Ability and efficiency of an automatic analysis software to measure microvascular parameters

Andrea Carsetti; Hollmann D Aya; Silvia Pierantozzi; Simone Bazurro; Abele Donati; Andrew Rhodes; Maurizio Cecconi

doi:10.1007/s10877-016-9928-3

Ability and efficiency of an automatic analysis software to measure microvascular parameters

J Clin Monit Comput. 2017 Aug;31(4):669-676. doi: 10.1007/s10877-016-9928-3. Epub 2016 Sep 1.

Authors

Andrea Carsetti^{1

2}, Hollmann D Aya¹, Silvia Pierantozzi^{1

2}, Simone Bazurro¹, Abele Donati², Andrew Rhodes^{1

3}, Maurizio Cecconi^{4

5}

Affiliations

¹ Department of Intensive Care Medicine, St George's University Hospitals NHS Foundation Trust, Blackshaw Road, London, SW17 0QT, UK.
² Department of Biomedical Sciences and Public Health, Università Politecnica delle Marche, Ancona, Italy.
³ St George's University of London, London, UK.
⁴ Department of Intensive Care Medicine, St George's University Hospitals NHS Foundation Trust, Blackshaw Road, London, SW17 0QT, UK. m.cecconi@nhs.net.
⁵ St George's University of London, London, UK. m.cecconi@nhs.net.

PMID: 27586243
DOI: 10.1007/s10877-016-9928-3

Abstract

Analysis of the microcirculation is currently performed offline, is time consuming and operator dependent. The aim of this study was to assess the ability and efficiency of the automatic analysis software CytoCamTools 1.7.12 (CC) to measure microvascular parameters in comparison with Automated Vascular Analysis (AVA) software 3.2. 22 patients admitted to the cardiothoracic intensive care unit following cardiac surgery were prospectively enrolled. Sublingual microcirculatory videos were analysed using AVA and CC software. The total vessel density (TVD) for small vessels, perfused vessel density (PVD) and proportion of perfused vessels (PPV) were calculated. Blood flow was assessed using the microvascular flow index (MFI) for AVA software and the averaged perfused speed indicator (APSI) for the CC software. The duration of the analysis was also recorded. Eighty-four videos from 22 patients were analysed. The bias between TVD-CC and TVD-AVA was 2.20 mm/mm² (95 % CI 1.37-3.03) with limits of agreement (LOA) of -4.39 (95 % CI -5.66 to -3.16) and 8.79 (95 % CI 7.50-10.01) mm/mm². The percentage error (PE) for TVD was ±32.2 %. TVD was positively correlated between CC and AVA (r = 0.74, p < 0.001). The bias between PVD-CC and PVD-AVA was 6.54 mm/mm² (95 % CI 5.60-7.48) with LOA of -4.25 (95 % CI -8.48 to -0.02) and 17.34 (95 % CI 13.11-21.57) mm/mm². The PE for PVD was ±61.2 %. PVD was positively correlated between CC and AVA (r = 0.66, p < 0.001). The median PPV-AVA was significantly higher than the median PPV-CC [97.39 % (95.25, 100 %) vs. 81.65 % (61.97, 88.99), p < 0.0001]. MFI categories cannot estimate or predict APSI values (p = 0.45). The time required for the analysis was shorter with CC than with AVA system [2'42″ (2'12″, 3'31″) vs. 16'12″ (13'38″, 17'57″), p < 0.001]. TVD is comparable between the two softwares, although faster with CC software. The values for PVD and PPV are not interchangeable given the different approach to assess microcirculatory flow.

Keywords: Analysis; CytoCam; Incident dark field imaging; Microcirculation.

MeSH terms

Automation
Blood Flow Velocity
Blood Vessels / diagnostic imaging
Critical Care
Humans
Image Processing, Computer-Assisted*
Intensive Care Units
Microcirculation*
Microscopy, Video
Observer Variation
Pattern Recognition, Automated*
Perfusion
Prospective Studies
Reproducibility of Results
Software*