TIME-ORIENTED MULTI-IMAGE CASE HISTORY – WAY TO
THE “DISEASE IMAGE” ANALYSIS
Saveli Goldberg
1
, Nikita Shklovskiy-Kordi
2
and Boris Zingerman
2
1
MGH, Boston, MA, USA
2
National Center for Hematology, Moscow, RF
Keywords: Integrate Medical Image Construction, Image Knowledge Base.
Abstract: Electron patient records clinical database allows creating integrative presentation links different types of
medical information for individual patient on real-time scale. Resulting image, can be confront with
"disease image" – non-formalised physician’s perceptions of nosology expression in individual patient’s
disease. Normalization of parameters permits data processing based on case-to-case and case-to-cluster
comparative and multivariate statistical analysis of the patient’s data.
1 INTRODUCTION
Recognition of diagnostic pictures, uniting different
data and parameters in the process of deagnostic of
individual patient’s disease can be naimed the
creation of "disease image" (Vorobiev et al., 1999).
Connecting individual features of patient suffering
with general medical knowledge remains the main
problem of medicine. This mission will never be
solved completely at a tool level because Human
Being is the extreme quantum device for making
decisions of this kind of problems (Liberman et al.,
1998). However organazing data and images from
databases and collecting knowledge base about their
possible interaction is a real problem for experts and
should be studied by mathematic theory.
The verbal characteristic of creative thinking of a
doctor hematologist-morphologist experienced in
both macro and microscopic investigation of a
patient can be put into words: "You are telling me
nonsense about this patient! I do see - (hand wave in
the air) - the picture of his bone marrow!».
Analysis of disease progression and treatment
outcome requires great amount of data with
reference to their occurrence on the time scale and
causal connections. Most part of this data is
avaliable via computer, but there is no instrument to
extract them from different uncoordinated databases,
to organize them, or at least to collect them on single
screen. In this paper we propose a real-time
computerized medical system for collecting and
storage of clinical information on an individual
patient with abilities of immediate verification,
analysis and creation of integrated presentation
which can be analyzed as the image of the individual
case and “disease image”. The Sysem presents way
to reduse number of analised variables, selecting the
tamplete. The redused set of variables should retain
as much of important for treatment information
existed in the original variables as possible. Our
previose experience was to solve similar problem
with Dr. Watson Type System (Goldberg et al.,
1995).
2 MATERIALS AND METHODS
We have taken advantage of a traditional system
used for logical structuring of data known as
“temperature sheets”. This system leading
parameters and therapeutic assignments on one sheet
of observation have a common time axis. This
approach is common in Hematology/Oncology units
and it has helped a lot to develop clinical parameters
for treatment of acute radiation disease (biological
dosimetry, in particular) (Vorobiev, 1970), as well
as to improve protocols for treatment of
hematological malignancies (Vorobiev, 1977). We
used the following steps to generate this instrument:
1. Integration of data stored in different formats
(text, tables, roentgenograms, microphotographs,
videos etc.);
200
Goldberg S., Shklovskiy-Kordi N. and Zingerman B. (2007).
TIME-ORIENTED MULTI-IMAGE CASE HISTORY WAY TO THE “DISEASE IMAGE” ANALYSIS.
In Proceedings of the Second International Conference on Computer Vision Theory and Applications, pages 200-203
DOI: 10.5220/0002070902000203
Copyright
c
SciTePress
Figure 1: Example of the Page of Dynamic Observation of a patient underwent chemotherapy treatment. Some information
presented as marks only, some as small windows. The dynamics of chosen parameters (Temperature, WBC and Platelet
counts) are normalized and colour-coded.
2. Compression of clinical data by highlighting
important and urgent information;
3. Display of information in an integrated fashion
on the same screen;
Automatic matching of entered data with stored
timetables derived from established protocols for
diagnostic procedures and treatment.
Exit from the assigned limits of selected
parameters is similarly controlled.
The data can be plotted manually on the basis of
standard case history (wherein the physician
acquires medical or diagnostic information and
enters it manually on a Page of Dynamic
Observation (PDO)), or based on a Time-Oriented
Multi-Image Case History (TOMICH) (Shklovskiy-
Kordi et al., 1998).
Manual method of entering data on PDO does
not require additional devices or software besides a
regular PC computer with Microsoft Office and a
PDO template in an Excel format. PDO data entry
form is an Excel spreadsheet, in which certain rows
are reserved for specific type of information (i.e.
clinical data, laboratory results, medications, etc.).
The software contains a set of templates for widely
used clinical protocols and a system for automatic
detection of protocol violations.
When a standard protocol is used for
management of a patient with known diagnosis, a
template is provided with required laboratory data
and medications to be used with a timetable for their
administration.
The field of the graphical presentation of the
“dynamic» indicators - any numerical indicators
which change in the time (as temperature, blood
pressure, blood counts) essential for the clinical
case. The field of flags represent:
- icons which open physician notes,
-icons from which specific additional
information can be open . The color of these icons
corresponds to the expert evaluation of the
importance: “red” - changes the opinion about the
illness and lead to a change of the cure tactic;
“green” - confirms the diagnosis and the cure tactic;
“gray” - does not give a definite answer.
- icons of the planed research, which color
depends on the fulfilling.
After filling the “event form”, a cursor tags the
event to a “window” for a brief description (i.e. CT
TIME-ORIENTED MULTI-IMAGE CASE HISTORY – WAY TO THE “DISEASE IMAGE” ANALYSIS
201
scan description on Fig.1). A double click opens a
map of all pictures and text files linked to the event.
Normalization. All numeric clinical data is
broken into normal, sub-normal and pathological
range. This provides normalization of all parameters
and presentation on common axes. To define the
range of sub-normal values, a team of experts
empirically established the scope of “acceptable”
(for the given diagnosis and, in some cases, for an
individual patient) parameters. If a parameter stays
within the defined sub-normal or normal range, no
special action is required. If the specific value is out
of the acceptance limits, the program generates an
automatic alarm signal.
Complications: Complications are recorded on a
specific line in PDO and serves to visualize the
dynamics of patient’s symptoms. After a
symptom/syndrome is selected from a pre-loaded
list, a window appears on the screen with a
definition and criteria to assist in the diagnosis and
management. 4-positions of scale, “x” - lack of the
syndrome, “+” - the growth of the syndrome, “-“ -
the decreasing of the syndrome, “0” - the state
without any changes (stabilization).
3 DISCUSSION
TOMICH has a standard format for presenting key
components of patient’s medical record (the constant
form of a positional relationship of the basic
semantic units of a case history), but also has the
flexibility for adding new templates, as necessary for
a specific diagnosis. These templates accumulate
pre-defined lists of medications, required lab tests
and syndromes, and define sub-normal and
pathological range of values, as well as color palette
for drugs and graphs. Also, the template may refer to
the standard protocols for specific diseases or
clinical trials stored in the database. Normalization
of parameters makes future perspective of the data
processing based on case-to-case and case-to-cluster
comparative and multivariate statistical analysis of
the patient’s data
The beforehand constructed template permits
standard recognized images for diagnosis and helps
to discriminate general characteristics and specific
features for an individual patient. For example, there
are accepted criteria for decrease in platelets,
leukocyte and hemoglobin in response to
chemotherapeutic treatment. We found that
comparison of shapes of drug-dependent changes in
blood counts is a valuable estimation of outcome
(Shklovskiy-Kordi et al., 2004).
In a real-time mode, TOMICH automatically
performed data validation and notified a user when
selected parameters were beyond acceptable ranges
or when the timetable set by the protocol was not
followed. These software features permit health care
personnel to monitor and correct, when needed,
individual actions taken by medical personnel.
TOMICH links the actions of medical staff with
requirements set by the protocols. Attention of
physicians and staff is prompted by a color indicator
(Shklovskiy-Kordi et al., 2003).
4 CONCLUSIONS
TOMICH is a convenient and easily automated
method for entering all available information about a
patient. It may be classified as a decision-support
and expert–oriented system, which allows a
physician to select a pre-entered template and to
modify it for creating the most appropriate template
for a particular patient. It provides easy access to
primary data and allows generation of a common
time-line axis format for multimedia presentation of
a patient’s record. The system links different
medical information and forms a cognate image of
diseases. This presentation allows real-time
evaluation of disease and of the response to
treatment. Use of TOMICH facilitates the analysis
of clinical course and compliance and reduces the
probability of medical errors.
Normalization of parameters makes future
perspective of the data processing based on case-to-
case and case-to-cluster comparative and
multivariate statistical analysis of the patient’s data
ACKNOWLEDGEMENTS
This work was supported, in part, by a Grants 05-07-
90231-в and 06-08-00610-а of RFBR, Russia.
REFERENCES
Vorobiev A., Vorobiev I., Kremeneckaia A., Harazichvily
D., Shklovskiy-Kordi N.: The Sintesis of Traditions
and Novations in contemorary Diagnstic, Clinic
laboratory practis (in Russian), Vol.9, Medicina,
Moscow, 1999, pp.6-7
Liberman E.A., Minina S.V., Shklovsky-Kordi N.E.:
Biological information and laws of nature Biosistem,
46, 1998, р. 103-106.
VISAPP 2007 - International Conference on Computer Vision Theory and Applications
202
Goldberg S., Saveliev D., Tsibulkin E.: The desision
support system for therapeutic departmemt physican as
an exemple of Dr. Watson Typr System,MEDINFO95
Vancuver, Canada, 1995, p. 1031
Vorobiev AI, Brilliant MD. WBC and Platelets dynamic
during acute irradiation syndrome as biological
dosimeter. (Russ.) Proc Conf. Institute of Biophysics,
Moscow, 1970,.p.12
Vorobiev AI, Brilliant MD. Experience of ambulance
treatment of hemablastose patients. (Russ.)
Arh.Sov.Med., Moscow,1977, 49(8), p.3-9.
Shklovskiy-Kordi N., Goldberg S., Zingerman B.: Time-
oriented multi-image presentation of dynamic case
history. Proc AMIA Symp. 1998, p.107
N.E.Shklovskiy-Kordi, N.G.Chernova, L.I.Erlih and
B.V.Zingerman: Base of knowledge for individual
case records, Abstr. for the 6th International
Symposium on Hodgkin’s Lymphoma, Cologne,
Germany, 2004, p.33
Shklovsky-Kordi N., Zingerman B., Rivkind N., Goldberg
S., Davis S., Vartikivski L., Krol M.,: Multimedia
Case History As A Tool For Management Of Patients
And Clinical Trials. Proceedings, Sixteenth IEEE
Symposium on Computer-Based Medical Systems,
2003, p.269-272
TIME-ORIENTED MULTI-IMAGE CASE HISTORY – WAY TO THE “DISEASE IMAGE” ANALYSIS
203