Preprototype of an Automated Microbial Detection and Identification System: a Developmental Investigation

(1)

Preprototype

of

an

Automated Microbial Detection and

Identification System:

a

Developmental Investigation

ALEX C.SONNENWIRTH

Departmentsof Microbiology and Immunology, and Pathology, Washington University School of Medicine; and Departments of PathologyandLaboratory Medicine, The JewishHospital of St. Louis, St. Louis,

Missouri 63110*

Received forpublication21 April1977

The AutoMicrobic System is an automated, computerized instrument that uses highly selective media and an optical system for detection, enumeration, and identification of bacteriaandsomeyeastsin13h. Apreprototypeinstrument (AutoMicrobic System-1) and its urine culture kit (Identi-Pak), developed for the detection, enumeration, and identification of eight species or groups of bacteria and of Candida species and Torulopsis glabrata in urine specimens, wasevaluated during its development. An overall agreement ofapproximately 90% between the preprototype instrument and conventional (manual) culture methodshas been obtained both with 1,473 seeded (simulated)and 1,688 clinical (mono- orpolymicrobial) specimens containing 70,000 (or more) colony-forming

unitspermlofEscherichiacoli,Klebsiella-Enterobacter species, Proteus species, Citrobacterfreundii, Serratia species,group Denterococci, oryeasts (Candida

speciesand T.glabrata). Loweragreementsinidentificationwereobtained with

Pseudomonasaeruginosa-containing (average of75%inclinical specimens) and Staphylococcus aureus-containing (76%) specimens. Comparison of specimens testedsimultaneouslyintwopreprototypesystems resulted inc4%disagreement;

true negativity agreements in all specimen groups tested were at least 94%. Among problems remaining are adaptation of system for specimens other than urine, improvementofsensitivityfor P. aeruginosaand S.aureus, and standard-ization of manual methods used forcomparisonandvalidation.

A variety of mechanized or automated sys-tems for detection, characterization, and

drug-susceptibility testing of microorganisms have beenrecently describedorareinthe process of

development(2).Practically allof these systems were originally designed to perform a single

functiononly, e.g., (i) detection, bysuch meth-ods as differential light scattering, radiometric measurement of

14CO2

fromlabeled substrates,

bioluminescence assay for adenosine

triphos-phate, measurement of electrical impedance, microcalorimetry, and gas chromatography; (ii)

identification, bycombinations of pyrolysis and eithergas-liquid chromatographyor mass spec-trometry, gaschromatography, opticalscanning ofgrowth stimulation orinhibition, microcalor-imetry, and forward light scattering (electrical impedance nowalsobeinginvestigated for pos-sible useinidentification); or (iii) antimicrobial

sensitivity testing, e.g., byforwardlight

scatter-ing forqualitativeorquantitativetesting in 4 to 5 h.

Arecentdevelopment, the AutoMicrobic

Sys-tem(AMS;McDonnellDouglas Corp.,St.Louis, Mo.) (1), was designed for the detection,

*enumeration, identification,

and

drug-suscepti-bilitytestingofmicroorganismsinclinical spec-imens, withallofits functions to be performed simultaneously or consecutively by the same

instrument. Ituses anovelarrayofhighly

selec-tive media in whichamixtureof substrates and inhibitors is used to permit the growth of one

or agroupofcloselyrelatedmicroorganisms, to

concurrently inhibit growth of any other orga-nismspresent, andtoenumeratethe total num-ber oforganismsin thesample. Thelyophilized

media are incorporated in wells of a

small,

sealed, disposable plastic cuvette (card),

inocu-lated,andincubatedinanautomated instrument

equipped with an optical system. The optical

system monitors light transmission changes in

the card wells and transmits optical

measure-mentsto aminicomputer,whichcompares them withpresetthresholds and thendisplaysresults in 4 to 13 h. A preprototype AMS instrument and its urine culture kit (Identi-Pak), for the detection, enumeration, and identification of eightspeciesorgroups of bacteria and of yeasts

(Candida species and Torulopsis glabrata) in urine specimenswasinvestigated during its de-velopment.

This report presents the results of a study

400

on February 7, 2020 by guest

http://jcm.asm.org/

(2)

AUTOMATED BACTERIAL DETECTION/IDENTIFICATION

designedto(i) comparethe AMS preprototype

system withconventional culture methods, (ii) ascertainitsreproducibility by testingidentical splitspecimensintwoidentical instruments, and (iii) define the performance limitsof the AMS

both with seeded (simulated) urines and with clinicalurinespecimens.

MATERIALS AND METHODS

Organisms. The organisms used for preparation of90% of simulatedspecimenswere currentisolates from clinicalspecimensobtained fromthree St.Louis

areahospitalsthatwerereidentified byconventional

methods (3) andmaintained onTrypticasesoy agar

slants;fortheremaining 10% ofsimulatedspecimens, laboratory stockstrainswereused.Single strainswere

employedforseeding 1,326 stimulated specimens, and 2organisms wereusedforseeding each of147 simu-latedspecimens.

Specimens. (i) Seeded. Simulated urine

speci-menswereprepared by seeding pooled, filter-sterilized

urinefromhealthyhuman malestoafinal concentra-tionof106to 107 colony-forming units (CFU)perml.

All seededspecimenswerecoded, and the identity of

the organism(s) was not known to the instrument's operator until thecodewasbrokenaftercompletion

of both the instrumentrunandconventionalculture.

Seededspecimenswereprocessedwithin 2 hof

prep-aration. A total of 1,473 simulated specimens were

tested.

(ii) Clinical.Thesewereunselected, clean, voided

urine specimens from patients hospitalized at The Jewish Hospitalof St. Louis.Specimenswere

refrig-eratedforamaximumof 12h,with thegreatmajority

tested within 4 h of refrigeration. A total of 1,688 clinicalspecimenswereused.

Conventionalcultureprotocol. Serial 10-fold di-lutions ofurinesamples (seededorclinical) were

pre-pared with saline, and 0.1-ml amounts ofdilutions

were spreadonthe surface ofMacConkey agarand

Trypticasesoybloodagarplateswithaglass

"hockey-stick."Afterovernightincubationat35°C,countswere

performed(i) manuallyforaperiodof3months with

aQuebec colonycounterand(ii)for thenext9months withaModel 480 Artek-Fisher AutomaticBacterial

colony counter. The colony counter was regularly

checked with manualcounts.Fermentative

gram-neg-ative rodswereidentified withtheAPI 20Ekit,and

oxidizers and nonfermenterswereidentified withthe

use ofconventional techniques (3), including oxida-tion-fermentation media.Coagulase testswith staph-ylococciweredone bythe tube method; bile-esculin

agar(40% bile)and 6.5%NaCltests(forgrowth)were

used for presumed enterococci, and germ-tube, fer-mentation,and assimilationtests wereperformedon

yeastlike organisms.

Preprototypeinstrument. Thepreprototype in-strument was similar to the production model de-scribedby Aldridge,etal.(1) with the following

excep-tions:(i)ithasonlya30-specimen capacity;(ii)results were displayed by teletypewriter instead of

electro-static printer, (iii) thecomputer (Digital Equipment Corp.model PDP 11-05) wasmodifiedwithtapesfor collection of time history profiles; and (iv) onlyone

light-emitting diode pair was available for reading

each growth well, resulting in a relative inability of the instrument to discriminate between growth or bubblesinthewells. Thedetection threshold of the instrumentwasset at 7x 104CFU/ml.

Urine test kit. The urine Identi-Pak card con-tainedseparate,specific,selective media(1)for Esch-erichiacoli,Pseudomonas aeruginosa,Proteus spe-cies, Citrobacter freundii, Serratia species, Klebsi-ella-Enterobactergroups, yeasts(Candidaspecies,T. glabrata), Staphylococcus aureus, and group D en-terococci andanenriched broth for positive growth

controlandenumeration (total count).The maximum (liquid) volume of each broth in the card wells was 32

p1.

Atwo-reservoirsample injectorfor diluent and specimenmixingandinoculation of the cardwellswas included in the kit; one reservoir (chamber) feeds throughaneedletoinoculate thespecificgrowth wells,

and the other feeds through a separate needle to inoculate the enumeration wells.

Instrumenttestprotocol.Thediluent usedwas unbuffered distilledwaterwith 0.5% NaCl. To cham-ber A of the sample injector, containing 1.8 ml of diluent, 0.2 ml of undiluted urine was added. After

mixing,0.05mlwastransferredtochamberB, which contained5ml of diluent. The finalspecimendilution in chamber A was 1:10, and in chamber B, 1:1,000. Beforefilling,each cardwasappropriatelymarked for instrument recognition with Arabic numerals. The

joined test units (injector and card) were placed in thefillingmodule, where the diluted urinewas trans-ferred from theinjectorthroughthetwoneedles into thecardwells(the needles piercing hermetic seals in the card) by apneumatically controlled

evacuation-repressurizationprocess(5min). Afterfilling,thecards wereseparated from the injector and loaded into the reader-incubator. At this point, the instrument as-sumedcontrol;every30minit recordedreadings (for atotal of12h),printed interim status reports every h, andattheend of13helapsedtime, furnished final reports.

Comparative study.Thenumber, nature, and des-ignation ofsimulatedandclinical specimens tested is

shown in Table 1. Group I, a total of 929 seeded urines,wastested withoneinstrument(AMS-1) and withconventional manual culture methods; group III, 544seededurines,wastested with AMS-1 and another identical preprototype instrument (AMS-2) in the manufacturer'slaboratory, employing split specimens and conventional culture methods. The split speci-mensweretransportedonice tothe second laboratory andweretested within3hafterdividing the specimen.

Group 11(648 clinical urines) and group IV (1,040 clinical urines) were tested exactly as groups I and III (Table1).Wheneverdiscrepancies were noted be-tween instrumentand manualmethod, the cardwells wereenteredwithasyringeand needle and cultured by conventional methods; presence of bubbles was noted and recorded.

Comparison of manual and instrument cultures weremadebydetermining correlationsbased on two formulas:

(i)percentpositivecorrelation:

T+

(T+)

+(F-)x

0

401

VOL. 6, 1977

on February 7, 2020 by guest

http://jcm.asm.org/

(3)

(ii)percent negativecorrelation:

T- x100

(T-)+(F+)

Percent positive correlation indicates the successof the AMS instrument in detecting a particular

orga-nism (T+,truepositive) when it is presentin

concen-trations of 7x 104CFU/mlormore(as shown by the

manual culturemethod); i.e., it denotes the sensitivity ofthe instrument. Percent negativecorrelation, where T- representsthenumberof allnegative challenged media, shows the accuracyof AMS in notreporting

an organism when, in fact, the organism was not present;i.e., it signifiesthespecificityofAMS.Results considered asfalse negative (F-) AMS readings (no detection or identification ofaparticular organism) werecalculated by deducting AMS positive readings

from positive conventional culture method results (considered as100%), while falsepositiveAMS read-ings (F+) representedthose notobtainedby

conven-tionalculture.

RESULTS

Detailed results obtained with 929 seeded ur-ines (group I) are shown in Table 2. Positive

TABLE 1. Specimens tested with AMSpreprototype Designa- No. Specimens Compared

tion with:

Group I 929 Seeded (simulated) CMa urine

GroupIlIP 544 Seeded (simulated) CM,

urine AMS-2¢

Total 1,473

GroupII 648 Clinicalurine(CVS)Y CM GroupIVh 1,040 Clinical urine(CVS)" CM,

AMS-2 Total 1,688

{CM, Conventional(manual)method. bSplit specimens.

"Second identicalpreprototypeinstrument.

' Clean, voidedspecimen.

correlation(sensitivity) of the AMSsystem

com-pared with manual culturewas91to 100%with

Serratia, Klebsiella-Enterobacter, the yeasts,

E. coli, C. freundii, and Proteus, withover90%

agreement for the majority of the organisms.

Negative correlation (specificity)was98to100%

forallorganisms testedexceptS. aureus (91%).

False negative readings by the AMS as

com-pared with conventional culture resultsoccurred

with c1.4% of Klebsiella-Enterobacter group

and the yeasts, c9% ofgroup Denterococci, P. aeruginosa, and E. coli (6.6%), and c20% of S.

aureus, C. freundii, and Proteus species (13%),

foran overall (weighted)false negativityrateof

4.9%. False positive readings ranged from 0 to 4% with the exception ofgroup D enterococci (9%), P. aeruginosa (15%), and S. aureus

(20%), for an overall (weighted) false positivity

rate of 3.5%. A second group ofseeded urines (Table 3) yielded positive correlations ofover 90% for allorganismstested,exceptforS.aureus (84%); negative correlationswere95 to 100%.

Clinical urinespecimenstestresultsareshown in Tables 4 and 5. In group II (648 clinical urines), positive correlation (sensitivity) of c91%

at levels of 70,000 CFU/ml or greater was achievedbythe AMS indetecting and identify-ing Klebsiella-Enterobacter, group D

entero-cocci, Proteus species and the yeasts; positive correlation was 88% with E. coli, 77% with P. aeruginosa, and 66% withasmallnumberof S.

aureus(Table 4), withnoSerratiaorC.freundii

isolated.IngroupIV (1,040 split clinicalurines),

percent positive correlationswereconsiderably higher for E. coli (98%) and S. aureus (86%) and lowerforgroupD enterococci (80%) and P. aeruginosa (73%), with an overall agreement

range of 87 to 100% for the other organisms

(Table 5).

TABLE2. Group I-seeded urinesa Challengesh

Organisms Correlation

T+ F- F+

T-%Negative %Positive

P. aeruginosa 77 8 13 831 91 98

Proteus 140 21 1 767 87 99

C.freundii 9 2 0 918 82 100

Serratia 10 0 0 61 100 100

E. coli 150 10 7 762 94 99

Klebsiella-Enterobacter 211 1 8 709 99 99

Yeast 68 1 2 858 98 99

S.aureus 4 1 1 196 80 91

Group D enterococci 10 1 1 103 91 99

Enumeration 728 173 1 27 81 96

Positivecontrol 860 52 0 17 94 100

a929specimens.

Number oftruepositive (T+),falsenegative (F-),falsepositive (F+),andtruenegative (T-) challenges.

on February 7, 2020 by guest

http://jcm.asm.org/

(4)

An overall summary of results obtained in

comparisonoftwoinstruments(reproducibility) with seeded and clinical specimensin termsof

percent positive correlation is shown in Table 6. The two instruments (AMS-1 and AMS-2)

wereusedby different workers in different

lab-oratories employing split clinical specimens. With seeded specimens, there was 6% or less

disagreementbetween thetwoinstruments with

Pseudomonas, Serratia, E. coli, and S. aureus

andcompleteagreementwithProteus, C. freun-dii, and Klebsiella-Enterobacter. With clinical specimens, therewas complete agreementwith

C. freundii, Serratia species, and yeasts, 2% disagreement withProteus and E.

coli,

13 and 14%disagreementwithKlebsiella-Enterobacter and S. aureus, respectively, and 21%

disagree-ment with P. aeruginosa. Since the positive correlationsfor eachinstrumentwere basedon comparisons with manual cultures (also

per-formed on each split sample in both

laborato-ries), analysis ofresults of the manual culture in the two laboratories was undertaken. The

resultsindicated that the greatmajority of

dis-agreements between the two instruments, as

expressed byeach percent positivecorrelation,

was due to discrepant results of the manual culturemethods, primarilyofcolonycounts, ob-tainedbythe twolaboratories. Whenthe

man-ual culture results were excluded from

consid-eration,disagreementsbetweenthe two

instru-mentswerereducedto4%orlessoverall.

The distributionandfrequency oforganisms in the clinicalspecimens tested,aslisted in Ta-bles 4 and 5, generally reflect the prevailing

patternoforganismsseeninurinarytract

infec-tions of thepatientpopulationinthisinstitution,

except for E. coli. Of 648 specimensshown in Table 4,atotal of 150yielded E. coli in concen-trations of 7x

104

CFU/mlorhigher,but

instru-TABLE 3. Group III-seededurinese Correlation

Organisms No. of

chal-lenges Posi- %

Neg-tive ative

P.aeruginosa 17 94 96

Proteus 19 100 99

C.freundii 64 97 95

Serratia 163 96 99

E. coli 24 92 100

Klebsiella-Enterobac- 11 100 97 ter

Yeast 28 100 99

S.aureus 90 84 95

GroupD enterococci 120 98 100

Enumeration 463 86 100

Positive control 468 86

a544specimens; split samples. b",Nonegative challenges.

TABLE4. GroupII-clinical urinesa Correlation

Organisms No. ofchal- t

Organisms

~~lenges

Posi- Neg-tive ative

Proteus 41 100 99

C.freundii 0 99

Serratia 0 100

E. coli 81h 88 98

Yeast 5 100 99

S.aureus 3 66 99

GroupDenterococci 16 94 96

Positivecontrol 170 97 91

a648specimens.

ITotal of 150 E. colichallenges-change in media

formu-lation.

TABLE 5. GroupIV-clinicalurinesa Correlation

Organisms No. of

chal-lenges %

Neg-Posi- atv

tive ative

Proteus 73 93 98

C.freundii 4 100 99

Serratia 3 100 99

E. coli 58b 98 98

Yeast 14 100 99

S.aureus 7 86 99

GroupD enterococci 44 80 98

Positive control 342 99 91'

a1040specimens; split samples.

'Totalof270E. colichallenges change in media formu-lation.

cIncludes bubbles.

mentresults for

only

81 are

listed,

since formu-lation of the E. coli selective brothwaschanged

during

the

study, and,

therefore,

results of the first69E.coli

challenges

couldnotbe included.

Likewise,

the second group of 1,040 clinical ur-ines

(Table

5)

yielded

270

specimens

containing E. coli with results of 212

challenges

not in-cluded because of the change in formulation. Among the positive

clinical

specimens, 12.2% containedmorethanone

organism.

Correct enumeration (percentpositive

corre-lation oftotal counts) was achieved

by

the

in-strument in an average of 83.5% of the two

groups of seededurines(range, 81to86%), and

in 86.5% of the two groups of clinical urines (range,85 to88%).

Overall

accuracyof enumer-ation in

negative

challenges

(percent negative

6,1977

on February 7, 2020 by guest

http://jcm.asm.org/

(5)

TABLF, 6. Comparisonofpercentageofpositive correlation betweentwopreprototype instruments(AMS-I versusAMS-2, employingspliturinespecimens)based on conventionalcolony counts

Group III-544seeded GroupIV-1040

clin-specimens No. ofchal- No. of chal- icalspecimens

Organisms

AMS-1 AMS-2 lenges lenges AMS-1 AMS-2

(%) (%) (%)

94 100 17 P.

aeruginosa

15 73 94

100 100 19 Proteus 73 93 91

97 97 64 C.freundii 4 100 100

96 94 163 Serratia 3 100 100

92 96 24a E. coli 58" 98 100

100 100 11 Klebsiella-Enterobacter 64 87 100

100 100 28 Yeast 14 100 100

84 89 90 S.aureus 7 86 100

98 99 120 GroupDenterococci 44 80 86

aTotal of150 E. colichallenges;see text.

hTotal of270E.coli

challenges;

see text.

correlation) was98% withseededurines (range, 96to100%)and 97.5%withclinical urines (range, 97 to98%). Overall detectionratesinthe positive

control broth were 90% for seeded.urines and 99% forclinical urines.

DISCTJSSION

TheAMSpreprototype instrument and urine kit employed in this investigation is the first, and todate only,systemcapable ofperforming detection, enumeration, and identification of

various bacteriaand yeastscommonlyfound in urine in a fully automated mode. It requires

only minimal initial manual handling, with all

steps beyondurinesample dilution andloading performed automatically by the instrument. In thesestudies, involving 1,473seeded urines and 1,688clinical urines, the preprototype AMS

sys-tem wasfound tohave an overall average

sen-sitivity andspecificity ofapproximately 90% in

detecting, enumerating, and correctly identify-ing (at the level of c7 x 104 CFU/ml) eight

different bacterialspecies/groupsandtwoyeast

species/groups

inurine

specimens.

In comparison with conventional (manual) culturemethods,the AMSwashighly successful

in detecting and identifying yeasts (Candida

species and T. glabrata) in both seededurines (average agreement, 99%), and mono- and

po-lymicrobial clinical urines (100% agreement). A

similar pattern was noted with E. coli (where agreementwas91% in seeded and 93% inclinical urines), Klebsiella-Enterobacter species (99.5 and 89%), Proteus species (93.5 and 96.5%), C.

freundii

(89.5 and 100%), and Serratia species (98 and 100%). Agreements below 90% were found in clinical urinesonly withP.aeruginosa

(75%), group D enterococci (87%), and with a small number (10 specimens) of S. aureus

(76%).

In the early portion of the study, a number of

discrepancies

betweenthe AMS andmanual culture results,

expecially

false

positive

AMS findings, weretracedtothe formation of visible

bubbles in the media-containing growth wells, particularlyinthoseservingforselectivegrowth

of P. aeruginosa.Adjustmentsinthe card man-ufacturing processwere made,and the number of discrepancies due to bubble formation did diminish; however, no allowance forthis occur-rence was made in the calculation of percent

positive correlations, a fact that partially ex-plains the low agreement rate

(75%)

with P.

aeruginosainclinicalurines. Anothersourceof

difficulty

withthisorganism wastherate of

02

diffusionthroughtheTeflontapeused for seal-ing the top and bottom of thecard; batches of

Teflon variedin their

permeability.

Some false positives with S. aureus were due to

problems

withrehydration of the S. aureusmedium. It is

noteworthy that the test results reportedhere

wereobtainedwith theuseofmorethanadozen

production

lots oftestkits,thusinevitably

intro-ducingamajor variable. Thismade the

testing

procedure both more realistic and more severe

than itwouldhave been with theuseofa

single

productionbatch oftestkits.

It wasalsofoundthat thepreprototype system

was capable of detecting, and in some cases identifying,certain

organisms

atlevels far below 7x 104

CFU/ml,

althoughthiswasnotexplored

systematically.

In fact, some Proteus

species

were detected and identified at levelsaslowas

3 x 103 CFU/ml. However, it needs to be em-phasized that the threshold for detection and

identification in this study was set at 7 x 104

CFU/ml and that the preprototype instrument and kit was not designed to

perform

at levels lower thanthis.

The use of a computer equipped with time

404

SONNENWIRTH

on February 7, 2020 by guest

http://jcm.asm.org/

(6)

405

history collection tapes afforded some insight

intothe detection andidentification timesversus

population size

relationship.

In

general, high

populationsweredetectedandidentifiedinless time thanlow

populations,

but thiswasnot true in all instances. The system occasionally de-tected and identified organisms in the 1 x 103 to 3 x

10'-CFU/ml

range; in some instances,

false positive AMS results were recorded be-cause the manual culture method missed such

lowcounts, withthepresence of the organisms

in the proper well (selective medium) proven

by conventional culture of the well. However, the AMS did correctly report enumerations of lessthan70,000/mlinsuchcases.

Comparative testing ofspliturinespecimens

in twoseparate, identical AMSinstruments

re-vealed yetanothervariable. Despite theuse of

astandardized manualculturemethod, discrep-ancies were noted in some

colony

counts ob-tained inthe twolaboratories ranging from0.4 to 1.4logs. Thesediscrepancies accounted fora

majorportion ofdisagreements between thetwo

instruments, since success of the instrumental results was predicated on the manual colony

counts. When the latter were excluded from consideration and results of thetwoinstruments

werecompared

directly,

disagreements between

the twoinstrumentswereless than 4%.

Insummary,thisstudy has shown that detec-tion,

enumeration,

and identification of eight bacterial andtwoyeast

species/groups

inurines by the AMS system is a

promising

approach,

anoverallagreementbetween the preprototype AMS instrument and conventionalmanual

pro-cedures of

approximately

90% having been

ob-tained. Further work is required to raise the system's sensitivity for P. aeruginosa and S.

aureus, toadapt the systemforuse with

speci-mensotherthan urine andtoimprove standard-ization of manual methods

(counts)

used for comparisonand validation of instrument results. Thesystemalso needstobe

challenged

with(i) organismsfor whichnoselective brothsare

pres-ently availabletofurtherdelineate the

specific-ity and limits ofsensitivity of the

existing

for-mulation and (ii) a larger number of

highly

polymicrobicspecimens (containing fourtofive differentspecies; theseare

relatively

rareamong

clinical specmens,exceptin those frompatients

with long-teroiindwelling catheters, e.g., quad-ruplegics andparaplegics, etc.).

A large-scale

multilaboratory

collaborative project for the assessmentof the AMS system

in urine

bacteriology

(1) should result in eluci-dating some of the

problems

delineated and fumishing broad-based data on the inter- and intralaboratoryreproducibility of thesystem.

ACKNOWLEDGMENT

Thisstudy wassupported by a grant from theMcDonnell DouglasCorp. to The Jewish Hospital of St. Louis.

LITERATURE CITED

1. Aldridge, C., S. Gibson, J. Lanham, M. Meyer, R. Vannest,P.Jones, and R.Charles. 1977. Automated microbiological detection/identification system. J. Clin. Microbiol. 6:406-413.

2.Isenberg,H. D., and J. D.MacLowry. 1976. Automated methodsand datahandling inbacteriology. Annu. Rev. Microbiol. 30:483-505.

3. Lennette,E. H., E. H. Spaulding, and J. P. Truant (ed.). 1974. Manual ofclinical microbiology, 2nd ed. AmericanSociety for Microbiology, Washington, D. C. VOL.