart function.

How will we improve the treatment?
The CPS project's overall goal is to increase the success rates of ablation therapy by improving the accuracy and efficiency of locating the optimal region of tissue to eliminate during the pace mapping procedure. Our research will involve using machine learning to find patterns in data which may be able to help locate diseased tissue. Machine learning involves passing some input and corresponding output data to a machine learning algorithm which can then 'learn' how to predict output data.

The first aim of the CPS project is to make the initial prediction about the location of diseased tissue more accurate in order to guide the initial placement of the catheter. Patient specific data known to affect ECGs, such as medication, and heart and chest size, as well as ECG data, will be incorporated into a machine learning algorithm. Once the computer has 'learnt' how these data correlate with the location of diseased tissue, it will be able to predict the location of diseased tissue in a new patient.

The second aim involves accurately locating the critical region of diseased tissue responsible for disrupting the heart rhythm, in order to pinpoint the optimal target for ablation. A machine learning algorithm will be used to compare template and paced ECGs, and understand how the differences between them, can indicate the specific location of the tissue which is causing the heart rhythm disturbance. Once the algorithm has 'learnt' how to locate the ablation target, we can use it to indicate where the clinician should ablate in new patients.

What is the significance of this research?
Increasing ablation therapy success rates will mean that patients will be unlikely to suffer from future heart rhythm disorders as a result of their heart attack, increasing the life expectancy of heart attack patients. Excess damage caused to the heart as a result of unnecessary ablation lesions will be limited, decreasing the likelihood of future complications. In addition, dangerous heart rhythms do not need to be induced in the patient, significantly decreasing the risk of death during the treatment.

Background
Reentrant waves can form around myocardial infarction scars, potentially resulting in VT. The reentrant pathway can be interrupted by radiofrequency ablation lesions, terminating and preventing VT. Pace mapping is performed to locate the exit site for ablation. Prediction of scar location is guided by an ECG of the clinical VT. Pacing is performed within the scar, and template and paced ECGs are compared. Good correlation indicates pacing from the exit site. The procedure is lengthy, difficult and VT recurrence rates are high.

Objectives
Regions of scar will be more accurately identified and human error mitigated, in guiding the catheter to the initial pacing site. Coordinates of scar location will be specified with an error boundary extending to cover 12% of the heart (average scar size).

Pace mapping accuracy will be increased; given 40 is the average number of pace maps generated per patient, I aim to locate the target by sampling less than 10 maps: an improvement of 75%.

Methods
A machine learning algorithm will use clinical ECG, patient geometry and medication data for input, with corresponding location of successful ablation, derived from an MRI patient model, as output data. Given input data from newly presenting patients, the algorithm will predict the scar location by tagging the model and registering it with the electroanatomical map.

Difference graphs calculated by comparing template and paced ECGs will provide input data for a machine learning algorithm, with corresponding location of successful ablation as output data. Given input data from new patients, the machine learning algorithm will indicate the ablation target.

Research Outputs
A software product will be produced and integrated into the CARTO mapping system for clinical use to improve ablation therapy success rates. IPR will be obtained and product information disseminated to clinicians via journals and conferences such as the European Heart Rhythm Association.