Retrieving similar cases for clinical decision support in the context of revascularization of lower limbs

Vascular surgeons face complex decisions regarding revascularization of lower limbs of patients, for which there is no decision support solution yet available.

Relying on Case-Based Reasoning approach, this paper addresses the issue of case retrieving in the context of lower limb revascularization. To this end we propose a similarity measure based on relevant clinical attributes selection and weighting. We analyze its performance for clinical decision support.

Retrieval process is used to anticipate the success or failure of a revascularization strategy for a new patient based on the surgical procedure outcome of similar past patient cases. Using clinical knowledge in Global vascular Guidelines, cases were structured and formalized. They were described by several attributes reflecting the patient risk estimation, limb pathology staging, localization of atherosclerotic lesions, vascular history and revascularization type. To construct the similarity measure for lower limb revascularization CBR, three distance metrics (Euclidean, Cosine and Heterogeneous), three attributes selection scenarios as well as two weighting strategies were identified and compared.

A case base of 17 patients was built and used to assess the relevance of the retrieval process when predicting the long-term outcome of the revascularization procedure. Similarity measure showed good performance and clinical relevance when using a heterogeneous distance with attributes and weights derived from information in official guidelines, vascular history and type of revascularization.

By investigating and validating the similar cases retrieving process, we showed that CBR appears to be a promising approach for clinical decision support in the context of lower extremity arterial disease.