Klinisk Biokemi i Norden Nr 4, vol. 18, 2006 - page 10

10
| 4 | 2006
Klinisk Biokemi i Norden
(Fortsat fra side 8)
the numbers of abnormal differential counts at the
referring hospitals. Subsequently it has become dif-
ficult or impossible to maintain the skills needed for
high quality microscopic differential counts. Thus
evaluation of highly abnormal differential counts
often requires external help, either from the depart-
ments of clinical biochemistry or pathology. There
is little uniformity among different laboratories on
criteria for manual action following automated cell
counters. An expert group has newly published 41
rules for action, most commonly for morphological
review of blood smears (1), which is still a neces-
sity.
Automated image processing
The informatics of morphological differential leuko-
cyte counting is complicated and specific compared
to many clinical chemistry data sets currently han-
dled. The differential count represents a great deal
of biological complexity. This can be characterised
by pattern recognition systems and the increase in
IT knowledge and the enhanced IT power of instru-
ments have facilitated this approach. The first gen-
eration of “computerised microscopes” arrived dur-
ing the 1970´s like Larc (Corning Glass) or Hematrak
(Geometric data Corporation) but these instruments
were useless for clinical purposes as they could only
classify normal blood cels (2-5). Neither the imag-
ing technology to sample cells nor the neuronal
networks used for classification were sophisticated
enough. During the 1990´s, the technologies of arti-
ficial neural networks and speed of computer proc-
essors developed and a second generation is now
arriving with much better capacity, both to localize
cells, collect images to present them to the operator
and even to a part classify them (6). Applications
both within haematology (diff count) and pathology
(“PAP smears”) are FDA-approved.
In collaboration with CellaVision AB (Lund,
Sweden) we (Departments of Clinical Chemistry in
Göteborg and Malmö, Sweden) have participated
during the development and evaluated DiffMaster™
Octavia (7). This instrument is basically a software
product on a hardware platform consisting of an
automated microscope, a high quality CCD cam-
era and a computer (figure 1). The slide holder
holds eight slides at a time, which are processed
Figure 1: The DiffMaster
at work in the Department
of Clinical Chemistry,
Malmö University
Hospital, Sweden.
1,2,3,4,5,6,7,8,9 11,12,13,14,15,16,17,18,19,20,...44
Powered by FlippingBook