A fuzzy classification system to predict renal artery restenosis after angioplasty

Percutaneous transluminal angioplasty (PTRA) has emerged as a promising treatment for patients with renovascular hypertension. However, the benefit of this procedure is hampered by restenosis that frequently occurs within around 6 months after a successful angioplasty. This paper presents a fuzzy classification system based on a fuzzy pattern matching model that is being developed to evaluate the risk of short-term restenosis. First, identified classes are represented by fuzzy prototypes that take into account the imprecision of the criteria. Second, the system is applied to angiographic features of given stenoses and provides the membership degree of these stenosis to the two classes "short term restenosis" or "no restenosis". The fuzzy classifier's performances have been tested on twenty two patients who underwent balloon angioplasty in the context of a French multicenter randomized trial EMMA. With the fuzzy classifier, restenosis were predicted prospectively with 100% while its sensitivity is about 73%. The fuzzy classification system is expected to become a reliable tool to predict PTRA outcomes.