COGNITIVE REASONING IN INTELLIGENT MEDICAL

INFORMATION SYSTEMS

Visual Data Perception Algorithms in Medical Decision Support Systems

Marek R. Ogiela

, Ryszard Tadeusiewicz

, Lidia Ogiela

AGH University of Science and Technology

Institute of Automatics,

Department of Company Management

Al. Mickiewicza 30, PL-30-059 Kraków, Poland

Keywords: Intelligent systems, image understanding, pattern classification, medical imaging, artificial intelligence

Abstract: This paper presents new approach in application of picture languages for cognitive analysis and reasoning of

selected medical visualization. It will be shown new opportunities for applying these methods to undertake

tasks of the automatic understanding of image semantics in intelligent medical information or computer-

aided diagnosis systems. These systems are applied in various tasks supporting decisions taken in the wide

area of health care and medical imaging. The possibility of obtaining the information about semantic

content of the medical images may contribute considerably to the creation of new intelligent cognitive

medical systems. This article shows that structural techniques of artificial intelligence may be applied in the

case of tasks related to automatic classification and machine perception of semantic pattern content in order

to determine the medical meaning of the images. In the paper, we describe some examples presenting ways

of applying such techniques in the creation of cognitive vision systems for selected classes of medical

images.

1 INTRODUCTION

In the foundations of image understanding there are

many algorithms and AI approaches to the task of

intelligent visual data perception and analysis.

Among them one is the most important enabling to

make a deeper semantic and cognitive analysis.

These are picture languages consisting formal

grammars for pictorial pattern analysis as well as

languages of shape features description allowing

multidimensional pattern classification. In this paper

will be presented the way of application of such

formalisms to the task of understanding of medical

visual data, especially in the intelligent medical

information systems. We try to show how the tasks

of automatic understanding of medical data may be

done using cognitive analysis approach, which allow

to make a semantic perception of analyzed

visualization. Generally, the perception of an image

requires a deeper analysis aimed at the determination

of significant semantic features (Albus, 2001). Such

features

enable a further semantic image

interpretation or a semantically oriented indexation

in databases (especially when objects are retrieved

from various diagnostic examinations or determine

different disease entities). A proper semantic

interpretation of the data being analyzed is very

important because in the case of medical images

often happens that the same illnesses are visualized

in various forms of images that are registered and

processed (Fig.1). This is the main reason why

attempts have been made to create a system that

would automatically find the message (i.e. the

content) carried by analyzed medical images.

Due to the fact that the number of combinations

of features that characterize images is not limited, it

can be assumed that perception may refer

the image

to potentially unlimited number of classes. This may

be achieved by cognitive analysis, in which

specified languages of image description must be

used.

The general approach in the cognitive analysis is

the initial interpretation of images and specification

of important features. The proper selection of such

features is conducted by means of image pre-

processing. Next features are subsequently described

with the use of a picture language generated by an

appropriately defined attributed grammar. Properties

described in this way can be later reproduced in the

222

R. Ogiela M., Tadeusiewicz R. and Ogiela L. (2004).

COGNITIVE REASONING IN INTELLIGENT MEDICAL INFORMATION SYSTEMS - Visual Data Perception Algorithms in Medical Decision Support

Systems.

In Proceedings of the Sixth International Conference on Enterprise Information Systems, pages 222-225

DOI: 10.5220/0002603402220225

 SciTePress

course of structural reasoning conducted by the

prepared information system.

The main advantage of this approach is its

possibility to interpret the meaning of a much bigger

class of images than the ones, which were used for

the writing of the formal language. This results from

the fact that the used grammar rules generalise the

descriptions introduced and allow one to interpret

new cases, previously not defined.

For such interpretation of the mentioned

structures and for a verification of lesion

advancement level, a graph grammar (Ogiela, 2003),

and an attributed grammar have been proposed.

Before coming to the cognitive interpretation of

the changes, it is necessary to preserve the sequence

of preliminary operations, which are included in the

image pre-processing. The goal of this analysis is to

obtain new representation in the form of width

graphs, which show the pathological changes

occurring in these structures.

Figure 1: Two images carrying the same message. In both cases, the symptoms of pancreas cancer are visible

During the initial analysis of visualisations, the

following operations are executed: segmentation,

skeletonization, and the application of a

straightening transformation to transform the

contour of the analysed structure into two-

dimensional graph, which shows a profile of a

straightened organ. The graphs obtained in such way

are the starting point in the classification of

morphological features by using context-free

grammars. In order to define primary components on

the obtained width graphs as terminal symbols

describing these components, an algorithm of linear

approximation was used. As a result of

approximation the sequences of terminal symbols

for every graph was received, which constitute an

input to syntax analysers and semantic classifiers.

2 PICTURE GRAMMAR FOR

COGNITIVE ANALYSIS

A possibility to conduct cognitive analysis will be

presented on the examples of patterns received

during the diagnostic examinations of renal pelvis,

and coronary arteries.

2.1 Coronary Image Interpretation

Analysis of coronary arteries is extremely important

from the point of view of correct diagnosis of

myocardial ischaemia states caused by coronary

atheromatosis sclerosis lesions resulting in stenoses

of artery lumen, which in consequence lead to

myocardial ischaemia disease. This disease can take

the form of either stable or unstable angina pectoris

or myocardial infraction (Khan, 1996).

The following attributed grammar has been

proposed to diagnose various types of stenosis

shapes:

= {SYMPTOM, U, H, D}

= {h, u, d} for h∈[-10°, 10°], u∈(10°, 90°),

d∈(-10°, -90°)

STS = STENOSIS

SP:

STENOSIS → D H U

STENOSIS → D U | D H

Lesion = Stenosis

H → H h | h

D → D d | d

U → U u | u

sym

= w

sym

+ w

sym

= h

sym

+ h

...

COGNITIVE REASONING IN INTELLIGENT MEDICAL INFORMATION SYSTEMS - Visual Data Perception

Algorithms in Medical Decision Support Systems

223

This grammar allows to detect different forms of

coronary artery stenosis, which may characterize the

different disease units (angina pectoris or infarct).

Using attributes permits to calculate the numerical

parameters and semantic information of detected

lesions, which allows to characterize the degree of

lesion development.

The simplicity of this grammar results mainly

from the big generation capacity of context-free

grammars, understood mainly as possibilities to

describe complex shapes by means of a small

number of introductory rules, that is grammar

productions.

2.2 Renal Pelvis Cognitive Analysis

In the case of analysis of renal radiograms, the main

task is to recognise local stenoses or dilations of

upper segments of urinary tracts and attempt to

define the correct morphology of renal pelvis and

renal calyxes. Lesions in those structures can

suggest the occurrence of renal calculi or deposits,

which causing ureter artresia can lead to diseases

such as acute extrarenal uraemia or hydronephrosis.

An analysis of the correct morphology of ureter

lumen will be conducted with the use of context-free

attributed grammar.

Diagnosing morphological lesions in the form of

ureter stenosis or dilations has been conducted with

the use of the following attributed grammar:

= {LESION, STENOSIS, DILATATION, HOR,

SLOPE_UP, SLOPE_DOWN}

= {h, v, nv} for h∈[-8°, 8°], su∈(8°, 180°),

sd∈(-8°, -180°)

STS = LESION

SP:

LESION → STENOSIS

STENOSIS → SLOPE_DOWN HOR

SLOPE_UP

STENOSIS → SLOPE_DOWN

SLOPE_UP

STENOSIS → SLOPE_DOWN HOR

Lesion = Stenosis

LESION → DILATATION

DILATATION → SLOPE_UP HOR

SLOPE_DOWN

DILATATION → SLOPE_UP

SLOPE_DOWN

DILATATION → SLOPE_UP HOR

Lesion =

Dilatation

HOR → HOR h | h

SLOPE_DOWN → SLOPE_DOWN

sd | sd

SLOPE_UP → SLOPE_UP su | su

sym

= w

sym

+ w

;

sym

= h

sym

+ h

...

3 SELECTED RESULTS

As a result of cognitive analysis using linguistic

approach it is possible to understand pathogenesis of

the deformations viewed on x-ray images of the

organs under consideration, what means the

possibility of recognize some kind of diseases even

on images absolutely not similar one to other.

Presented approach is applicable even if no

templates of healthy and pathological organs at all or

if number of recognized classes goes to infinity. In

particularly applications of the presented grammars

deliver almost complete information concerning the

visual morphological irregularities of investigated

organs. An analysis of the morphological changes

was carried out based on a set containing few dozens

of images. The efficiency of gaining recognition of

information with semantic character, in all cases

exceeded the threshold of 93%. In Fig. 2 are

presented examples, which show the description of

the changes in ureter ducts, and coronary arteries.

The results obtained owing to the application of

the characterized methods, confirm the immense

opportunities offered by syntactic methods in the

cognitive analysis of medical visualizations showing

dangerous pathological lesions.

4 CONCLUSION

Development of the intelligent information systems

and techniques of visual data semantics analysis

made possible to understand the medical meaning of

any images coming from diagnostic research.

However, the full automatic analysis and

interpretation of such data is still a real problem,

advanced techniques of artificial intelligence must

be applied to enable the creation of systems that can

both recognize and understand visual data (Ogiela

2003).

Thus the aim of the presented techniques was to

show an innovative concept of the application of

structural pattern analysis in the creation of

cognitive information systems.

ICEIS 2004 - ARTIFICIAL INTELLIGENCE AND DECISION SUPPORT SYSTEMS

224

Figure 2: Results of disease symptom interpretation in intelligent diagnosis support system

Such systems are able to understanding and

determining the semantic meaning of medical

images of certain classes. It is worth mentioning that

machine perception using such methods may lead to

an automatic interpretation of medical images in the

way it is done by a specialist. It may enable the

determination of not only crucial changes but also

the consequences of existing irregularities and

finally the optimal directions and methods of

conducting a suitable therapy. Automatic

understanding of the image content can have

numerous further applications for example such

information can be used to monitor therapeutic

processes or to forecast disease development as well

as the patient’s future state of health.

ACKNOWLEDGEMENT

This work was supported by the AGH University of

Science and Technology under Grant No.

10.10.120.39.

REFERENCES

Albus, J.S., Meystel, A.M., 2001. Engineering of Mind:

An Introduction to the Science of Intelligent Systems.

John Wiley & Sons.

Khan, M.G., 1996. Heart Disease Diagnosis and Therapy.

Williams & Wilkins, Baltimore.

Ogiela, M.R., Tadeusiewicz, R., New approach for

cognitive analysis and understanding of medical

patterns and visualizations. Proceedings of SPIE, vol.

5203 Andrew G. Tescher (eds.), SPIE, Bellingham

WA, 2003, pp.615-622.

Ogiela, M.R., Tadeusiewicz, R., Cognitive Vision Systems

in Medical Applications. Lecture Notes in Artificial

Intelligence, 2871 (2003) 116-123.

Ogiela, M.R., Tadeusiewicz, R., Artificial Intelligence

structural imaging techniques in visual pattern analysis

and medical data understanding. Pattern Recognition,

36 (2003) 2441-2452.

Sonka, M., Fitzpatrick, J.M. (eds.), 2000. Handbook of

Medical Imaging: Vol. 2- Medical image processing

and analysis. SPIE PRESS, Bellingham WA.

COGNITIVE REASONING IN INTELLIGENT MEDICAL INFORMATION SYSTEMS - Visual Data Perception

Algorithms in Medical Decision Support Systems

225