![]()
14
| 4 | 2006
Klinisk Biokemi i Norden
Conclusions
In conclusion, artificial neural networks can pro-
vide a decision support system that can help the
morphologist to generate haematological reports
of high quality in routine laboratory medicine.
Automated cell location and pre-classification
improves efficiency but the safeguard of manual
validation remains. A common standard trace-
able to an expert cell atlas can be created in the
future. Image storage and retrieval is simple and
networking possibilities are considerable. In the
future the laboratory will depend more and more
on such systems.
It is therefore appropriate to raise the ques-
tion if the traditional microscope will still have
a place in the future. Will not immunophenotyp-
ing, molecular biology, etc replace microscopy?
However, there are certain strengths in morphol-
ogy that are very hard to bypass. The pattern
recognition capacity of the human expert is still
necessary for high quality haematology and it
is indeed a challenge to develop new technical
systems that are capable of automated pattern
recognition similar to that of the human micro-
scopist. On the other hand a flow cytometer will
recognize aspects of cells that are not discern-
able to the human eye. Also, a flow cytometer
performs quantitative measurements, and will
quantitate at least seven parameters of each cell
on 10 000 cells. Microscopic morphology and
immunophenotyping are both methods that can
be used to assess the complexity of a sample.
Molecular biology methods, on the other hand,
have gained use in detecting presence or absence
of a certain aspect in the total sample but are
useless for assessing the complexity of a sam-
ple. PCR methods are now being used to detect
minimal residual disease (MRD) where very few
malignant cells with a specific genetic aberration
may be present.
Still, when the patient with acute leukaemia
arrives to the hospital late Friday afternoon,
a glance in the microscope by an experienced
microscopist will establish the diagnosis with-
out the need for expert laboratory methods or
advanced technology. Therefore the microscope
and the skilled human will survive for the time
being.
References
1. Barnes PW, Mc Fadden SL, Machin SJ, Simson
E. The international consensus group for hema-
tology review: Suggested criteria for action
following automated CBC and WBC differential
analysis. Lab Hematol 2005;11:83-90.
2. Rosvoll RV, Mengason AP, Smith L, Patel HJ,
Maynard J, Connor F. Visual and automated
differential leukocyte counts. A comparison
study of three instruments. Am J Clin Pathol.
1979; 71:695-703.
3. Pierre RV, Payne BA, Lee WK, Hyma BA,
Melchert LM, Scheidt RM.. Comparison of
four leukocyte differential methods with the
National Committee for Clinical Laboratory
Standards (NCCLS) reference method. Am J Clin
Pathol. 1987; 87:201-9.
4. Perel ID, Herrmann NR, Watson LJ. Automated
differential leucocyte counting by the Geometric
Data Hematrak system: eighteen months
experience in a private pathology laboratory.
Pathology 1980;12:449-60.
5. Penttilä IM, Mahlamäki E, Mononen I,
Kärkkäinen P Adaptation of the May-Grunwald-
Giemsa staining method for automated dif-
ferential counting of blood leukocytes by a
Hematrak analyzer. Scand J Haematology 1985;
34: 274-280
6. Tatsumi N, Pierre R V. Automated image
processing. Past, present, and future of blood
cell morphology identification. Clinics Lab Med
2002, 22, 299-315.
7. Swolin B, Simonsson P, Backman S, Löfquist I,
Bredin I, Johnsson M: Differential counting of
blood leukocytes using automated microscopy
and a decision support system based on artifi-
cial neural networks- evaluation of DiffMaster
TM Octavia. Clin Lab Haem 2003;25:139-147.
8. Koepke JA, Dotson MA, Shifman MA, . A
critical evaluation of the manual/visual differ-
ential leukocyte counting method. Blood Cells.
1985;11:173-86.
9. Rumke CL.. Imprecision of ratio-derived dif-
ferential leukocyte counts. Blood Cells.
1985;11:311-4, 315.
10. Berend Houwen The Differential Cell Count
Laboratory Hematology 2001; 7: 89-100
(Fortsat fra side 13)
