Cardiovascular diseases, a leading global cause of death, can be treated using Minimally Invasive Surgery (MIS) for various heart conditions. Cardiac ablation is an example of MIS, treating heart rhythm disorders like atrial fibrillation and the operation outcomes are highly dependent on the surgeon's skills. This procedure utilizes catheters, flexible endovascular devices inserted into the patient's blood vessels through a small incision. Traditionally, novice surgeons' performance is assessed in the Operating Room (OR) through surgical tasks. Unskilled behavior can lead to longer operations and inferior surgical outcomes. However, an alternative approach can be capturing surgeons' maneuvers and using them as input for an AI model to evaluate their skills outside the OR. To this end, two experimental setups were proposed to study the skills modelling for surgical behaviours. The first setup simulates the ablation procedure using a mechanical system with a synthetic heartbeat mechanism that measures contact forces between the catheter's tip and tissue. The second one simulates the cardiac catheterization procedure for the surgeon’s practice and records the user's maneuvers at the same time. The first task involved maintaining the force within a safe range while the tip of the catheter is touching the surface. The second task was passing a catheter’s tip through curves and level-intersection on a transparent blood vessel phantom. To evaluate attendees' demonstrations, it is crucial to extract maneuver models for both expert and novice surgeons. Data from participants, including novices and experts, performing the task using the experimental setups, is compiled. Deep recurrent neural networks are employed to extract the model of skills by solving a binary classification problem, distinguishing between expert and novice maneuvers. The results demonstrate the proposed networks' ability to accurately distinguish between novice and expert surgical skills, achieving an accuracy of over 92%.