AJR 2010; 195:W319 0361–803X/10/1954–W319 © American Roentgen Ray Society Time to Change the Philosophy of Medicine We have read the article by Bowles and colleagues [1] with great interest. The authors suggest the use of computerized systems that automatically link assessment and recommendations as an effective method that can further increase the consistency of BI-RADS guidance. Although we agree with the authors, we believe that limitations of the BI-RADS system that arise from the methodology of system development must be emphasized. Despite the fact that scoring systems perform well in predicting outcomes in the original cohorts on which the models were developed, they may underestimate or overestimate outcomes in different populations [2]. Standardized scoring and reporting systems, such as BI-RADS, have been developed by use of conventional statistical methods, which are completely reductionist in nature. Contrary to the belief of Rene Descartes, biologic systems are complex systems, and their behavior cannot be explained by reductionist approaches [3]. In recent years, various methods based on artificial intelligence techniques, claiming to be universal approximators, have been proposed as alternatives to statistical methods, especially to model highly nonlinear functional relationships [2]. Artificial neural networks (ANNs), which are useful in detecting complex nonlinear relationships between a set of inputs and outputs, can be used as an example. ANNs have been shown to be as effective and sometimes superior to polynomial equations in predicting quantitative nonlinear relationships between variables and responses. Fuzzy logic is another problemsolving technique that defines and generates responses based on ambiguous, incomplete, and imprecise information. Fuzzy systems have drawn growing attention and interest in bioinformatics applications, decision-making studies, pattern recognition, and data analysis. By a combination of ANNs and fuzzy logic, the so called neurofuzzy approach, deficiencies of both can be compensated to some extent and their explanatory power can be increased further. Neurofuzzy systems have already been successfully applied to computer-aided diagnosis (e.g., detection of microcalcification, automatic detection of distorted plethysmogram pulses in neonates and pediatric patients, detection of erythematosquamous diseases, and lung-nodule detection) and have been useful for predicting the presence of prostate cancer [2]. In conclusion, we must change our reductionistic philosophy to understand disease states and their treatment. Selda Tez 19 Mayis Hospital Ankara, Turkey Yusuf A. Kilic Hacettepe University School of Medicine Ankara, Turkey Mesut Tez Numune Training and Research Hospital Ankara, Turkey