Design and simulation of an intelligent medical ventilator system for patient-ventilator synchrony and lung protection

Abstract

Respiratory failure can be life threatening for patients with hypoxemic and hypercapnic conditions, and the failure requires fast medical response to save the patient(s). Respiratory diseases such as acute respiratory distress syndrome (ARDS), pneumonia, and corona virus disease 2019 (COVID-19) have increased the need for mechanical ventilation (MV). Mechanical ventilators are devices that support patients in delivering oxygen to the lungs during inspiration and removing the carbondioxide during expiration. The ventilators are mostly controlled manually by physicians, in changing the ventilator settings, such as positive end expiratory pressure (PEEP). Physicians in most hospitals are few and the number of patients with respiratory failures is on the rise due to viruses, smoke inhalation, and injuries that cause fluids to collect in the air sacs of the lungs. The manual control of the ventilators comes with human errors caused by some physicians especially when they are overwhelmed with lots Of work, which may lead to inaccuracy or delay in the adjustments made on the mechanical ventilators; hence, resulting into delay in recovering of the patient or to the worst, increasing the death rate due to inaccurate ventilator setting adjustments. In this study, firstly, fuzzy logic method was used to develop a computer aided decision making to improve on the accuracy of the reasoning done during the ventilator setting adjustments. The ARDS Network (ARDSNet) PEEP values were used in building the fuzzy rules of the fuzzy algorithm and four algorithms were trained, that is: incremental high PEEP, decremental high PEEP, incremental low PEEP, and decremental low PEEP. The need for use of the fuzzy logic method to guide the physicians was with a purpose of reducing the physician's time spent in thinking on who parameters to adjust on the ventilator, especially during an emergency. Secondly, the fuzzy neural network model was further trained to increase the accuracy and reduce the time spent comparing the fuzzy logic algorithm output and the ARDSNet PEEP values. Thirdly, the convolutional neural network (CNN) was then used to classify the different brain and lung conditions, so as to add a variable and alarm to already existing ventilator variables and alarms respectively, Finally, a single lung compartment transfer function was derived and used to test the simulation of the fuzzy-image based system, with trained fuzzified PEEP and CNN algorithms connected to the derived lung transfer function. On testing the system, the first consideration was based on healthy brains after the brain medical scans, whereby the incremental high PEEP algorithm was selected for patients with pneumonia, and decremental high PEEP algorithm was selected for patients with COVID-19. The second consideration was based on healthy lungs after the lung medical scans in which incremental low PEEP algorithm was selected for patients with intracranial hemorrhage or acute ischemic stroke, and a decremental low PEEP algorithm was selected for patients with normal/ healthy brains. This guides the doctors in determining other causes of respiratory failure, hence, acting as an alarm on the ventilator. The neural fuzzy algorithm was able to mimic the ARDSNet PEEP values for high and low PEEP. The CNN algorithm had 97% validation accuracy for the three lung conditions, and validation accuracy for the three brain conditions. Three lung transfer functions were used in the simulation, that is: transfer functions for healthy lung, sick lung, and recovering lung. The lung residual volume (RV) and total lung capacity (TLC) values indicate that for healthy lungs, lower PEEP is desirable. For sick lungs, more PEEP is required to attain the recommended lung RV and TLC ranges. For recovering lungs, lower PEEP is required to balance the lung protection and oxygenation. The intelligent ventilator system can therefore, be applied as a physician-ventilator interface to guide the physician during the ventilation, so as to reduce on the human errors; ensure lung protection; and reduce on the time spent by physicians in making decisions. This can be achieved by the system, when it guides the physicians on what parameters to adjust at minimal possible time during the ventilation, and increase the possibility of patient survival. The efficiency of physicians that operate on the mechanical ventilators, especially during emergency times, can now be improved by not necessarily increasing the number of the physicians controlling one ventilator, but by adding the fuzzy-image reasoning to the mechanical ventilator systems, to help in providing suggestions to the physician(s).