Adriana Scandurra

Lung recruitment and positive end-expiratory pressure have different effects on CO2 elimination in healthy and sick lungs.

BACKGROUND:We studied the effects that the lung recruitment maneuver (RM) and positive end-expiratory pressure (PEEP) have on the elimination of CO2 per breath (VTCO2,br). METHODS:In 7 healthy and 7 lung-lavaged pigs at constant ventilation, PEEP was increased from 0 to 18 cm H2O and then decreased to 0 in steps of 6 cm H2O every 10 minutes. Cycling RMs with plateau pressure/PEEP of 40/20 (healthy) and 50/25 (lavaged) cm H2O were applied for 2 minutes between 18-PEEP steps. Volumetric capnography, respiratory mechanics, blood gas, and hemodynamic data were recorded. RESULTS:In healthy lungs before the RM, VTCO2,br was inversely proportional to PEEP decreasing from 4.0 (3.6–4.4) mL (median and interquartile range) at 0-PEEP to 3.1 (2.8–3.4) mL at 18-PEEP (P < 0.05). After the RM, VTCO2,br increased from 3.3 (3–3.6) mL at 18-PEEP to 4.0 (3.5–4.5) mL at 0-PEEP (P < 0.05). In lavaged lungs before the RM, VTCO2,br increased initially from 2.0 (1.7–2.3) mL at 0-PEEP to 2.6 (2.2–3) mL at 12-PEEP (P < 0.05) but then decreased to 2.4 (2–2.8) mL when PEEP was increased further to 18 cm H2O (P < 0.05). After the RM, the highest VTCO2,br of 2.9 (2.1–3.7) mL was observed at 12-PEEP and then decreased to 2.5 (1.9–3.1) mL at 0-PEEP (P < 0.05). VTCO2,br was directly related to changes in lung perfusion, the area of gas exchange, and alveolar ventilation but inversely related to changes in dead space. CONCLUSIONS:CO2 elimination by the lungs was dependent on PEEP and recruitment and showed major differences between healthy and lavaged lungs.

Noninvasive monitoring of lung recruitment maneuvers in morbidly obese patients: the role of pulse oximetry and volumetric capnography

BACKGROUND:We conducted this study to determine whether pulse oximetry and volumetric capnography (VCap) can determine the opening and closing pressures of lungs of anesthetized morbidly obese patients. METHODS:Twenty morbidly obese patients undergoing laparoscopic bariatric surgery with capnoperitoneum were studied. A lung recruitment maneuver was performed in pressure control ventilation as follows: (1) During an ascending limb, the lungs’ opening pressure was detected. After increasing positive end-expiratory pressure (PEEP) from 8 to 16 cm H2O, fraction of inspired oxygen (FIO2) was decreased until pulse oximetric arterial saturation (SpO2) was <92%. Thereafter, end-inspiratory pressure was increased in steps of 2 cm H2O, from 36 to a maximum of 50 cm H2O. The opening pressure was attained when SpO2 exceeded 97%. (2) During a subsequent decreasing limb, the lungs’ closing pressure was identified. PEEP was decreased from 22 to 10 cm H2O in steps of 2 cm H2O. The closing pressure was determined as the PEEP value at which respiratory compliance decreased from its maximum value. We continuously recorded lung mechanics, SpO2, and VCap. RESULTS:The lungs’ opening pressures were detected at 44 (4) cm H2O (median and interquartile range) and the closing pressure at 14 (2) cm H2O. Therefore, the level of PEEP that kept the lungs without collapse was found to be 16 (3) cm H2O. Using respiratory compliance as a reference, receiver operating characteristic analysis showed that SpO2 (area under the curve [AUC] 0.80 [SE 0.07], sensitivity 0.65, and specificity 0.94), the elimination of CO2 per breath (AUC 0.91 [SE 0.05], sensitivity 0.85, and specificity 0.98), and Bohr’s dead space (AUC 0.83 [SE 0.06], sensitivity 0.70, and specificity 0.95] were relatively accurate for detecting lung collapse during the decreasing limb of a recruitment maneuver. CONCLUSIONS:Lung recruitment in morbidly obese patients could be effectively monitored by combining noninvasive pulse oximetry and VCap. SpO2, the elimination of CO2, and Bohr’s dead space detected the individual’s opening and closing pressures.

Automatic design of interpretable fuzzy predicate systems for clustering using self-organizing maps

Abstract In the area of pattern recognition, clustering algorithms are a family of unsupervised classifiers designed with the aim to discover unrevealed structures in the data. While this is a never ending research topic, many methods have been developed with good theoretical and practical properties. One of such methods is based on self organizing maps (SOM), which have been successfully used for data clustering, using a two levels clustering approach. Newer on the field, clustering systems based on fuzzy logic improve the performance of traditional approaches. In this paper we combine both approaches. Most of the previous works on fuzzy clustering are based on fuzzy inference systems, but we propose the design of a new clustering system in which we use predicate fuzzy logic to perform the clustering task, being automatically designed based on data. Given a datum, degrees of truth of fuzzy predicates associated with each cluster are computed using continuous membership functions defined over data features. The predicate with the maximum degree of truth determines the cluster to be assigned. Knowledge is discovered from data, obtained using the SOM generalization aptitude and taking advantage of the well-known SOM abilities to discover natural data grouping when compared with direct clustering. In addition, the proposed approach adds linguistic interpretability when membership functions are analyzed by a field expert. We also present how this approach can be used to deal with partitioned data. Results show that clustering accuracy obtained is high and it outperforms other methods in the majority of datasets tested.

The Sensitivity and Specificity of Pulmonary Carbon Dioxide Elimination for Noninvasive Assessment of Fluid Responsiveness.

BACKGROUND:We sought to determine whether the response of pulmonary elimination of CO2 (VCO2) to a sudden increase in positive end-expiratory pressure (PEEP) could predict fluid responsiveness and serve as a noninvasive surrogate for cardiac index (CI). METHODS:Fifty-two patients undergoing cardiovascular surgery were included in this study. By using a constant-flow ventilation mode, we performed a PEEP challenge of 1-minute increase in PEEP from 5 to 10 cm H2O. At PEEP of 5 cm H2O, patients were preloaded with 500 mL IV saline solution after which a second PEEP challenge was performed. Patients in whom fluid administration increased CI by ≥15% from the individual baseline value were defined as volume responders. Beat-by-beat CI was derived from arterial pulse contour analysis, and breath-by-breath VCO2 data were collected during the protocol. The sensitivity and specificity of VCO2 for detecting the fluid responders according to CI was performed by the receiver operating characteristic curves. RESULTS:Twenty-one of 52 patients were identified as fluid responders (40%). The PEEP maneuver before fluid administration decreased CI from 2.65 ± 0.34 to 2.21 ± 0.32 L/min/m2 (P = 0.0011) and VCO2 from 150 ± 23 to 123 ± 23 mL/min (P = 0.0036) in responders, whereas the changes in CI and VCO2 were not significant in nonresponders. The PEEP challenge after fluid administration induced no significant changes in CI and VCO2, in neither responders nor nonresponders. PEEP-induced decreases in CI and VCO2 before fluid administration were well correlated (r2 = 0.75, P < 0.0001) but not thereafter. The area under the receiver operating characteristic curves for a PEEP-induced decrease in &Dgr;CI and &Dgr;VCO2 was 0.99, with a 95% confidence interval from 0.96 to 0.99 for &Dgr;CI and from 0.97 to 0.99 for &Dgr;VCO2. During the PEEP challenge, a decrease in VCO2 by 11% predicted fluid responsiveness with a sensitivity of 0.90 (95% confidence interval, 0.87–0.93) and a specificity of 0.95 (95% confidence interval, 0.92–0.98). CONCLUSIONS:PEEP-induced changes in VCO2 predicted fluid responsiveness with accuracy in patients undergoing cardiac surgery.

Using SOM as a Tool for Automated Design of Clustering Systems Based on Fuzzy Predicates

Clustering task is a never-ending research topic. New methods are permanently proposed. In particular, Fuzzy Logic and Self-organizing Maps and their mutual cooperation have demonstrated to be interesting paradigms. We propose a general approach to obtain membership functions for a ranked clustering system based on fuzzy predicates logical operations, considering Gaussian-shaped curves. We find membership functions parameters from trained Self-organizing Maps, which generalize the statistical characteristics of data. The system is self-configured and it has the advantages of other fuzzy approaches. Clustering quality is assessed by labeled data, which allow computing accuracy. The proposal must be tested with more real datasets, though the preliminary results obtained in well-known datasets suggest that it is a promising clustering scheme.

Improvements in the visualization of segmented areas of patterns of dynamic laser speckle

This paper proposes a method to visualize different regions into image of biospeckle patterns using Self-Organizing Maps. Images are obtained from sequences of laser speckle images of biological specimens. The dynamic speckle is a phenomenon that occurs when a beam of coherent light illuminates a sample in which there is some type of activity, not visible, which results in a variable pattern over time. Self-Organizing Maps have shown an efficient behavior for the identification of regions according to the activity of the phenomenon involved. In this paper we show results obtained in the segmentation of regions in corn seeds, particularly the detection of the floury zone.

Optimization of arterial age prediction models based in pulse wave

We propose the detection of early arterial ageing through a prediction model of arterial age based in the coherence assumption between the pulse wave morphology and the patients chronological age. Whereas we evaluate several methods, a Sugeno fuzzy inference system is selected. Models optimization is approached using hybrid methods: parameter adaptation with Artificial Neural Networks and Genetic Algorithms. Features selection was performed according with their projection on main factors of the Principal Components Analysis. The model performance was tested using the bootstrap error type .632E. The model presented an error smaller than 8.5%. This result encourages including this process as a diagnosis module into the device for pulse analysis that has been developed by the Bioengineering Laboratory staff.

Dynamic speckle image segmentation using self-organizing maps

The aim of this work is to build a computational model able to automatically identify, after training, dynamic speckle pattern regions with similar properties. The process is carried out using a set of descriptors applied to the intensity variations with time in every pixel of a speckle image sequence. An image obtained by projecting a self-organized map is converted into regions of similar activity that can be easily distinguished. We propose a general procedure that could be applied to numerous situations. As examples we show different situations: (a) an activity test in a simplified situation; (b) a non-biological example and (c) biological active specimens. The results obtained are encouraging; they significantly improve upon those obtained using a single descriptor and will eventually permit automatic quantitative assessment.