Limits...
Making it possible to measure knowledge, experience and intuition in diagnosing lung injury severity: a fuzzy logic vision based on the Murray score.

D'Negri CE, De Vito EL - BMC Med Inform Decis Mak (2010)

Bottom Line: Murray score is the result of an equation that gives all its variables the same linear contribution and weight and makes use of consented cut-offs.An overestimation was found in the surveyed group's interpretation of severity.FL methodology could overcome a series of restrictions that current tests have due to cut-offs.

View Article: PubMed Central - HTML - PubMed

Affiliation: Consejo Nacional de Investigaciones Científicas y Técnicas, Combatientes de Malvinas 3150, CP 1427, Buenos Aires, Argentina. cdnegri@lanari.fmed.uba.ar

ABSTRACT

Background: Murray score is the result of an equation that gives all its variables the same linear contribution and weight and makes use of consented cut-offs. Everyday physicians' vocabulary is full of terms (adjectives) like: little, small, low, high, etc. that they handle in an intuitive and not always linear way to make therapeutic decisions. The purpose of this paper is to develop a fuzzy logic (FL) vision of Murray's score variables to enable the measurement of physicians' knowledge, experience and intuition in diagnosing lung injury and test if they followed Murray's equation predictions.

Methods: For a prospective survey carried out among a team of professionals (aged 29 to 53) in a University Hospital Intensive Care Unit, twelve physicians filled in two questionnaires. In the first one they had to define the ranks which should be categorized as normal, moderate and severe for three of four Murray variables. In another questionnaire, which represented all probable combinations of those categories, they had to tick the pulmonary condition as: no injury, mild, moderate, and ARDS. This procedure gave rise to a Fuzzy Inference System designed to provide the degree of severity as sensed by the group.

Results: The survey showed fuzzy frontiers for the categories and fuzzy diagnosis. In all, 45% of the hypothetical patients (n 18,013) were equally diagnosed by the survey and Murray's equation, whereas another 51% was overestimated in one level by the survey. Physicians agreed with 96.5% of ARDS cases according to Murray's test but only 11.6% of its mild cases were equally diagnosed by the survey. Nonlinearity of the survey reasoning (high relevance to gas exchange and chest film) was apparent.

Conclusions: The contiguous categories of the variables confirm the existence of fuzzy frontiers. An overestimation was found in the surveyed group's interpretation of severity. This overestimation was mainly due to the different weight assigned to PO2/FiO2 and chest film variables. The FL approach made it possible to measure knowledge, experience and intuition as they appear in physicians' thinking. FL methodology could overcome a series of restrictions that current tests have due to cut-offs.

Show MeSH

Related in: MedlinePlus

Fuzzification of Murray's equation. (a) PO2/FiO2 membership functions. (b) Compliance membership functions. (c) PEEP membership functions. These are the membership functions of an equivalent fuzzy inference system that responds as Murray test does, obtained through a neuro-fuzzy adaptative inference system imposing the same number of categories as the Survey fuzzy system. Murray's bands are shown as in figure 1.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC2991274&req=5

Figure 3: Fuzzification of Murray's equation. (a) PO2/FiO2 membership functions. (b) Compliance membership functions. (c) PEEP membership functions. These are the membership functions of an equivalent fuzzy inference system that responds as Murray test does, obtained through a neuro-fuzzy adaptative inference system imposing the same number of categories as the Survey fuzzy system. Murray's bands are shown as in figure 1.

Mentions: Based on the Murray output for our input cases, we went all the way in the opposite direction and obtained its corresponding FIS by appealing to a neuro-fuzzy adaptative inference system conditioned to have the same number of categories as the survey's fuzzy system and with the scores predicted by Murray's score for the 18,013 quadruplets as the output. Figure 3 shows the remarkable symmetry of its membership functions. The reason must be sought in the fact that a linear equation, with no weights in any of its four variables, is being represented. Coincidence between Murray and his FIS appeared in 86% of the cases, while the rest were equally distributed at both sides of the central bar, supposing a histogram as the one shown in Figure 2.


Making it possible to measure knowledge, experience and intuition in diagnosing lung injury severity: a fuzzy logic vision based on the Murray score.

D'Negri CE, De Vito EL - BMC Med Inform Decis Mak (2010)

Fuzzification of Murray's equation. (a) PO2/FiO2 membership functions. (b) Compliance membership functions. (c) PEEP membership functions. These are the membership functions of an equivalent fuzzy inference system that responds as Murray test does, obtained through a neuro-fuzzy adaptative inference system imposing the same number of categories as the Survey fuzzy system. Murray's bands are shown as in figure 1.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC2991274&req=5

Figure 3: Fuzzification of Murray's equation. (a) PO2/FiO2 membership functions. (b) Compliance membership functions. (c) PEEP membership functions. These are the membership functions of an equivalent fuzzy inference system that responds as Murray test does, obtained through a neuro-fuzzy adaptative inference system imposing the same number of categories as the Survey fuzzy system. Murray's bands are shown as in figure 1.
Mentions: Based on the Murray output for our input cases, we went all the way in the opposite direction and obtained its corresponding FIS by appealing to a neuro-fuzzy adaptative inference system conditioned to have the same number of categories as the survey's fuzzy system and with the scores predicted by Murray's score for the 18,013 quadruplets as the output. Figure 3 shows the remarkable symmetry of its membership functions. The reason must be sought in the fact that a linear equation, with no weights in any of its four variables, is being represented. Coincidence between Murray and his FIS appeared in 86% of the cases, while the rest were equally distributed at both sides of the central bar, supposing a histogram as the one shown in Figure 2.

Bottom Line: Murray score is the result of an equation that gives all its variables the same linear contribution and weight and makes use of consented cut-offs.An overestimation was found in the surveyed group's interpretation of severity.FL methodology could overcome a series of restrictions that current tests have due to cut-offs.

View Article: PubMed Central - HTML - PubMed

Affiliation: Consejo Nacional de Investigaciones Científicas y Técnicas, Combatientes de Malvinas 3150, CP 1427, Buenos Aires, Argentina. cdnegri@lanari.fmed.uba.ar

ABSTRACT

Background: Murray score is the result of an equation that gives all its variables the same linear contribution and weight and makes use of consented cut-offs. Everyday physicians' vocabulary is full of terms (adjectives) like: little, small, low, high, etc. that they handle in an intuitive and not always linear way to make therapeutic decisions. The purpose of this paper is to develop a fuzzy logic (FL) vision of Murray's score variables to enable the measurement of physicians' knowledge, experience and intuition in diagnosing lung injury and test if they followed Murray's equation predictions.

Methods: For a prospective survey carried out among a team of professionals (aged 29 to 53) in a University Hospital Intensive Care Unit, twelve physicians filled in two questionnaires. In the first one they had to define the ranks which should be categorized as normal, moderate and severe for three of four Murray variables. In another questionnaire, which represented all probable combinations of those categories, they had to tick the pulmonary condition as: no injury, mild, moderate, and ARDS. This procedure gave rise to a Fuzzy Inference System designed to provide the degree of severity as sensed by the group.

Results: The survey showed fuzzy frontiers for the categories and fuzzy diagnosis. In all, 45% of the hypothetical patients (n 18,013) were equally diagnosed by the survey and Murray's equation, whereas another 51% was overestimated in one level by the survey. Physicians agreed with 96.5% of ARDS cases according to Murray's test but only 11.6% of its mild cases were equally diagnosed by the survey. Nonlinearity of the survey reasoning (high relevance to gas exchange and chest film) was apparent.

Conclusions: The contiguous categories of the variables confirm the existence of fuzzy frontiers. An overestimation was found in the surveyed group's interpretation of severity. This overestimation was mainly due to the different weight assigned to PO2/FiO2 and chest film variables. The FL approach made it possible to measure knowledge, experience and intuition as they appear in physicians' thinking. FL methodology could overcome a series of restrictions that current tests have due to cut-offs.

Show MeSH
Related in: MedlinePlus