Possibilities of brain SPECT with perfusion radiopharmaceuticals for the quantitative assessment of cognitive changes in patients with hypertensive encephalopathy

Background. Vascular diseases of the brain, which lead to encephalopathy, are a significant medical and social problem. The main clinical tool for diagnosing cognitive impairments is a neuropsychological testing. Its disadvantages are a big number of different tests, which are used in clinical institutions, and thus, make it extremely complicated to compare the data; in monitoring studies, patients can learn answers, which somewhat distorts the results; there is also a possibility of a non-objective doctor’s impact on the results of the conducted test. Therefore, the development of methods for assessing the neuropsychological and cognitive state of patients based on objective data is an urgent task. Besides, to this date, it is not completely known which segments of the brain directly or indirectly affect this or that cognitive function. Purpose – to develop a methodology for assessing the scores of neuropsychological testing (NPT) in patients with atherosclerotic hypertensive encephalopathy (ATHE) based on data from single-photon emission computed tomography (SPECT) with perfusion radiopharmaceuticals (RPh). Methods. atherosclerotic The study was performed within the framework of the research project at the State Institution «Romodanov Neurosurgery Institute of the National Academy of Med ical Sciences of Ukraine» «To analyze the structural and functional impairments of central nervous system in interim and long-term periods of a blast-induced mild trauma-tic brain injury». State registration number 0117U004278 using the data of the research project of Department of Neurology and Reflexology of Shupyk National Healthcare University of Ukraine of the of Health of Ukraine «Neurological, neuropsychological, neurovisual characteristics of the patients with hypertensive ence-phalopathy» (state registration number 0110V005690) which were included in the dissertation of Zalisna Yu.D. for the degree «Candidate of Medical Sciences» on the topic «Neurological, neuropsychological, neurovisual characteristics of the patients with hypertensive ence-phalopathy», 2015, Kyiv. contribution: conduction of scintigraphic examinations, participation in the development of a scintigraphic atlas of the brain, clinical data analysis. Author’s contribution: conduction of scintigraphic examinations, general management, participation in writing and editing the text of the article.

Background. Vascular diseases of the brain, which lead to encephalopathy, are a significant medical and social problem. The main clinical tool for diagnosing cognitive impairments is a neuropsychological testing. Its disadvantages are a big number of different tests, which are used in clinical institutions, and thus, make it extremely complicated to compare the data; in monitoring studies, patients can learn answers, which somewhat distorts the results; there is also a possibility of a non-objective doctor's impact on the results of the conducted test. Therefore, the development of methods for assessing the neuropsychological and cognitive state of patients based on objective data is an urgent task. Besides, to this date, it is not completely known which segments of the brain directly or indirectly affect this or that cognitive function. Purpose -to develop a methodology for assessing the scores of neuropsychological testing (NPT) in patients with atherosclerotic hypertensive encephalopathy (ATHE) based on data from single-photon emission computed tomography (SPECT) with perfusion radiopharmaceuticals (RPh). Materials and Methods. NPT and SPECT data of twenty patients with clinical diagnosis of atherosclerotic hypertensive encephalopathy were analyzed. The principal scales used during the study were the following: Montgomery -Asberg Depression Rating Scale (MADRS), Hamilton Anxiety Rating Scale (HARS), Schulte Table, O. Luria. Tomographic images of the brain were obtained on the gamma camera «E. Cam» (Siemens) using perfusion lipophilic radiopharmaceutical 99m Tc-hexamethylpropyleneamineoxime ( 99m Tc-HMPAO). Processing and analysis of the scintigraphic images were conducted in the original software «ScintiBrain», which is implemented in the Matlab-2018 environment. The quantitative characteristics of accumulation and spatial distribution of RPh in the brain were: specific accumulation of RPh (Upt) and parameter of intrahemispheric symmetry (IHS) of perfusion in the brain segments. Based on machine learning method, which can be attributed to regression methods, NPT data were compared with the ratio of Upt and IHS between different segments of the brain Results. As a result of the analysis (machine learning) of 190 Upt and IHS ratios between different segments of the brain with NPT values, the most informative ratios in terms of regression analysis were highlighted. The independence of Upt and IHS makes it possible to increase the accuracy of calculations of NPT values by algebraic averaging of calculations by Upt and IHS ratios.
The study was performed within the framework of the research project at the State Institution «Romodanov Neurosurgery Institute of the National Academy of Medical Sciences of Ukraine» «To analyze the structural and functional impairments of central nervous system in interim and long-term periods of a blast-induced mild traumatic brain injury». State registration number 0117U004278 using the data of the research project of Department of Neurology and Reflexology of Shupyk National Healthcare University of Ukraine of the Ministry of Health of Ukraine «Neurological, neuropsychological, neurovisual characteristics of the patients with hypertensive encephalopathy» (state registration number 0110V005690) which were included in the dissertation of Zalisna Yu.D. for the degree «Candidate of Medical Sciences» on the topic «Neurological, neuropsychological, neurovisual characteristics of the patients with hypertensive encephalopathy», 2015, Kyiv.
Vascular diseases of the brain, which lead to encephalopathy, are a significant medical and social problem. The most significant factors that cause neuropsychological and cognitive changes are thought to be: chronic blood circulation disorder of central nervous system, arterial hypertension, atherosclerosis, diabetes mellitus, cardiac pathology, blood pathologies, etc. [1]. At the same time, diffuse and focal changes in the brain are observed during the progression of the disease [2,3]. In such cases, single-photon emission computed tomography (SPECT) is a high-informative method for diagnosing the hemo-dynamic disorders [4 -6]. However, it should be taken into account that hypertensive encephalopathy is not only focal changes in the brain tissues. The disruption of blood microcirculation and respective cerebral functions are observed in the brain in general [7 -9].
Machine learning results by both Upt and IHS values had one order of correlation and a mean squared error in the tests. The correlation coefficient of linear approximation of values between the SPECT and NPT data was in the range of 0.75-0.93 (p < 0.01), which corresponded to the average relative error in calculations of test values from 7-22%. Conclusions. For the first time, a method of quantitative assessment of the neuropsychological and cognitive state of patients was developed on the basis of an objective research method, namely SPECT. The average relative error in calculations of equivalent values of NPT was in the range of 7 -22%. At the same time, it was shown that the neuropsychological and cognitive state of patients, in particular patients with ATHE, according to the analysis of effective brain perfusion, didn't correspond to one specific segment of the brain, but instead to a group of three interrelations of RPh accumulation in the highlighted regions of interest. The presence of a sufficiently high correlation between the NPT data and the indicators of the specific accumulation of RPh and the IHS in the brain segments showed that the neuropsychological and cognitive state of the patient depended not only on the level of accumulation of RPh in segments, but also on its spatial distribution. Церебральна ОФЕКТ рекомендована пацієнтам з енцефалопатіями для оцінки поширеності ураження ГМ. При виявленні вогнищевих змін таке дослідження стає показовим в оцінці ефективності терапії та нормалізації ефективної перфузії. Також перфузійні зображення ОФЕКТ ГМ є інформативними і при додементних розладах [10]. Так в [10] відмічається статистичний взаємозв'язок між показниками накопичення радіофармпрепарату (РФП) в структурнофункціональних блоках за О. Лурія з типом додиментних когнітивних розладів.
Cerebral SPECT is recommended for patients with encephalopathy for assessing the brain lesion spread. In cases of focal changes, this examination becomes demonstrative in assessment of the efficacy of the therapy and normalization of effective perfusion. Perfusion CT images of the brain are also informative in the cases of predementia disorders [10]. Indeed, in [10] statistical connection between the indicators of radiopharmaceutical (RPh) accumulation in structural functional blocks according to O. Luria and the type of predementia cognitive disorders are observed.
The main clinical tool for diagnosing cognitive impairments is a neuropsychological test (NPT) [11,12]. But the possibilities of electroencephalography are also worth mentioning [13], as well as rapidly developing methods of functional MRI, which make it possible to estimate the brain connectivity -compatibility of system components, their synchronization [14]. The disadvantages of NPT are a too big number of different tests, which are used in clinical institutions, and thus, make it extremely complicated to compare the data; in monitoring studies, patients can learn answers, which somewhat distorts the results. Moreover, no matter if the test is aimed at diagnostics of a certain cognitive function, all of them somehow activate the whole range of different groups of neurons which get «indirectly» involved in the processes. In other words, NPT doesn't allow physiologically caused reasons of cognitive changes to be determined. Therefore, we set a task of estimating the possibility of quantitative assessment of cognitive impairmens which are equivalent to the main NPT on the basis of the analysis of SPECT brain images with perfusion radiopharmaceuticals.
Objective -of the study was to develop methodology for assessing the scores of NPTs in patients with atherosclerotic hypertensive encephalopathy (ATHE) based on data from SPECT with perfusion radiopharmaceuticals, in particular with 99m Tc-hexamethylpropyleneamineoxime ( 99m Tc-HMPAO).
Нейропсихологічне тестування У відповідності до мети в роботі співставляються дані ОФЕКТ з такими методиками НПТ: -MADRS -шкала Монтгомері-Асберга для оцінки депресії (Montgomery-Asberg Depression Rating Scale); Patients 20 patients aged from 45 to 60 years with stage 2 ATHE were examined. The disease duration was 9 ± 4,2 years and the average age of the patients -53,1 ± 2,5 years. The diagnosis was established, taking into account the characteristics of clinical neurological syndrome according to the classification of vascular diseases of the brain, recommended by ICD-10. Every patient underwent the following examinations: a clinical neurological dynamic examination, a neuropsychological test, magnetic resonance imaging of the brain, ultrasound duplex scanning of head and neck vessels, laboratory examinations, brain SPECT imaging. Arterial pressure was assessed according to the data of daily monitoring.

Neuropsychologic testing
In accordance with the purpose of the work the SPECT data were correlated with the following NPT methods: -MADRS (Montgomery-Asberg Depression Rating Scale) -a scale for measuring the severity of depression; ISSN 2708-7166 (Print) ISSN 2708-7174 (Online)
-HARS (Hamilton Anxiety Rating Scale) -a clinical rating scale for measung the severity of anxiety symptons in patients; -Schulte Table -an attention switching test, performance capability. This method consists of five Schulte tests which allow the influence of fatigue on the attention to be observed. Therefore, along with the results of immediate separate testing (Sh 1 , Sh 2 , Sh 3 , Sh 4 , Sh 5 ), the generalized value -the average score of the tests (Sh av ) was calculated; -O. Luria's 10-word memory test. It is aimed at estimating the volume of short-term and long-term memory. This test includes the examination of fatigue influence on the result, so it was performed seven consecutive times (L 1 , L 2 , L 3 , L 4 , L 5 , L 6 , L 7 ). The total score value (a number of words, remembered by a patient) of all the tests (L ∑7 ) was used as a generalized indicator of memory assessment.
For the purpose of integral quantitative assessment of the degree of cognitive impairments, the generalized indicator of Schulte-Luria test (G 1 ) and MADRS -HARS (G 2 ) was used [15]: where Sh is average time of Schulte test completion by the patient, Sh 0 is an ideal time of Schulte test completion (30 s), L ∑7 is a total number of words, remembered by the patient in seven Luria tests, L 0 is an ideal number of words that the patient shoud remember in 7 Luria tests (70 words), MADRS and HARS -the score of respective tests.
The patients also underwent NPT according to the following methods: -MMSE (Mini-Mental State Examination); -FAB (Frontal Assessment Batter), which was developed for detecting dementia with prevalent disorders of frontal lobe or subcortex of the brain. According to the SPECT data, the adequate signs of perfusion change in these lobes weren't always observed; -Single Leg Stance Test; -HADS (Hospital Anxiety and Depression Scale), which was developed by Zigmond A.S. and Snaith R.P to detect and measure the severity of depression and anxiety; -CLOCK -the clock-drawing test in which the patient is asked to draw a clock. The test is used for differential diagnosis of dementia of frontal lobe and in the cases of disorders of subcortex structures -of Alzheimer's disease.
However, the latter tests were the criteria for the formation of the group of patients, therefore regression analysis between these tests and SPECT image is uninformative and is not presented in this work.
General characteristics of the patients according to the mentioned NPT data are given in Table 1.

SPECT images
Scintigraphic examinations of the brain were performed on the gamma camera «E. Cam» (Siemens) with LEHR collimator 15-20 minutes after 99m Tc-HMPAO administration. Tomographic examination included the collection of 128 projections for 128×128 matrix (3 patients) or 64 projections for 64×64 matrix (17 patients); the injected radiopharmaceutical activity amounted to 740 MBq. Ukrainian journal of radiology and oncology. 2022;30 (1) Processing and analysis of the images were conducted in Matlab-2018 environment in the application with the user interface, created by the «ScintiBrain» developers [16].
The main calculation parameters were intrahemispheric symmetry (IHS) of the brain (q) [17] and specific accumulation of RPh in the brain segments: Specific accumulation of RPh in a given segment (Upt) was calculated as ratio of absolute value of registered gamma-impulses in the region of interest (Q) to its volume (V) in cm 3 : where index i is an ordinal number of the brain segments (regions of interest).
Unlike the IHS value, which is mentioned and described in the publications [12], for increasing the stability (reduction in variability) of the calculation of respective parameter, normalization of the q value was used in this work: where i is an ordinal number of the brain segments (ROI); indexes «19» and «20» are the right and the left hemisphere respectively.
The use of normalization of (3) and (4) permitted variability of the output data calculations to be reduced: Upt -from 5 -10% to 3 -7%, q -from 5 -17% to 5 -8%. These variabilities were estimated according to the processing of scintigraphic images of the patients five times without automatic smoothing parameters selection and spatial orientation of the brain; an expert method for assessing the quality of prior image processing was used.
Зазначимо, що на перших етапах дослідження замість показника Upt використовувались значення Таблиця 1. Загальна характеристика хворих на АТГЕ за даними НПТ  [18,19]. However, the mentioned methods are based on calculating the ratio of RPh accumulation in ROI to the cerebellum or the brain. Therefore, using volume of cerebral blood flow as a basic value for implementation of the method given in this work, is not relevant since it requires additional calculation with its errors. In particular, Lassen's method of calculating volume of cerebral blood flow is irrelevant, because in this case RPh accumulation in ROI «cerebellum» which have low correlation relationships with NPT data, is used as a reference zone.  ( 1) 2

Method for comparing SPECT and NPT data
After finding the matrices of ratios, scatterograms between SPECT data and NPT results were built for every patient. The obtained scatterograms were approximated by the line: where x is SPECT data, y is the result of a certain test, a, b are equation approximation coefficients. As a result of analyzing and comparing the ratios of SPECT values to NPT data, the scatterograms, in which correlation coefficients (R) between the approximation curve and values on the scatterogram were maximal, were highlighted. Thus, the search for the most optimal combinations of SPECT values that characterizes neurocognitive condition of the patients can be classified as regressional method of machine learning.
The values of the relations of IV order were also analyzed, Nevertheless, the latter didn't show more accurate results of the calculation of NPT values according to SPECT data than the relations of III order in most cases. That's why the respective data are not presented in the study.
It should be noted that the most significant factors for 3D SPECT images processing for the given method of calculation of NPT values which affect accuracy, are variability of filtering quality (smoothing) of the images and, most importantly, the accuracy of the segmentation.

The results of machine learning for
The data in Tables 2 and 3 show that results of machine learning according to both values of ratios of RPh accumulation and IHS values have the same R order and a mean squared error D (the latter can act as a criterion for a credible interval). It indicates a similar diagnostic value of quantitative criteria Upt and q. It should be emphasiszed that Upt is a quantitave characteristic of RPh accumulation in ROI, whereas q is mostly a characteristic of spatial distribution of RPh. The independence of Upt and q makes it possible to increase the accuracy of calculations of NPT values by algebraic averaging of calculations by U Upt і U q ratios. The result of this Ukrainian journal of radiology and oncology. 2022;30 (1) averaging is given in Table 4, where intead of SPECT data, which were used in the learning, the quantitative values of the third variant of scintigraphic data processing of the patients were used. The given results indicate quite a high correlation dependence between SPECT and NPT data -R is in the range of 0,75 -0,93; the average relative error -in the range of 7 -22%. The graphic demonstration of the results of machine learning are shown in Fig. 2 . Thus, it can be concluded that, for the first time, the method of quantitative assessment of the neuropsychological and cognitive state of patients was developed on the basis of an objective research method. The accuracy of calculations can be significantly increased Рис. 2. Скатерограми між бальними оцінками тестів НПТ та відповідними результатами машинного навчання за даними ОФЕКТ: o -навчання за значеннями ВПС; × -навчання за Upt; • -усереднені результати навчання й відповідні їм пунктирні лінії апроксимації Fig. 2. Scatterograms between the scores of NPT tests and respective results of machine learning according to SPECT data: o -learning according to IHS values; × -learning according to Upt; • -averaged results of the learning and respective dotted lines of approximation by minimizing the subjective assessment of the quality of prior processing of scintigraphic images (in our study this examination was deliberately roughened for estimating the error of calculations taking into account a relatively small number of patients); by increasing the accuracy of 3D segmentation of the brain images; by increasing the matrix of SPECT project frames from 64 to 128. методу дослідження. Відзначимо, що точність розрахунків може бути суттєво збільшена за рахунок мінімізації суб'єктивної оцінки якості попередньої обробки сцинтиграфічних зображень (в даній роботі ця операція навмисно була загрублена, для оцінки похибки розрахунків з урахуванням не досить великої кількості пацієнтів); підвищення точності сегментації 3D зображень ГМ; збільшення матриці проекційних кадрів ОФЕКТ з 64 до 128.
To increase the accuracy of assessment of the neuropsychological and cognitive state of the patients it is appropriate to maximally automate the process of prior processing of scintigraphic data, to increase the accuracy of 3D segmentation of the brain images and to use SPECT with 128×128 matrices.