دانلود رایگان ترجمه مقاله یک سیستم استنباط منطق فازی جدید برای پشتیبانی از تصمیم گیری (نشریه اسپرینگر ۲۰۱۰)

یک سیستم استنباط منطق فازی تازه برای پشتیبانی از تصمیمات مربوط به جداسازی دستگاه تنفس مصنوعی از بیماران

A Novel Fuzzy Logic Inference System for Decision Support in Weaning from Mechanical Ventilation

Abstract

Weaning from mechanical ventilation represents one of the most challenging issues in management of critically ill patients. Currently used weaning predictors ignore many important dimensions of weaning outcome and have not been uniformly successful. A fuzzy logic inference system that uses nine variables, and five rule blocks within two layers, has been designed and implemented over mathematical simulations and random clinical scenarios, to compare its behavior and performance in predicting expert opinion with those for rapid shallow breathing index (RSBI), pressure time index and Jabour’ weaning index. RSBI has failed to predict expert opinion in 52% of scenarios. Fuzzy logic inference system has shown the best discriminative power (ROC: 0.9288), and RSBI the worst (ROC: 0.6556) in predicting expert opinion. Fuzzy logic provides an approach which can handle multi-attribute decision making, and is a very powerful tool to overcome the weaknesses of currently used weaning predictors.

۱ Introduction

Mechanical ventilation offers essential ventilatory support, while the respiratory system recovers from acute respiratory failure [1]. While often life-saving, mechanical ventilation is the most complex of critical care procedures, and is an invasive, costly therapy with well-documented complications [2–۶]. Besides increased risk of ventilator associated pneumonia, barotrauma and nosocomial infections, mortality is known to be significantly higher in patients with prolonged mechanical ventilation [1]. That’s why we want to wean and extubate patients at the earliest possible time. But failure to predict the right time to wean, may result in respiratory failure and reintubation of the patient, which is also known to increase morbidity and mortality [2].

The search for a reliable predictor of weaning success has yielded numerous predictors, but without reproducible results [7, 8]. Among them rapid shallow breathing index (RSBI) is the most widely used, in part due to its ease of calculation. It is measured after one minute of spontaneous breathing as the ratio of frequency to tidal volume, and a RSBI below the threshold of 105 (breaths/min)/L predicts weaning success with a sensitivity of up to 97% [9]. But its sensitivity decreases with prolonged ventilation, and its specificity is influenced by the disease state [7]. Specificity of 65% observed in patients with COPD, decreases to 28% in patients with acute respiratory failure [7, 10]. The other predictors of weaning represent more detailed parameters of respiratory dynamics, but are not widely used because of the difficulties in required measurements or bedside calculations. That’s why the focus of clinical studies was mostly on finding a simple weaning predictor that is easy to measure and calculate.

Currently used weaning predictors almost solely focus on respiratory parameters, and fail to represent the whole clinical picture that forms the basis for readiness for weaning. Most of them ignore systemic perfusion, ventilation and acid–base balance, which have significant impact on the success of weaning trial.

In a recent review Siner and Manthous emphasized the diversity of variables that affect weaning outcomes and state that the process of weaning “will always be art guided but not predicted entirely by science” [۱۱]. This notion, which we partly agree, actually points to the inadequacy of currently used weaning predictors in explaining the chaotic nature of weaning, and the superiority of human reasoning which is much powerful in solving multi-attribute problems. An approach closer to human reasoning may allow design of better predictive systems, and fuzzy-logic is a powerful tool in that respect.