Massimo Pifferi1, Andrew Bush2, Giovanni Pioggia3, Maria Di Cicco4, Iolanda Chinellato5, Alessandro Bodini5, Pierantonio Macchia4, Attilio L Boner5. 1. Department of Pediatrics, University of Pisa, Pisa, Italy. Electronic address: m.pifferi@med.unipi.it. 2. Imperial School of Medicine at the National Heart and Lung Institute, London, England. 3. Institute of Clinical Physiology, CNR, Pisa, Italy. 4. Department of Pediatrics, University of Pisa, Pisa, Italy. 5. Department of Pediatrics, University of Verona, Verona, Italy.
Abstract
BACKGROUND: Asthma control is emphasized by new guidelines but remains poor in many children. Evaluation of control relies on subjective patient recall and may be overestimated by health-care professionals. This study assessed the value of spirometry and fractional exhaled nitric oxide (FeNO) measurements, used alone or in combination, in models developed by a machine learning approach in the objective classification of asthma control according to Global Initiative for Asthma guidelines and tested the model in a second group of children with asthma. METHODS: Fifty-three children with persistent atopic asthma underwent two to six evaluations of asthma control, including spirometry and FeNO. Soft computing evaluation was performed by means of artificial neural networks and principal component analysis. The model was then tested in a cross-sectional study in an additional 77 children with allergic asthma. RESULTS: The machine learning method was not able to distinguish different levels of control using either spirometry or FeNO values alone. However, their use in combination modeled by soft computing was able to discriminate levels of asthma control. In particular, the model is able to recognize all children with uncontrolled asthma and correctly identify 99.0% of children with totally controlled asthma. In the cross-sectional study, the model prospectively identified correctly all the uncontrolled children and 79.6% of the controlled children. CONCLUSIONS: Soft computing analysis of spirometry and FeNO allows objective categorization of asthma control status.
BACKGROUND: Asthma control is emphasized by new guidelines but remains poor in many children. Evaluation of control relies on subjective patient recall and may be overestimated by health-care professionals. This study assessed the value of spirometry and fractional exhaled nitric oxide (FeNO) measurements, used alone or in combination, in models developed by a machine learning approach in the objective classification of asthma control according to Global Initiative for Asthma guidelines and tested the model in a second group of children with asthma. METHODS: Fifty-three children with persistent atopic asthma underwent two to six evaluations of asthma control, including spirometry and FeNO. Soft computing evaluation was performed by means of artificial neural networks and principal component analysis. The model was then tested in a cross-sectional study in an additional 77 children with allergic asthma. RESULTS: The machine learning method was not able to distinguish different levels of control using either spirometry or FeNO values alone. However, their use in combination modeled by soft computing was able to discriminate levels of asthma control. In particular, the model is able to recognize all children with uncontrolled asthma and correctly identify 99.0% of children with totally controlled asthma. In the cross-sectional study, the model prospectively identified correctly all the uncontrolled children and 79.6% of the controlled children. CONCLUSIONS: Soft computing analysis of spirometry and FeNO allows objective categorization of asthma control status.