BACKGROUND: Brain damage after ischemic stroke results in changes in a wide range of structural and functional brain networks [1]. Scientific studies show that although stroke is primarily a focal lesion, it also affects the functional connectivity of anatomical and functional regions, often resulting in altered integration of brain networks and affecting whole-brain function, leading to cognitive and emotional impairment [2, 3]. AIM: The aim of the study was to determine changes in functional brain connectivity during hypersomnia in patients with acute ischemic stroke. MATERIALS AND METHODS: A total of 44 patients with acute ischemic stroke were examined. The participants were divided into two groups based on the presence of sleep disorders. Group 1 included 22 patients with hypersomnia, which was objectively confirmed by polysomnography. Group 2 also included 22 patients who did not have sleep disorders and constituted the control group. The age of patients in both groups ranged from 45 to 65 years. All patients underwent magnetic resonance imaging on tomographs with a magnetic field induction strength of 1.5 Tesla, using the standard protocol and special pulse sequences of T-gradient echo 3D MPRAGE and BOLD. Resting-state functional magnetic resonance imaging of the brain was employed to assess functional connectivity. Postprocessing was conducted on specialized software, CONN-TOOLBOX, which generated appropriate graphical representations of quantitative results based on the selection of zones of interest. RESULTS: In patients experiencing the acute phase of ischemic stroke, hypersomnia results in the strengthening of functional connections, predominantly in the temporo-occipital and parietal regions. This may be associated with impaired visual perception, memory, and spatial orientation. Additionally, there is a weakening of functional connections in the frontal and occipital cortex, which may indicate confusion of thinking and disorders of speech, arbitrary movements, and the regulation of complex behaviors. The disruption of the functional connections between the medial prefrontal cortex and the posterior cingulate cortex and the cerebellum is indicative of impaired coordination and regulation of balance and muscle tone. However, it also has the potential to affect emotional, cognitive, and behavioral changes in the brain. CONCLUSIONS: Resting-state functional magnetic resonance imaging is a technique that allows for the determination of changes in functional brain connections during hypersomnia in patients with acute ischemic stroke. Additionally, it enables the identification of neuroimaging markers corresponding to this pathology.
{"title":"Changes in the functional connections of the brain at rest in patients with acute ischemic stroke and hypersomnia","authors":"L. I. Trushina","doi":"10.17816/dd627000","DOIUrl":"https://doi.org/10.17816/dd627000","url":null,"abstract":"BACKGROUND: Brain damage after ischemic stroke results in changes in a wide range of structural and functional brain networks [1]. Scientific studies show that although stroke is primarily a focal lesion, it also affects the functional connectivity of anatomical and functional regions, often resulting in altered integration of brain networks and affecting whole-brain function, leading to cognitive and emotional impairment [2, 3]. \u0000AIM: The aim of the study was to determine changes in functional brain connectivity during hypersomnia in patients with acute ischemic stroke. \u0000MATERIALS AND METHODS: A total of 44 patients with acute ischemic stroke were examined. The participants were divided into two groups based on the presence of sleep disorders. Group 1 included 22 patients with hypersomnia, which was objectively confirmed by polysomnography. Group 2 also included 22 patients who did not have sleep disorders and constituted the control group. The age of patients in both groups ranged from 45 to 65 years. \u0000All patients underwent magnetic resonance imaging on tomographs with a magnetic field induction strength of 1.5 Tesla, using the standard protocol and special pulse sequences of T-gradient echo 3D MPRAGE and BOLD. Resting-state functional magnetic resonance imaging of the brain was employed to assess functional connectivity. Postprocessing was conducted on specialized software, CONN-TOOLBOX, which generated appropriate graphical representations of quantitative results based on the selection of zones of interest. \u0000RESULTS: In patients experiencing the acute phase of ischemic stroke, hypersomnia results in the strengthening of functional connections, predominantly in the temporo-occipital and parietal regions. This may be associated with impaired visual perception, memory, and spatial orientation. Additionally, there is a weakening of functional connections in the frontal and occipital cortex, which may indicate confusion of thinking and disorders of speech, arbitrary movements, and the regulation of complex behaviors. \u0000The disruption of the functional connections between the medial prefrontal cortex and the posterior cingulate cortex and the cerebellum is indicative of impaired coordination and regulation of balance and muscle tone. However, it also has the potential to affect emotional, cognitive, and behavioral changes in the brain. \u0000CONCLUSIONS: Resting-state functional magnetic resonance imaging is a technique that allows for the determination of changes in functional brain connections during hypersomnia in patients with acute ischemic stroke. Additionally, it enables the identification of neuroimaging markers corresponding to this pathology.","PeriodicalId":34831,"journal":{"name":"Digital Diagnostics","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-07-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141681566","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
O. V. Savva, U. N. Tumanova, V. Bychenko, A. I. Shchegolev
BACKGROUND: During pathological and forensic autopsies, the bodies of the deceased are examined to identify nonspecific cadaveric changes. These changes include internal hypostases, which are characterized by the redistribution of blood in tissues and organs under the influence of gravity [1, 2]. Such postmortem hypostases reflect the age of death, but they also complicate the differential diagnosis of lifetime pathological processes and lesions with nonspecific cadaveric changes [3, 4]. Postmortem magnetic resonance imaging represents an objective and noninvasive method of investigation, particularly in cases of neonatal death characterized by relative immaturity of organs and tissues. It may therefore prove to be a promising approach to visualize and evaluate cadaveric hypostases [5, 6]. AIM: The aim of this study was to investigate the manifestations of cadaveric hypostases in the liver of deceased neonates, with a focus on the impact of postmortem period duration. This was achieved through the use of postmortem magnetic resonance imaging and morphologic examination. MATERIALS AND METHODS: The study was based on a comprehensive postmortem radiology and pathological anatomical examination of the bodies of 62 newborns and infants who died at the age of 1.5 hours to 49 days. The subjects were selected to exclude those with developmental anomalies and liver diseases. A postmortem magnetic resonance imaging examination was conducted on a 3T Siemens Magnetom Verio apparatus, followed by a subsequent pathological and anatomic autopsy. The T1- and T2-weighted images were evaluated to determine the presence and severity of the magnetic resonance signal intensity gradient line in the ventral (superior) and dorsal (inferior) regions of the liver tissue. Following the autopsy, tissue samples were obtained from the ventral and dorsal regions of the liver, and subsequently subjected to microscopic analysis of hematoxylin and eosin-stained preparations. RESULTS: The results of postmortem magnetic resonance imaging have enabled the establishment of the radiation characteristics and histological changes in liver tissue caused by cadaveric hypostases. The most notable manifestation of cadaveric hypostases in the liver at postmortem magnetic resonance imaging is the change in magnetic resonance signal intensities in the above and below-located regions of the organ, accompanied by the emergence of a signal intensity gradient. This gradient reflects the location of the body after death and varies depending on the duration of the postmortem period. The signal intensity gradient was more frequently observed on T1-weighted images compared to T2-weighted images. Histological examination of liver tissue preparations revealed an increase in the size of sinusoids and a decrease in the area of hepatic beams, which was observed to progress with increasing age at death and was expressed to a greater extent in the lower liver region. These changes are undoubtedly a morpho
{"title":"Postmortem liver hypostases in newborns: radiation and pathological characteristics","authors":"O. V. Savva, U. N. Tumanova, V. Bychenko, A. I. Shchegolev","doi":"10.17816/dd625987","DOIUrl":"https://doi.org/10.17816/dd625987","url":null,"abstract":"BACKGROUND: During pathological and forensic autopsies, the bodies of the deceased are examined to identify nonspecific cadaveric changes. These changes include internal hypostases, which are characterized by the redistribution of blood in tissues and organs under the influence of gravity [1, 2]. Such postmortem hypostases reflect the age of death, but they also complicate the differential diagnosis of lifetime pathological processes and lesions with nonspecific cadaveric changes [3, 4]. Postmortem magnetic resonance imaging represents an objective and noninvasive method of investigation, particularly in cases of neonatal death characterized by relative immaturity of organs and tissues. It may therefore prove to be a promising approach to visualize and evaluate cadaveric hypostases [5, 6]. \u0000AIM: The aim of this study was to investigate the manifestations of cadaveric hypostases in the liver of deceased neonates, with a focus on the impact of postmortem period duration. This was achieved through the use of postmortem magnetic resonance imaging and morphologic examination. \u0000MATERIALS AND METHODS: The study was based on a comprehensive postmortem radiology and pathological anatomical examination of the bodies of 62 newborns and infants who died at the age of 1.5 hours to 49 days. The subjects were selected to exclude those with developmental anomalies and liver diseases. A postmortem magnetic resonance imaging examination was conducted on a 3T Siemens Magnetom Verio apparatus, followed by a subsequent pathological and anatomic autopsy. The T1- and T2-weighted images were evaluated to determine the presence and severity of the magnetic resonance signal intensity gradient line in the ventral (superior) and dorsal (inferior) regions of the liver tissue. Following the autopsy, tissue samples were obtained from the ventral and dorsal regions of the liver, and subsequently subjected to microscopic analysis of hematoxylin and eosin-stained preparations. \u0000RESULTS: The results of postmortem magnetic resonance imaging have enabled the establishment of the radiation characteristics and histological changes in liver tissue caused by cadaveric hypostases. The most notable manifestation of cadaveric hypostases in the liver at postmortem magnetic resonance imaging is the change in magnetic resonance signal intensities in the above and below-located regions of the organ, accompanied by the emergence of a signal intensity gradient. This gradient reflects the location of the body after death and varies depending on the duration of the postmortem period. The signal intensity gradient was more frequently observed on T1-weighted images compared to T2-weighted images. Histological examination of liver tissue preparations revealed an increase in the size of sinusoids and a decrease in the area of hepatic beams, which was observed to progress with increasing age at death and was expressed to a greater extent in the lower liver region. These changes are undoubtedly a morpho","PeriodicalId":34831,"journal":{"name":"Digital Diagnostics","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-07-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141682661","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
A. Dushkin, M. Afanasiev, S. S. Afanasiev, T. Grishacheva, A. Karaulov
BACKGROUND: Visual inspection and colposcopy are subjective methods of cervical evaluation. Currently, the majority of colposcopes are equipped with the capacity to digitally transmit and record cervical images, in addition to modern software for image processing. For the objective assessment, prevention of development, and risk assessment of precancerous changes (SIL+) and cervical cancer, it is essential to use modern methods of image processing. AIM: The study aimed at demonstrating the capabilities of digital analysis of cervical images based on ImageJ software [1]. MATERIALS AND METHODS: A total of 500 colposcopic images of the Schiller test were obtained during dilated colposcopy. Digital analysis was performed using ImageJ software, which employed minimum (MinGV) and maximum (MaxGV) gray pixel values (0–255) and lesion surface area (%Area) as parameters. The images were divided into 4 groups according to the cytologic examination performed: healthy donors (n=19; 3.8%), mild grade squamous cell intraepithelial lesion (n=113; 22.6%), severe grade squamous cell intraepithelial lesion (n=327; 65.4%), and invasive cervical cancer (n=41; 8.2%). Mathematical and statistical analysis of the obtained data was performed using Python programming language packages in the Google Colab environment. Comparisons of quantitative measures between three or more groups were conducted using the Kruskal-Wallis criterion and posteriori comparisons by Dunn’s criterion with Holm’s correction. RESULSTS: Statistical significance was observed in the increase of MinGV (p=0.035), MaxGV (p0.001) and %Area (p=0.022) from the mild (88/141/31) to the severe (83/142/32) degree of squamous cell intraepithelial lesion and cervical cancer (88/162/36). Objective parameters for the assessment of the degree of cervical surface lesions during digital colposcopy were obtained. Digital analysis of the cervical surface may assist the clinical specialist in determining further management strategies, including scarification or incisional biopsy with subsequent morphological examination. CONCLUSIONS: The application of digital analysis to colposcopic images has the potential to reduce the subjective assessment of cervical condition, enhance the efficiency of the initial appointment with a gynecologist, and facilitate the selection of patients for cytologic examination.
{"title":"Digital approach to estimate clinical images of the cervix with ImageJ software","authors":"A. Dushkin, M. Afanasiev, S. S. Afanasiev, T. Grishacheva, A. Karaulov","doi":"10.17816/dd626768","DOIUrl":"https://doi.org/10.17816/dd626768","url":null,"abstract":"BACKGROUND: Visual inspection and colposcopy are subjective methods of cervical evaluation. Currently, the majority of colposcopes are equipped with the capacity to digitally transmit and record cervical images, in addition to modern software for image processing. For the objective assessment, prevention of development, and risk assessment of precancerous changes (SIL+) and cervical cancer, it is essential to use modern methods of image processing. \u0000AIM: The study aimed at demonstrating the capabilities of digital analysis of cervical images based on ImageJ software [1]. \u0000MATERIALS AND METHODS: A total of 500 colposcopic images of the Schiller test were obtained during dilated colposcopy. Digital analysis was performed using ImageJ software, which employed minimum (MinGV) and maximum (MaxGV) gray pixel values (0–255) and lesion surface area (%Area) as parameters. The images were divided into 4 groups according to the cytologic examination performed: healthy donors (n=19; 3.8%), mild grade squamous cell intraepithelial lesion (n=113; 22.6%), severe grade squamous cell intraepithelial lesion (n=327; 65.4%), and invasive cervical cancer (n=41; 8.2%). Mathematical and statistical analysis of the obtained data was performed using Python programming language packages in the Google Colab environment. Comparisons of quantitative measures between three or more groups were conducted using the Kruskal-Wallis criterion and posteriori comparisons by Dunn’s criterion with Holm’s correction. \u0000RESULSTS: Statistical significance was observed in the increase of MinGV (p=0.035), MaxGV (p0.001) and %Area (p=0.022) from the mild (88/141/31) to the severe (83/142/32) degree of squamous cell intraepithelial lesion and cervical cancer (88/162/36). Objective parameters for the assessment of the degree of cervical surface lesions during digital colposcopy were obtained. Digital analysis of the cervical surface may assist the clinical specialist in determining further management strategies, including scarification or incisional biopsy with subsequent morphological examination. \u0000CONCLUSIONS: The application of digital analysis to colposcopic images has the potential to reduce the subjective assessment of cervical condition, enhance the efficiency of the initial appointment with a gynecologist, and facilitate the selection of patients for cytologic examination.","PeriodicalId":34831,"journal":{"name":"Digital Diagnostics","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-07-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141682773","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
B. Borodulin, Y.T. Gogoberidze, K. Zhilinskaya, I. A. Prosvirkin, R. A. Sabitov
BACKGROUND: The volume of medical diagnostic studies continues to increase annually, intensifying the desire to implement advanced technologies in the field of medical diagnostics. One of the promising approaches that has attracted attention is the use of artificial intelligence in this area. A study was conducted on the automated analysis of chest radiographs using the AI service PhthisisBioMed at a city hospital specializing in the treatment of respiratory diseases. AIM: The study aimed to assess the diagnostic accuracy of the artificial intelligence service “PhthisisBioMed” for the detection of respiratory pathologies in the context of a city hospital that provides 24-hour specialized care in the field of pulmonology. MATERIALS AND METHODS: This study employed a prospective design, with the results of the artificial intelligence service available to the physician on request. This enabled the physician to review the results of the service if an alternative opinion was needed. The reference test was conducted by radiologists at Samara City Hospital No. 4, who described the examinations performed during the testing period. The index test was performed on the software “Program for Automated Analysis of Digital Chest Radiographs/Fluorograms according to TU 62.01.29-001-96876180-2019” produced by PhthisisBioMed LLC. The PhthisisBioMed software was employed to analyze digital fluorograms of the lungs in direct anterior projection. The software automatically identified the following radiological signs of pathologies: pleural effusion, pneumothorax, atelectasis, darkening, infiltration/consolidation, dissemination, cavity, calcification/calcified shadow, and cortical layer integrity violation. Fluorograms of patients over the age of 18 were included in the analysis. The study was conducted within the framework of research and development work No. 121051700033-3, entitled “Lung Damage of Infectious Etiology. Improvement of Methods of Detection, Diagnosis and Treatment” (14.05.2021). RESULTS: Following the pilot operation of the PhthisisBioMed artificial intelligence service and subsequent ROC analysis, the diagnostic accuracy metrics claimed by the manufacturer of the artificial intelligence medical device were confirmed. The service provided the probability of the presence of various pathologies. According to the highlighted labels, 63 patients (4.8%) were suspected of tuberculosis based on characteristic radiologic features. The conclusion was made independently by the radiologist, and the results were evaluated by the attending physician. The attending physician had the opportunity to compare the results and discuss them with the radiologist if differences were found. The results of the survey of pulmonologists who participated in the study indicated that the conclusion of the artificial intelligence service was received automatically within 15 seconds, while the conclusion of the physician was received within 30 minutes or more. CONCLUSIONS:
{"title":"The experience of using artificial intelligence for automated analysis of digital radiographs in a city hospital","authors":"B. Borodulin, Y.T. Gogoberidze, K. Zhilinskaya, I. A. Prosvirkin, R. A. Sabitov","doi":"10.17816/dd629896","DOIUrl":"https://doi.org/10.17816/dd629896","url":null,"abstract":"BACKGROUND: The volume of medical diagnostic studies continues to increase annually, intensifying the desire to implement advanced technologies in the field of medical diagnostics. One of the promising approaches that has attracted attention is the use of artificial intelligence in this area. A study was conducted on the automated analysis of chest radiographs using the AI service PhthisisBioMed at a city hospital specializing in the treatment of respiratory diseases. \u0000AIM: The study aimed to assess the diagnostic accuracy of the artificial intelligence service “PhthisisBioMed” for the detection of respiratory pathologies in the context of a city hospital that provides 24-hour specialized care in the field of pulmonology. \u0000MATERIALS AND METHODS: This study employed a prospective design, with the results of the artificial intelligence service available to the physician on request. This enabled the physician to review the results of the service if an alternative opinion was needed. \u0000The reference test was conducted by radiologists at Samara City Hospital No. 4, who described the examinations performed during the testing period. The index test was performed on the software “Program for Automated Analysis of Digital Chest Radiographs/Fluorograms according to TU 62.01.29-001-96876180-2019” produced by PhthisisBioMed LLC. The PhthisisBioMed software was employed to analyze digital fluorograms of the lungs in direct anterior projection. The software automatically identified the following radiological signs of pathologies: pleural effusion, pneumothorax, atelectasis, darkening, infiltration/consolidation, dissemination, cavity, calcification/calcified shadow, and cortical layer integrity violation. \u0000Fluorograms of patients over the age of 18 were included in the analysis. The study was conducted within the framework of research and development work No. 121051700033-3, entitled “Lung Damage of Infectious Etiology. Improvement of Methods of Detection, Diagnosis and Treatment” (14.05.2021). \u0000RESULTS: Following the pilot operation of the PhthisisBioMed artificial intelligence service and subsequent ROC analysis, the diagnostic accuracy metrics claimed by the manufacturer of the artificial intelligence medical device were confirmed. \u0000The service provided the probability of the presence of various pathologies. According to the highlighted labels, 63 patients (4.8%) were suspected of tuberculosis based on characteristic radiologic features. The conclusion was made independently by the radiologist, and the results were evaluated by the attending physician. The attending physician had the opportunity to compare the results and discuss them with the radiologist if differences were found. \u0000The results of the survey of pulmonologists who participated in the study indicated that the conclusion of the artificial intelligence service was received automatically within 15 seconds, while the conclusion of the physician was received within 30 minutes or more. \u0000CONCLUSIONS:","PeriodicalId":34831,"journal":{"name":"Digital Diagnostics","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-07-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141681313","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Maksim V. Solominov, Denis V. Pakhomov, Tatiana A. Zagriazkina
BACKGROUND: Ischemic stroke is a significant contributor to mortality rates in Russia and globally [1]. Computed tomographic angiography is a primary diagnostic tool for ischemic stroke, enabling the identification of stenosis or occlusion in cerebral arteries. The majority of ischemic strokes (51%) occur in the middle cerebral artery region [2], underscoring the growing interest in evaluating blood flow in this area of the brain. The manual detection of stenoses is characterised by subjective evaluation and requires a considerable amount of time. The automation of middle cerebral artery narrowing detection represents a significant challenge in computed tomographic angiography image analysis. AIM: The study aims to develop an algorithm for the automatic detection of stenoses in the middle cerebral artery on DICOM images of computed tomographic angiography based on the application of artificial neural networks, vascularity assessment and skeletonization algorithms. MATERIALS AND METHODS: A total of 262 computed tomographic angiography series from patients at the N.V. Sklifosovsky Emergency Medical Research Institute were analyzed. Of these, 94 series exhibited stenosis in the M1/M2 segment of the middle cerebral artery. The image processing was conducted using an artificial neural network with a CFPNet-M architecture [3]. The reconstruction of the vascular tree was based on the calculation of the "vesselness" measure [4] with subsequent skeletonization of the identified structures. RESULTS: In the initial stage, a neural network for the segmentation of the middle cerebral artery basin was trained. The training array was generated using the MNI152 template with affine transformations and subsequent expert evaluation. In this case, the IoU (Intersection over Union) measure was 0.81. The primary objective was the segmentation of the middle cerebral artery vascular tree, which was achieved through the use of the vesselness filter, followed by an evaluation of voxel intensities and the identification of the connected object with the longest length. The next stage involved the construction of the skeleton of the middle cerebral artery. This entailed determining the centerline of the vessel and representing the resulting skeleton as a graph with the vessels as edges and their bifurcation points as vertices. The subsequent stage was the calculation of morphological features (diameter, area, and perimeter) in the cross-sectional plane for each segment (the area between the bifurcation points). Finally, the area of constriction was determined based on the analysis of the behavior of the segment cross-sections and the identification of any deviation from the threshold value. The overall accuracy of the algorithm was 79.39% (95% confidence interval 73.98–84.12), with a sensitivity of 80.85% (95% confidence interval 71.44–88.24) and a specificity of 78.57% (95% confidence interval 71.59–84.52). CONCLUSIONS: Thus, we developed an algorithm for the detection
{"title":"Application of machine learning methods and medical image processing in solving the problem of detecting stenoses of the middle cerebral artery according to computed tomographic angiography data","authors":"Maksim V. Solominov, Denis V. Pakhomov, Tatiana A. Zagriazkina","doi":"10.17816/dd626181","DOIUrl":"https://doi.org/10.17816/dd626181","url":null,"abstract":"BACKGROUND: Ischemic stroke is a significant contributor to mortality rates in Russia and globally [1]. Computed tomographic angiography is a primary diagnostic tool for ischemic stroke, enabling the identification of stenosis or occlusion in cerebral arteries. The majority of ischemic strokes (51%) occur in the middle cerebral artery region [2], underscoring the growing interest in evaluating blood flow in this area of the brain. The manual detection of stenoses is characterised by subjective evaluation and requires a considerable amount of time. The automation of middle cerebral artery narrowing detection represents a significant challenge in computed tomographic angiography image analysis. \u0000AIM: The study aims to develop an algorithm for the automatic detection of stenoses in the middle cerebral artery on DICOM images of computed tomographic angiography based on the application of artificial neural networks, vascularity assessment and skeletonization algorithms. \u0000MATERIALS AND METHODS: A total of 262 computed tomographic angiography series from patients at the N.V. Sklifosovsky Emergency Medical Research Institute were analyzed. Of these, 94 series exhibited stenosis in the M1/M2 segment of the middle cerebral artery. The image processing was conducted using an artificial neural network with a CFPNet-M architecture [3]. The reconstruction of the vascular tree was based on the calculation of the \"vesselness\" measure [4] with subsequent skeletonization of the identified structures. \u0000RESULTS: In the initial stage, a neural network for the segmentation of the middle cerebral artery basin was trained. The training array was generated using the MNI152 template with affine transformations and subsequent expert evaluation. In this case, the IoU (Intersection over Union) measure was 0.81. The primary objective was the segmentation of the middle cerebral artery vascular tree, which was achieved through the use of the vesselness filter, followed by an evaluation of voxel intensities and the identification of the connected object with the longest length. The next stage involved the construction of the skeleton of the middle cerebral artery. This entailed determining the centerline of the vessel and representing the resulting skeleton as a graph with the vessels as edges and their bifurcation points as vertices. The subsequent stage was the calculation of morphological features (diameter, area, and perimeter) in the cross-sectional plane for each segment (the area between the bifurcation points). Finally, the area of constriction was determined based on the analysis of the behavior of the segment cross-sections and the identification of any deviation from the threshold value. The overall accuracy of the algorithm was 79.39% (95% confidence interval 73.98–84.12), with a sensitivity of 80.85% (95% confidence interval 71.44–88.24) and a specificity of 78.57% (95% confidence interval 71.59–84.52). \u0000CONCLUSIONS: Thus, we developed an algorithm for the detection","PeriodicalId":34831,"journal":{"name":"Digital Diagnostics","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-07-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141680932","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
BACKGROUND: Medical workers are one of the professions that significantly strengthen the country’s economy [1, 2]. The development and implementation of health-saving technologies to prolong the effective professional life of medical workers of older age groups will preserve them as a labor resource of the country, which will exclude economic losses of the state. The results of a sociological survey of doctors providing medical care in the polyclinic segment of Moscow and the Moscow region were used to assess the professional competencies of specialists [3]. An organizational technology was then formed based on these findings, which was subsequently proposed for expert assessment. AIM: The aim of this study was to ascertain the significance of organizational technology measures for the professional longevity of doctors in older age groups. MATERIALS AND METHODS: The study employed a multi-methodological approach, integrating sociological, statistical, and expert evaluation techniques. A total of 50 experts were invited to rank the activities comprising the integrated technology in terms of their perceived importance for achieving the desired outcome, namely, the support of effective professional longevity of doctors in the event of the implementation of such technology. The experts were specialists in the field of health care and public health, including chief physicians and heads of departments of urban polyclinics in Moscow and the Moscow region, who were accredited in the specialty 03.02.03. They had experience of management in the field of health care ranging from one to 29 years. RESULTS: All experts concur that a medical organization should implement measures to prevent the deterioration of doctors due to aging. The necessity to test doctors over the age of 50 for cognitive disorders and dementia was confirmed by 90% of experts. Additionally, 60% of experts agreed that doctors over the age of 50 require a less demanding work schedule, including a reduction in intellectual workload and an extension of rest periods. At the same time, 20% of experts approve of the transition to a lighter work regime on an individual basis after testing, 10% agreed only with the prolongation of rest, and 10% gave a negative answer. In the opinion of 90% of experts, the widespread introduction of medical information systems (and training in working with them) will help to support the effective professional longevity of doctors of older age groups. A mere 40% of experts concurred that the transfer of senior physicians to monoprofile appointments would assist in prolonging their effective professional longevity. The majority of experts (80%) recommend regular cognitive training for doctors of advanced age, while 10% believe it is only necessary in specific cases and 10% are opposed to the idea. Only 70% of experts in the medical field implement organizational measures to maintain effective professional longevity, while the remaining 30% employ single measures.
{"title":"Expert assessment of the organization of comprehensive support for the prolongation of professional effective activity of a doctor","authors":"A. Vorobeva, M. Yakushin","doi":"10.17816/dd627052","DOIUrl":"https://doi.org/10.17816/dd627052","url":null,"abstract":"BACKGROUND: Medical workers are one of the professions that significantly strengthen the country’s economy [1, 2]. The development and implementation of health-saving technologies to prolong the effective professional life of medical workers of older age groups will preserve them as a labor resource of the country, which will exclude economic losses of the state. \u0000The results of a sociological survey of doctors providing medical care in the polyclinic segment of Moscow and the Moscow region were used to assess the professional competencies of specialists [3]. An organizational technology was then formed based on these findings, which was subsequently proposed for expert assessment. \u0000AIM: The aim of this study was to ascertain the significance of organizational technology measures for the professional longevity of doctors in older age groups. \u0000MATERIALS AND METHODS: The study employed a multi-methodological approach, integrating sociological, statistical, and expert evaluation techniques. A total of 50 experts were invited to rank the activities comprising the integrated technology in terms of their perceived importance for achieving the desired outcome, namely, the support of effective professional longevity of doctors in the event of the implementation of such technology. The experts were specialists in the field of health care and public health, including chief physicians and heads of departments of urban polyclinics in Moscow and the Moscow region, who were accredited in the specialty 03.02.03. They had experience of management in the field of health care ranging from one to 29 years. \u0000RESULTS: All experts concur that a medical organization should implement measures to prevent the deterioration of doctors due to aging. The necessity to test doctors over the age of 50 for cognitive disorders and dementia was confirmed by 90% of experts. Additionally, 60% of experts agreed that doctors over the age of 50 require a less demanding work schedule, including a reduction in intellectual workload and an extension of rest periods. At the same time, 20% of experts approve of the transition to a lighter work regime on an individual basis after testing, 10% agreed only with the prolongation of rest, and 10% gave a negative answer. In the opinion of 90% of experts, the widespread introduction of medical information systems (and training in working with them) will help to support the effective professional longevity of doctors of older age groups. A mere 40% of experts concurred that the transfer of senior physicians to monoprofile appointments would assist in prolonging their effective professional longevity. The majority of experts (80%) recommend regular cognitive training for doctors of advanced age, while 10% believe it is only necessary in specific cases and 10% are opposed to the idea. Only 70% of experts in the medical field implement organizational measures to maintain effective professional longevity, while the remaining 30% employ single measures.","PeriodicalId":34831,"journal":{"name":"Digital Diagnostics","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-07-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141682121","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
E. Poliker, Konstantin A. Koshechkin, Alexander M. Timokhin, Ekaterina V. Klyukina, Ekaterina D. Belyakova, Artem M. Brovko, Alina S. Lalayan, A. Ermolaeva
BACKGROUND: In the last decade, there has been a significant increase in interest in non-invasive monitoring of blood glucose levels [1]. This is driven by the desire to reduce patient discomfort, as well as the risk of infections associated with traditional invasive methods [2]. Raman spectroscopy, considered as a promising approach for non-invasive measurements [3], combined with machine learning, has the potential to lead to more accurate and faster diagnostic methods for conditions related to glucose imbalances [4]. AIMS: Development and validation of a new portable glucometer based on Raman spectroscopy using machine learning methods for non-invasive determination of glycated hemoglobin (HbA1c) levels. MATERIALS AND METHODS: The study was conducted on a sample of 100 volunteers of different age groups and genders, with varying health statuses, including individuals with type 1 and type 2 diabetes and those without diabetes. To collect data, we used a portable device developed by us, based on the registration of Raman spectra with laser excitation at 638 nm. The data were analyzed using Support Vector Machine neural networks. RESULTS: After processing the spectroscopic measurements using Support Vector Machine, the system showed sensitivity (95,7%) and specificity (84,2%) in determining HbA1c levels comparable to traditional methods such as high-performance liquid chromatography. It was found that the algorithm is sufficiently adaptive and can be used across a wide range of skin types, regardless of the age and gender of the participants. The results suggest the possibility of using the developed device in clinical practice. CONCLUSION: The developed portable glucometer based on Raman spectroscopy combined with machine learning algorithms could be a promising step towards non-invasive and continuous monitoring of glycemic levels in patients with diabetes.
{"title":"Using neural networks for non-invasive determination of glycated hemoglobin levels, illustrated by the application of an innovative portable glucometer in clinical practice","authors":"E. Poliker, Konstantin A. Koshechkin, Alexander M. Timokhin, Ekaterina V. Klyukina, Ekaterina D. Belyakova, Artem M. Brovko, Alina S. Lalayan, A. Ermolaeva","doi":"10.17816/dd627099","DOIUrl":"https://doi.org/10.17816/dd627099","url":null,"abstract":"BACKGROUND: In the last decade, there has been a significant increase in interest in non-invasive monitoring of blood glucose levels [1]. This is driven by the desire to reduce patient discomfort, as well as the risk of infections associated with traditional invasive methods [2]. Raman spectroscopy, considered as a promising approach for non-invasive measurements [3], combined with machine learning, has the potential to lead to more accurate and faster diagnostic methods for conditions related to glucose imbalances [4]. \u0000AIMS: Development and validation of a new portable glucometer based on Raman spectroscopy using machine learning methods for non-invasive determination of glycated hemoglobin (HbA1c) levels. \u0000MATERIALS AND METHODS: The study was conducted on a sample of 100 volunteers of different age groups and genders, with varying health statuses, including individuals with type 1 and type 2 diabetes and those without diabetes. To collect data, we used a portable device developed by us, based on the registration of Raman spectra with laser excitation at 638 nm. The data were analyzed using Support Vector Machine neural networks. \u0000RESULTS: After processing the spectroscopic measurements using Support Vector Machine, the system showed sensitivity (95,7%) and specificity (84,2%) in determining HbA1c levels comparable to traditional methods such as high-performance liquid chromatography. It was found that the algorithm is sufficiently adaptive and can be used across a wide range of skin types, regardless of the age and gender of the participants. The results suggest the possibility of using the developed device in clinical practice. \u0000CONCLUSION: The developed portable glucometer based on Raman spectroscopy combined with machine learning algorithms could be a promising step towards non-invasive and continuous monitoring of glycemic levels in patients with diabetes.","PeriodicalId":34831,"journal":{"name":"Digital Diagnostics","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-07-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141681491","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
E. V. Kazantseva, A. Ivannikov, A. Tarzimanova, V. Podzolkov
BACKGROUND: Arterial hypertension and chronic obstructive pulmonary disease have a deleterious effect on the structure of the heart, leading to the development of atrial fibrillation, which remains the leading cause of cerebral stroke and premature death [1]. Consequently, the early identification of atrial fibrillation risk factors in patients with arterial hypertension and chronic obstructive pulmonary disease is of paramount importance for the prevention of such conditions. This is why predictive cardiology employs machine learning methods, which are demonstrably superior to classical statistical methods of prediction [2–4]. AIM: The study aimed to develop a prognostic model of atrial fibrillation in comorbid patients with arterial hypertension and chronic obstructive pulmonary disease based on multilayer perceptron. MATERIALS AND METHODS: The study included 419 patients treated at the University Clinical Hospital No. 4 of the I.M. Sechenov First Moscow State Medical University. Group 1 consisted of 91 (21.7%) patients with a verified diagnosis of atrial fibrillation, while Group 2 comprised 328 (78.3%) patients without atrial fibrillation. The random forest machine learning algorithm was used to identify predictors, which were then utilized to develop a neural network of the multilayer perceptron type. This consisted of two layers: an input layer of 12 neurons with the ReLU activation function and an output layer that receives input data from the previous layer and transmits them to one output with the sigmoid activation function. The threshold value, sensitivity, specificity, and diagnostic efficiency of the obtained model were determined using receiver operating characteristic analysis with the calculation of the area under the curve (AUC). RESULTS: By the first stage of prognostic model development, the most significant predictors of atrial fibrillation development were selected by the random forest machine learning algorithm. The model was developed using three variables: C-reactive protein concentration (odds ratio, OR 1.04; 95% confidence interval, CI 1.015–1.067; p=0.002), erythrocyte sedimentation rate (OR 1.04; 95% CI 1.019–1.069; p=0.002), and creatinine concentration (OR 1.03; 95% CI 1.011–1.042; p 0.001). These variables were used to train a multilayer perceptron model on a test sample for 500 epochs. Following training, the developed model exhibited a sensitivity of 85%, a specificity of 80%, and a diagnostic efficiency of 79.6%. AUC amounted to 0.900. CONCLUSIONS: The study resulted in the development of a prognostic model based on the application of machine learning methods, which exhibited favorable metrics. This model may be considered a valuable tool for clinical practice.
{"title":"Predicting atrial fibrillation in comorbid patients with arterial hypertension and chronic obstructive pulmonary disease using laboratory research methods: a machine learning approach","authors":"E. V. Kazantseva, A. Ivannikov, A. Tarzimanova, V. Podzolkov","doi":"10.17816/dd626797","DOIUrl":"https://doi.org/10.17816/dd626797","url":null,"abstract":"BACKGROUND: Arterial hypertension and chronic obstructive pulmonary disease have a deleterious effect on the structure of the heart, leading to the development of atrial fibrillation, which remains the leading cause of cerebral stroke and premature death [1]. Consequently, the early identification of atrial fibrillation risk factors in patients with arterial hypertension and chronic obstructive pulmonary disease is of paramount importance for the prevention of such conditions. This is why predictive cardiology employs machine learning methods, which are demonstrably superior to classical statistical methods of prediction [2–4]. \u0000AIM: The study aimed to develop a prognostic model of atrial fibrillation in comorbid patients with arterial hypertension and chronic obstructive pulmonary disease based on multilayer perceptron. \u0000MATERIALS AND METHODS: The study included 419 patients treated at the University Clinical Hospital No. 4 of the I.M. Sechenov First Moscow State Medical University. Group 1 consisted of 91 (21.7%) patients with a verified diagnosis of atrial fibrillation, while Group 2 comprised 328 (78.3%) patients without atrial fibrillation. The random forest machine learning algorithm was used to identify predictors, which were then utilized to develop a neural network of the multilayer perceptron type. This consisted of two layers: an input layer of 12 neurons with the ReLU activation function and an output layer that receives input data from the previous layer and transmits them to one output with the sigmoid activation function. The threshold value, sensitivity, specificity, and diagnostic efficiency of the obtained model were determined using receiver operating characteristic analysis with the calculation of the area under the curve (AUC). \u0000RESULTS: By the first stage of prognostic model development, the most significant predictors of atrial fibrillation development were selected by the random forest machine learning algorithm. The model was developed using three variables: C-reactive protein concentration (odds ratio, OR 1.04; 95% confidence interval, CI 1.015–1.067; p=0.002), erythrocyte sedimentation rate (OR 1.04; 95% CI 1.019–1.069; p=0.002), and creatinine concentration (OR 1.03; 95% CI 1.011–1.042; p 0.001). These variables were used to train a multilayer perceptron model on a test sample for 500 epochs. \u0000Following training, the developed model exhibited a sensitivity of 85%, a specificity of 80%, and a diagnostic efficiency of 79.6%. AUC amounted to 0.900. \u0000CONCLUSIONS: The study resulted in the development of a prognostic model based on the application of machine learning methods, which exhibited favorable metrics. This model may be considered a valuable tool for clinical practice.","PeriodicalId":34831,"journal":{"name":"Digital Diagnostics","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-07-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141683244","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
E. D. Nikitin, Nikita S. Plaksin, Maria B. Garetz, Evgeniy M. Gutin
BACKGROUND: Up to 95% of digital fluoroscopy screening studies are free of pathologic changes. Radiologists typically spend the majority of their time reviewing and describing such studies. In these cases, artificial intelligence systems can be used to automate the description, thereby saving physicians’ time [1–3]. AIM: The aim of this study was to compare the efficacy of various algorithms within an existing artificial intelligence system in an ultra-high sensitivity scenario and to estimate the percentage of X-rays that could be automatically characterized. MATERIALS AND METHODS: The artificial intelligence system “Cels.Fluorography” version 0.15.3 was used for the analysis. A dataset derived from disparate medical organizations, comprising 11,707 studies devoid of pathology and 5,846 studies exhibiting pathology, was selected for comparison. A subsample of 500 studies with pathology and 9,500 studies without pathology (5% to 95% balance) was randomly selected 1,000 times from the dataset to calculate the metrics. The resulting metrics were then averaged. The markup of two physicians was used as the source of the target variable. In the event of a discrepancy in opinion, the study was subjected to an expert physician evaluation. An X-ray was considered pathological if the final markup contained at least one of 12 radiological features [4]. Five methods were used to compare metrics: by maximum (1) and mean (2) probability of radiological features localized by the neural network-detector; by maximum (3) and mean (4) probability of feature presence derived from dedicated “heads” of the neural network trained to determine the presence of each feature on the image (0 for no feature, 1 for presence); by probability (5) derived from a separate “head” of the neural network trained to determine the binary presence of pathology on the study (0 for normal, 1 for pathology). For each method, a response threshold was selected to ensure that no more than one missed pathology was identified per 1,000 examinations in the current subsample. The percentage of X-rays that could be correctly identified as pathology-free by artificial intelligence was calculated as the main quality metric. RESULTS: The methods demonstrated the following average percentages of norm dropout: 66.4%, 72.2%, 69.0%, 74.1%, 68.7%—and the following area under the ROC curve: 0.948, 0.957, 0.964, 0.967, 0.971. The 95% confidence interval for the dropout rate associated with the optimal method was found to be 66.1% to 79.4%. CONCLUSIONS: Modern artificial intelligence systems can be used to automate the description of a significant portion of screenings. The most efficacious method for norm screening (over 74% of the flow) was demonstrated by the averaging of probabilities derived from special “heads” of the neural network trained to identify the presence of pathology.
{"title":"Comparison of the methods of operation of the artificial intelligence system in the ultra-high sensitivity mode for the autonomous description of chest X-rays without pathology","authors":"E. D. Nikitin, Nikita S. Plaksin, Maria B. Garetz, Evgeniy M. Gutin","doi":"10.17816/dd626001","DOIUrl":"https://doi.org/10.17816/dd626001","url":null,"abstract":"BACKGROUND: Up to 95% of digital fluoroscopy screening studies are free of pathologic changes. Radiologists typically spend the majority of their time reviewing and describing such studies. In these cases, artificial intelligence systems can be used to automate the description, thereby saving physicians’ time [1–3]. \u0000AIM: The aim of this study was to compare the efficacy of various algorithms within an existing artificial intelligence system in an ultra-high sensitivity scenario and to estimate the percentage of X-rays that could be automatically characterized. \u0000MATERIALS AND METHODS: The artificial intelligence system “Cels.Fluorography” version 0.15.3 was used for the analysis. A dataset derived from disparate medical organizations, comprising 11,707 studies devoid of pathology and 5,846 studies exhibiting pathology, was selected for comparison. A subsample of 500 studies with pathology and 9,500 studies without pathology (5% to 95% balance) was randomly selected 1,000 times from the dataset to calculate the metrics. The resulting metrics were then averaged. \u0000The markup of two physicians was used as the source of the target variable. In the event of a discrepancy in opinion, the study was subjected to an expert physician evaluation. An X-ray was considered pathological if the final markup contained at least one of 12 radiological features [4]. \u0000Five methods were used to compare metrics: by maximum (1) and mean (2) probability of radiological features localized by the neural network-detector; by maximum (3) and mean (4) probability of feature presence derived from dedicated “heads” of the neural network trained to determine the presence of each feature on the image (0 for no feature, 1 for presence); by probability (5) derived from a separate “head” of the neural network trained to determine the binary presence of pathology on the study (0 for normal, 1 for pathology). \u0000For each method, a response threshold was selected to ensure that no more than one missed pathology was identified per 1,000 examinations in the current subsample. The percentage of X-rays that could be correctly identified as pathology-free by artificial intelligence was calculated as the main quality metric. \u0000RESULTS: The methods demonstrated the following average percentages of norm dropout: 66.4%, 72.2%, 69.0%, 74.1%, 68.7%—and the following area under the ROC curve: 0.948, 0.957, 0.964, 0.967, 0.971. The 95% confidence interval for the dropout rate associated with the optimal method was found to be 66.1% to 79.4%. \u0000CONCLUSIONS: Modern artificial intelligence systems can be used to automate the description of a significant portion of screenings. The most efficacious method for norm screening (over 74% of the flow) was demonstrated by the averaging of probabilities derived from special “heads” of the neural network trained to identify the presence of pathology.","PeriodicalId":34831,"journal":{"name":"Digital Diagnostics","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-07-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141683681","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Fedor A. Laputin, Ivan V. Sidorov, Andrey S. Moshkin
BACKGROUND: Ovarian reserve reflects a woman's ability to successfully realize reproductive function. The assessment of ovarian reserve is an urgent task for clinical practice [1] and is important in scientific research. The use of computerized diagnostic image processing methods can accelerate and facilitate the performance of routine tasks in clinical practice. Their use in retrospective data analysis for scientific purposes allows to increase the objectivity of the study and supplement it with auxiliary information [2]. The issue of ovarian localization and follicle segmentation on ultrasound images has been previously investigated in other works. For instance, Z. Chen et al. [3] employed the U-net model to identify follicles on ultrasound images. Similarly, V.K. Singh et al. [4] addressed a related problem using a variant of U-net, namely UNet++ [5], which has gained considerable traction in the field of medical image analysis [6]. AIM: The study aimed to develop machine learning models for analyzing ovarian images obtained from an ultrasound machine. MATERIALS AND METHODS: An open dataset with a labeled ovary region was used for pre-training ovarian segmentation and follicle detection models. Subsequently, the dataset, which contains marked-up ovarian and follicle regions, was employed for training and testing. It encompasses a total of approximately 800 examples from 50 unique patients. The localization of follicles in an ultrasound image is a challenging task. To address this, the designed detector system was divided into two parts: ovary segmentation and follicle detection within the selected region. This approach allows the model to focus on a region where there are no other organs and various ultrasound artifacts that can be falsely perceived as the object under investigation. For the purpose of ovarian segmentation, the UNet++ architecture [5] was employed in conjunction with the ResNeSt encoder [8], which incorporates the SE-Net [9] and SK-Net [10] attention mechanisms. The object detection model is employed to identify the location of follicles within the ovary, as it enables precise enumeration of the number of follicles, even in the presence of overlapping structures, a capability that the segmentation model lacks. In our study, we used the YOLOv8 model [11]. Furthermore, data preprocessing has been employed to enhance the quality of model predictions. This has involved the identification and removal of regions with auxiliary information, the reduction of noise, and the augmentation of data. RESULTS: Two ovarian localization models are presented based on the results of this study. The first model is a segmentation model with an IoU quality of at least 50%. The second model is a detection model with a mAP quality of at least 65%. A third model is a model for follicle detection with subsequent follicle counting. This model has an MAPE error not exceeding 35%. CONCLUSIONS: The study resulted in the proposal of a method for app
{"title":"Assessment of ovarian follicular reserve according to ultrasound data based on machine learning methods","authors":"Fedor A. Laputin, Ivan V. Sidorov, Andrey S. Moshkin","doi":"10.17816/dd626171","DOIUrl":"https://doi.org/10.17816/dd626171","url":null,"abstract":"BACKGROUND: Ovarian reserve reflects a woman's ability to successfully realize reproductive function. The assessment of ovarian reserve is an urgent task for clinical practice [1] and is important in scientific research. The use of computerized diagnostic image processing methods can accelerate and facilitate the performance of routine tasks in clinical practice. Their use in retrospective data analysis for scientific purposes allows to increase the objectivity of the study and supplement it with auxiliary information [2]. \u0000The issue of ovarian localization and follicle segmentation on ultrasound images has been previously investigated in other works. For instance, Z. Chen et al. [3] employed the U-net model to identify follicles on ultrasound images. Similarly, V.K. Singh et al. [4] addressed a related problem using a variant of U-net, namely UNet++ [5], which has gained considerable traction in the field of medical image analysis [6]. \u0000AIM: The study aimed to develop machine learning models for analyzing ovarian images obtained from an ultrasound machine. \u0000MATERIALS AND METHODS: An open dataset with a labeled ovary region was used for pre-training ovarian segmentation and follicle detection models. Subsequently, the dataset, which contains marked-up ovarian and follicle regions, was employed for training and testing. It encompasses a total of approximately 800 examples from 50 unique patients. \u0000The localization of follicles in an ultrasound image is a challenging task. To address this, the designed detector system was divided into two parts: ovary segmentation and follicle detection within the selected region. This approach allows the model to focus on a region where there are no other organs and various ultrasound artifacts that can be falsely perceived as the object under investigation. For the purpose of ovarian segmentation, the UNet++ architecture [5] was employed in conjunction with the ResNeSt encoder [8], which incorporates the SE-Net [9] and SK-Net [10] attention mechanisms. \u0000The object detection model is employed to identify the location of follicles within the ovary, as it enables precise enumeration of the number of follicles, even in the presence of overlapping structures, a capability that the segmentation model lacks. In our study, we used the YOLOv8 model [11]. \u0000Furthermore, data preprocessing has been employed to enhance the quality of model predictions. This has involved the identification and removal of regions with auxiliary information, the reduction of noise, and the augmentation of data. \u0000RESULTS: Two ovarian localization models are presented based on the results of this study. The first model is a segmentation model with an IoU quality of at least 50%. The second model is a detection model with a mAP quality of at least 65%. A third model is a model for follicle detection with subsequent follicle counting. This model has an MAPE error not exceeding 35%. \u0000CONCLUSIONS: The study resulted in the proposal of a method for app","PeriodicalId":34831,"journal":{"name":"Digital Diagnostics","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-07-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141681564","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：