objectives methods results conclusions

Automated reading of urine cultures using a novel
image analysis device
J.H. Glasson1, R. Hill2, M.J. Summerford1
1LBT Innovations Ltd, Adelaide, South Australia,
2Australian Centre for Visual Technologies, University of Adelaide, South Australia
OBJECTIVES
Contact: [email protected]
Fig 1. Report screen showing summary results and alerts
In recent years, the application of image analysis technologies within the clinical
laboratory, particularly in the cell-based fields of haematology, anatomical pathology and
cytopathology, has increased.
In the microbiology laboratory, the agar plate remains an important diagnostic tool and
alternate technologies have not yet provided a comprehensive replacement. This is
despite the reality that agar plates require numbers of highly trained staff, time and a great
deal of laboratory space.
The Automated Plate Assessment System (APAS®) is a novel device dedicated to
screening agar plates. Using image analysis technology, it is able to detect colony growth,
enumerate the various colony types detected and apply standard microbiological rule sets
to sort each plate into categories suitable for further processing. In this study, APAS®
also applied an expert system specifically for the interpretation of urine cultures.
A total of 526 urine specimens from two separate laboratories were cultured and analysed
by an APAS® prototype device. These results were then compared with those generated
by experienced microbiologists.
Fig 2. Plate report review screen for a urine sample plated onto Blood Agar and Brilliance UTI Agar
METHODS
In one laboratory, two hundred and seventy six urines submitted for routine culture were
inoculated onto Horse Blood Agar and Brilliance™ UTI Clarity Agar bi-plates (Thermo
Fisher Scientific, Thebarton, Australia). In the other laboratory, two hundred and fifty
urines were inoculated onto Horse Blood Agar and MacConkey (No Salt) Agar bi-plates
(Thermo Fisher Scientific, Thebarton, Australia).
Following incubation, the plates were read by experienced microbiologists and then
analysed by a prototype APAS® instrument (LBT Innovations Ltd, Adelaide, Australia).
The results were then compared.
RESULTS
Microbiologists reported that 178 urines showed growth at >=10^5 CFU/ml, 164 at 10^4
CFU/ml, 116 at 10^3 CFU/ml and 68 with no growth. The instrument showed agreement
in 512/526 (97%) cases.
Fig 3. Plate report review screen for a urine sample plated onto Blood Agar and MacConkey Agar
The differences were seen where APAS® overestimated colony counts predominantly in
the “No Growth” and 10^3 CFU/ml regions. In each case, the instrument determined a
higher count than the microbiologists.
A summary of these results can be seen below in Table 1.
Table 1. Comparison of APAS® urine colony counts with manually read plates
Colony Count (CFU/ml)
No Growth
10^3
10^4
10^5
Total
Counts determined by
Microbiologists
68
116
164
178
256
Counts determined by
APAS®
62
122
163
179
256
Preliminary colony identification on morphological and biochemical reactions from the
indicator agars for the main isolates showed agreement with 94% of isolates from the
MacConkey agar and 98% from the Chromogenic agar.
During the study, the instrument produced 33 alerts for the presence of swarming Proteus
spp. and 29 alerts for colony counts close to the 10^3, 10^4 or 10^5 CFU/ml borderlines.
In 47 cases, the system also flagged the presence of haemolysis.
CONCLUSIONS
This study demonstrated the ability of image analysis technology to read and interpret
cultures on agar plates.
Using APAS® to screen routine urine cultures produced results with an overall diagnostic
sensitivity of 97% and specificity of 95%.
APAS® was able to detect and enumerate colonies and provided a preliminary identification
for the primary isolates on the different agar combinations used by the two laboratories.
Results for each case were summarised in a report (Figure 1) which contained links to the
images of the individual plates (Figures 2 and 3).
The incorporation of an expert system with standard reporting rules further enhanced the
usefulness of the system.
When the automated results from the two agars were combined and the interpretive rule
set applied by the expert system the diagnostic sensitivity for these 526 urines was 97%
and the specificity 95%.
Acknowledgements
We thank Professor J. Turnidge and Dr A. Butcher of SA Pathology, Adelaide, South Australia and Dr S Giglio of
Healthscope Pathology, Wayville, South Australia. The assistance provided by the staff of the two laboratories was
particularly appreciated.