DIAGNOSIS MACHINE (grech, diagnostikos capable to distinguish) — use of computing methods and computer facilities for diagnosis. Of m is applied when use of computing methods and the equipment allows to make the diagnosis more reliably or more quickly in comparison with traditional methods diagnoses (see) on the basis of the same (or smaller number) data on the patient. Value D. of m consists also that by drawing up diagnostic programs ideas of the mechanism of development and disease are accurately and obviously formalized. In 60 — the 70th of 20 EL of m the wedge gradually is implemented into practice in research establishments, a profile and large to lay down. - professional, institutions.
In modern D. the m can allocate two main directions. The first is connected with attempts to put the formalized ideas of rules of diagnosis in the diagnostic program. Such programs are formed with participation of leading experts of the respective areas of medicine and assume usually implementation on the COMPUTER of algorithms of logical processing of a set of data on the patient for the purpose of establishment of the diagnosis (see. Algorithm diagnostic ). However the impossibility of full formalization of process of establishment of the diagnosis by the person, and also inability of programs to change the principle of the work on a measure of accumulation of experience led to the fact that works of this direction found only limited application.
The second direction D. of m is connected with use of so-called students of programs. The essence of «training» in this case consists in processing and the analysis of an array of case histories with the specified diagnosis and development on this basis of the algorithm allowing to make the diagnosis in each case (the so-called decisive rule), a cut further is used for establishment of the diagnosis and in the cases which were not used during the training. In process of accumulation of new data the decisive rule can change. The studying programs in the course of development of the decisive rule often allow to estimate the diagnostic importance of various symptoms and syndromes and, in particular, it is essential to reduce number of required data on the patient. It allows to make during work the diagnosis more quickly or with smaller expenses. Depending on complexity of the decisive rule it is possible to allocate two groups of students of diagnostic programs. In programs of the first group use of the COMPUTER is necessary only at a stage of development of the decisive rule. The decisive rule can be expressed as the diagnostic table or a formula and can be used further without use of the COMPUTER, manually or with use of rather simple analog and electromechanical devices. Such way considerably expands D.'s possibilities of m as allows to use it in those to lay down. institutions where there are no computer centers or their power is insufficient. In more difficult cases the decisive rule itself represents the diagnostic program and can be used only using the COMPUTER. As the studying programs in works of this direction universal programs of training are widely used to recognition of images (see). The simplest, but the m of the program which found broad application in D. are based on Bayes's principle (see Bayes the rule) which can be effectively used only at the assumption of independence of separate signs (symptoms) of a disease or in case of normally distributed parameters. The second group of the studying programs is based on direct sampling of the decisive rule on material of the training sequence. Such programs choose from more or less wide class of decisive rules the best by a certain criterion, the criterion of a minimum of mistakes on material of training is most often used.
In the Soviet and foreign works linear and kusochnolineyny decisive rules are widely used. The linear decisive rule is set by set of weight coefficients of each symptom for each diagnosis. If to designate through λij coefficient of a symptom of i for the diagnosis j, and value of a symptom i-Xi (Xi = 1 if the symptom is, and Xi = 0 if it is absent), then the linear decisive rule chooses that diagnosis, for to-rogo the sum (Xi x λij) reaches the greatest value. The studying program building the linear decisive rule shall pick up in the best way coefficients λij. Set of such coefficients forms the elementary diagnostic table. Geometrically linear decisive rule represents breakdown of space of signs on the parts corresponding to various diagnoses by means of the multidimensional planes. Piecewise linear decisive rules break space by means of the surface made of several such planes. Among the Soviet studying programs of these type in D. the m widely use system of programs of a method of the generalized portrait. Other types of decisive rules are used in the programs based on the principle of the moving apart Yu. I. Neymark (1967) metrics, «Bark» programs of M. M. Bongard and M. N. Vayntsvayga (1962), programs of finding of informative subspaces, etc. All these programs found application in D. for m.
It should be noted that by no means it is not enough to apply the appropriate program to creation of decisive rules for D. of m. The solution of each task demands huge pretreatment of archives of case histories. Only the specialist can perform this work. Besides, successful use of D. of m requires development of the formalized description of a condition of the patient, and the wide spread occurance of the received decisive rules assumes standardization of such description for various to lay down. institutions.
In a row to lay down. institutions methods D. of m are realized in the form of the diagnostic systems constructed on the basis of modern computer facilities. The most perspective are the systems working in the mode of dialogue between car and the doctor. In such systems the car is often used not only for diagnosis, but also for the choice of the optimum sequence of the researches of the patient necessary for diagnosis. Perhaps various division of tasks between car and the person; e.g., the car carries out the analysis of the curve ECG, and the final diagnosis is made by the doctor. Systems where, on the contrary, the person analyzes x-ray films are known, and the car according to the received description makes the diagnosis. Methods D. of m are used also in systems of continuous overseeing by the patient. At the same time data come to the car directly from the sensors installed on the patient.
In diagnostic systems it is desirable to use as it is possible big data arrays for training. Therefore in to lay down. institutions machine archives of case histories are created. In process of standardization of data and improvement of communication between computer centers an opportunity to use all material which is saved up in various during the training to lay down opens. institutions.
Methods D. of m in the long term, probably, will develop in the direction of creation of the adjusted models of a disease considering as these medical sciences and the accumulated experience, and specific features of the patient. The orientation of these methods at choice of optimum treatment will amplify. Systems D. of m will join in the general management information systems to lay down. institutions.
Of m gains ground in a number of the fields of medicine. In the field of surgical diagnosis in Ying-those surgeries to them. And. M apply V. Vishnevsky D. to the solution of a number of problems of differential diagnosis of the inborn and acquired heart diseases, thromboses of the left auricle, acute diseases of a stomach, diseases of a liver and stomach, etc. In the field of diagnosis of cancer diseases methods D. of m are actively developed for complex diagnosis of cancer diseases of a digestive tract, a lung and some other bodies. Such works are conducted, e.g., in Oncological scientific center of the USSR Academy of Medical Sciences. In the field of diagnosis of cardiovascular diseases D. by m of atherosclerotic stenocardia and a preinfarction angina on a wedge, to data is widely applied. Works are carried out by scientific research institute of applied mathematics and cybernetics Gorky un-that together with a number of medical institutions of Gorky. Broad work in this area is conducted also in Kaunas medical in-those and in other institutions. In the field of diagnosis of nervous diseases works in Minsk state medical in-those, etc. are conducted. Abroad methods D. of m are successfully applied in a row to lay down. institutions of the USA, France and Japan.
See also Diagnosis, cybernetic methods of a research .
Bibliography Algorithms of training in recognition of images, under the editorship of V. N. Vapnik, M., 1973; Vapnik V. N. and Chervoneyakis A.N. Theory of recognition of images, M., 1974, bibliogr.; Computer facilities in physiology and medicine, under the editorship of E. B. Babsky and V. V. Larin, M., 1968; Danilenko S. I. idr. Possibilities of use of the electronic computer for clinicoradiological differential diagnosis of cancer and high-quality damage of a gullet, Vestn, rentgenol, and radio-gramophones., No. 2, page 36, 1971, bibliogr.; Machine diagnosis and information search in medicine, under the editorship of A. A. Vishnevsky, M., 1969; H e y-mark Yu. I. and d river. Recognition of images and medical diagnosis, M., 1972, bibliogr.
A. Ya. Chervonenkis.