BACKGROUND: The majority of current algorithms for blood flow surface extraction in the context of hemomodeling of abdominal aortic aneurysms are derived through a segmentation step, rather than directly from CT scans [1]. This approach introduces a degree of complexity, as the segmentation neural network is trained without consideration of the fact that the blood flow is a simply-connected region. Consequently, post-processing may be required to fulfill the simple connectivity criterion. In addition, the blood flow surface obtained from the segmentation mask using marching cubes is too coarse and requires smoothing. To provide one-stage surface extraction, Voxel2Mesh [2] was the first to be proposed. Voxel2Mesh shows good performance in extracting relatively simple geometries, while for more complex ones, its modifications have been proposed in the literature [3, 4]. AIM: The study aimed to develop an algorithm for single-stage extraction of the lumen surface of an abdominal aortic aneurysm. MATERIALS AND METHODS: A total of 90 contrast-enhanced CT images and segmentation masks with blood flow region labeling were prepared and divided into three groups: 40, 20, and 30 images for training, validation, and testing, respectively. Affine and non-linear augmentations were applied to increase the effective training sample size. A hybrid neural network consisting of a voxel encoder, a voxel decoder, and a grid decoder was proposed for single-stage surface extraction. The architectural design of the encoder is based on the Atto-sized ConvNeXtV2 architecture. The voxel decoder is comprised of five blocks, beginning with an interpolation layer and concluding with two super-precision words with packet normalization layers and ReLU. The voxel decoder and encoder are linked by means of analogous connections to those observed in the Unet architecture. The grid decoder comprises four GraphSAGE convolutions, with GeLU intervening between each pair. It is connected to the voxel decoder. The input to the encoder is a computed tomography image, while the input to the grid decoder is an initial approximation of the surface in the form of a ball. The output of the voxel decorrelation is a segmentation mask, while the output of the mesh decorrelation is the extracted surface. A combination of voxel and mesh loss functions was employed for the purposes of training. The surface generated from the segmentation mask by the Marching Cubes algorithm was employed as the reference surface. The mesh loss function was regularized to set the necessary parameters for the generated mesh. The quality of the generated mesh was evaluated using the Dice coefficient, which compares the true segmentation mask with the rasterized generated surface. RESULTS: We proposed the first hybrid neural network with an encoder based on the state-of-the-art ConvNeXtV2 architecture for the direct generation of abdominal aortic aneurysm blood flow meshes. A 14.01% improvement in generation was ach
{"title":"One shot lumen mesh generation of abdominal aortic aneurysm by hybrid neural network","authors":"R. Epifanov, R. Mullyadzhanov, Andrey A. Karpenko","doi":"10.17816/dd626155","DOIUrl":"https://doi.org/10.17816/dd626155","url":null,"abstract":"BACKGROUND: The majority of current algorithms for blood flow surface extraction in the context of hemomodeling of abdominal aortic aneurysms are derived through a segmentation step, rather than directly from CT scans [1]. This approach introduces a degree of complexity, as the segmentation neural network is trained without consideration of the fact that the blood flow is a simply-connected region. Consequently, post-processing may be required to fulfill the simple connectivity criterion. In addition, the blood flow surface obtained from the segmentation mask using marching cubes is too coarse and requires smoothing. To provide one-stage surface extraction, Voxel2Mesh [2] was the first to be proposed. Voxel2Mesh shows good performance in extracting relatively simple geometries, while for more complex ones, its modifications have been proposed in the literature [3, 4]. \u0000AIM: The study aimed to develop an algorithm for single-stage extraction of the lumen surface of an abdominal aortic aneurysm. \u0000MATERIALS AND METHODS: A total of 90 contrast-enhanced CT images and segmentation masks with blood flow region labeling were prepared and divided into three groups: 40, 20, and 30 images for training, validation, and testing, respectively. Affine and non-linear augmentations were applied to increase the effective training sample size. A hybrid neural network consisting of a voxel encoder, a voxel decoder, and a grid decoder was proposed for single-stage surface extraction. The architectural design of the encoder is based on the Atto-sized ConvNeXtV2 architecture. The voxel decoder is comprised of five blocks, beginning with an interpolation layer and concluding with two super-precision words with packet normalization layers and ReLU. The voxel decoder and encoder are linked by means of analogous connections to those observed in the Unet architecture. The grid decoder comprises four GraphSAGE convolutions, with GeLU intervening between each pair. It is connected to the voxel decoder. The input to the encoder is a computed tomography image, while the input to the grid decoder is an initial approximation of the surface in the form of a ball. The output of the voxel decorrelation is a segmentation mask, while the output of the mesh decorrelation is the extracted surface. A combination of voxel and mesh loss functions was employed for the purposes of training. The surface generated from the segmentation mask by the Marching Cubes algorithm was employed as the reference surface. The mesh loss function was regularized to set the necessary parameters for the generated mesh. The quality of the generated mesh was evaluated using the Dice coefficient, which compares the true segmentation mask with the rasterized generated surface. \u0000RESULTS: We proposed the first hybrid neural network with an encoder based on the state-of-the-art ConvNeXtV2 architecture for the direct generation of abdominal aortic aneurysm blood flow meshes. A 14.01% improvement in generation was ach","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":"141681697","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: Prostate cancer is one of the most common cancers among men [1, 2]. In recent years, a number of prognostic models based on texture analysis of biparametric magnetic resonance images have been created. The research has shown that radiomics features extracted from apparent diffusion coefficient maps are the most reproducible [3]. However, the models were limited in accuracy, since they are built using a single machine learning algorithm, which takes into account only linear dependences [4–6]. AIM: Increasing the accuracy of a prognostic model diagnosing prostate cancer through the use of stacking machine learning algorithms that takes into account not only linear, but also nonlinear dependencies based on radiomics of biparametric magnetic resonance imaging apparent diffusion coefficient maps. MATERIALS AND METHODS: A single-center cohort retrospective study of patients with suspected prostate cancer was conducted in the X-ray Diagnostics and Tomography Department of the United Hospital and Polyclinic (Moscow, Russia) from 2017 to 2023. The presence of prostate cancer was confirmed by biopsy or radical prostatectomy. Statistical analyses was performed using Python 3.11. RESULTS: The study involved 67 men aged 60 [54; 66] years, of which 57 were diagnosed with prostate cancer, and 10 — with benign prostate formation. The LIFEx software identified 96 radiomic features. Statistically significant differences were found for: PARAMS_ZSpatialResampling (the voxel size of the image: Z dimension) (p=0.001), SHAPE_Sphericity[onlyFor3DROI] (how spherical a Volume of Interest is) (p=0.006), SHAPE_Compacity[onlyFor3DROI] (how compact the Volume of Interest is) (p=0.004), GLRLM_HGRE (p=0.039), GLRLM_SRHGE (p=0.041), GLRLM_RLNU (p=0.039), where GLRLM — Grey-Level Run Length Matrix. Univariate logistic regression showed that SHAPE_Compacity[onlyFor3DROI] (R2=15%) and PARAMS_ZSpatialResampling (R2=18%) had a statistically significant effect on the outcome. First, using the multivariate logistic regression method, a prognostic model was built that takes into account only linear dependencies. The model includes 3 features that together have a statistically significant effect on the outcome (R2=23%): SHAPE_Sphericity[onlyFor3DROI], PARAMS_ZSpatialResampling and GLRLM_RLNU. To describe nonlinear relationships, another model was built based on the “Decision Tree” algorithm. It included 4 indicators (R2=58%): DISCRETIZED_HISTO_Entropy_log10 (the randomness of the distribution), SHAPE_Sphericity[onlyFor3DROI], PARAMS_ZSpatialResampling and GLRLM_SRE. Stacking of algorithms, which consists of calculating the arithmetic mean between the predictions of the multivariate logistic regression and “Decision Tree” algorithms, made it possible to construct a model that takes into account both linear and nonlinear dependencies. The model includes 5 features (R2=77%). The constructed model formed the basis of the developed calculator program [7], currently introduce
{"title":"Development of a prognostic model for diagnosis of prostate cancer based on radiomics of biparametric magnetic resonance imaging apparent diffusion coefficient maps and stacking of machine learning algorithms","authors":"A. I. Kuznetsov","doi":"10.17816/dd626145","DOIUrl":"https://doi.org/10.17816/dd626145","url":null,"abstract":"BACKGROUND: Prostate cancer is one of the most common cancers among men [1, 2]. In recent years, a number of prognostic models based on texture analysis of biparametric magnetic resonance images have been created. The research has shown that radiomics features extracted from apparent diffusion coefficient maps are the most reproducible [3]. However, the models were limited in accuracy, since they are built using a single machine learning algorithm, which takes into account only linear dependences [4–6]. \u0000AIM: Increasing the accuracy of a prognostic model diagnosing prostate cancer through the use of stacking machine learning algorithms that takes into account not only linear, but also nonlinear dependencies based on radiomics of biparametric magnetic resonance imaging apparent diffusion coefficient maps. \u0000MATERIALS AND METHODS: A single-center cohort retrospective study of patients with suspected prostate cancer was conducted in the X-ray Diagnostics and Tomography Department of the United Hospital and Polyclinic (Moscow, Russia) from 2017 to 2023. The presence of prostate cancer was confirmed by biopsy or radical prostatectomy. Statistical analyses was performed using Python 3.11. \u0000RESULTS: The study involved 67 men aged 60 [54; 66] years, of which 57 were diagnosed with prostate cancer, and 10 — with benign prostate formation. The LIFEx software identified 96 radiomic features. \u0000Statistically significant differences were found for: PARAMS_ZSpatialResampling (the voxel size of the image: Z dimension) (p=0.001), SHAPE_Sphericity[onlyFor3DROI] (how spherical a Volume of Interest is) (p=0.006), SHAPE_Compacity[onlyFor3DROI] (how compact the Volume of Interest is) (p=0.004), GLRLM_HGRE (p=0.039), GLRLM_SRHGE (p=0.041), GLRLM_RLNU (p=0.039), where GLRLM — Grey-Level Run Length Matrix. Univariate logistic regression showed that SHAPE_Compacity[onlyFor3DROI] (R2=15%) and PARAMS_ZSpatialResampling (R2=18%) had a statistically significant effect on the outcome. First, using the multivariate logistic regression method, a prognostic model was built that takes into account only linear dependencies. The model includes 3 features that together have a statistically significant effect on the outcome (R2=23%): SHAPE_Sphericity[onlyFor3DROI], PARAMS_ZSpatialResampling and GLRLM_RLNU. \u0000To describe nonlinear relationships, another model was built based on the “Decision Tree” algorithm. It included 4 indicators (R2=58%): DISCRETIZED_HISTO_Entropy_log10 (the randomness of the distribution), SHAPE_Sphericity[onlyFor3DROI], PARAMS_ZSpatialResampling and GLRLM_SRE. \u0000Stacking of algorithms, which consists of calculating the arithmetic mean between the predictions of the multivariate logistic regression and “Decision Tree” algorithms, made it possible to construct a model that takes into account both linear and nonlinear dependencies. The model includes 5 features (R2=77%). The constructed model formed the basis of the developed calculator program [7], currently introduce","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":"141683370","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: In recent years, the healthcare sector has emerged as a key area where artificial intelligence technologies are gaining strategic importance. In particular, the implementation of these technologies in primary healthcare has demonstrated particular relevance and importance [1–3]. AIM: The aim of the study is to characterize the stages of implementation of artificial intelligence technologies in the activities of urban polyclinics in Moscow. MATERIALS AND METHODS: Analytical, statistical, socio-hygienic, and experimental methods were used. RESULTS: The primary objective of integrating artificial intelligence into the operations of city polyclinics was to enhance the efficacy of medical data processing, mitigate the likelihood of professional missteps, and optimize the coordination of interactions between different medical professionals. The initial challenge of processing a vast quantity of information was met by the implementation of artificial intelligence in the analysis of electronic medical records. This approach resulted in the development of integrated and secure systems that facilitate the accessibility of patient data to physicians and medical staff for the purpose of quality of care analysis. In addressing the second task of using artificial intelligence technologies to provide consulting services to physicians in making a diagnosis, the work was carried out in several stages. In 2020, the top three medical decision support systems were implemented, which assist therapists in making preliminary diagnoses based on the International Classification of Diseases 10th revision (ICD-10). Since 2023, the Diagnostic Assistant system, which analyzes data from a patient’s electronic medical record and offers a second opinion on a confirmed diagnosis, has been actively used. Currently, this system includes 95 codes of ICD-10 and similar diagnoses, with plans to expand its functionality to 268 diagnoses. As a consequence of the training and implementation of the expansion, the system will be capable of covering approximately 85% of the most frequently established confirmed diagnoses. A considerable number of expert physicians were involved in the establishment and evaluation of the systems, with over 10,000 cases being handled. In December 2023, a pilot project was conducted at the City Polyclinic No. 64 (Moscow) with the involvement of almost 100 doctors of this medical institution to identify the possibility of improving the reliability of the model. According to its results, it was found that the diagnoses made by the doctor and the artificial intelligence system coincide by 89%. Despite the impressive achievements of technology, it is important to emphasize that the use of artificial intelligence is not intended to replace the doctor, but rather serves as a second opinion in the work of a specialist. CONCLUSIONS: The integration of artificial intelligence into the operations of Moscow’s polyclinics not only reduces the time re
{"title":"Artificial intelligence technologies in the activities of primary healthcare in Moscow","authors":"E. V. Blokhina, A. S. Bezymyannyy","doi":"10.17816/dd626790","DOIUrl":"https://doi.org/10.17816/dd626790","url":null,"abstract":"BACKGROUND: In recent years, the healthcare sector has emerged as a key area where artificial intelligence technologies are gaining strategic importance. In particular, the implementation of these technologies in primary healthcare has demonstrated particular relevance and importance [1–3]. \u0000AIM: The aim of the study is to characterize the stages of implementation of artificial intelligence technologies in the activities of urban polyclinics in Moscow. \u0000MATERIALS AND METHODS: Analytical, statistical, socio-hygienic, and experimental methods were used. \u0000RESULTS: The primary objective of integrating artificial intelligence into the operations of city polyclinics was to enhance the efficacy of medical data processing, mitigate the likelihood of professional missteps, and optimize the coordination of interactions between different medical professionals. \u0000The initial challenge of processing a vast quantity of information was met by the implementation of artificial intelligence in the analysis of electronic medical records. This approach resulted in the development of integrated and secure systems that facilitate the accessibility of patient data to physicians and medical staff for the purpose of quality of care analysis. \u0000In addressing the second task of using artificial intelligence technologies to provide consulting services to physicians in making a diagnosis, the work was carried out in several stages. In 2020, the top three medical decision support systems were implemented, which assist therapists in making preliminary diagnoses based on the International Classification of Diseases 10th revision (ICD-10). \u0000Since 2023, the Diagnostic Assistant system, which analyzes data from a patient’s electronic medical record and offers a second opinion on a confirmed diagnosis, has been actively used. Currently, this system includes 95 codes of ICD-10 and similar diagnoses, with plans to expand its functionality to 268 diagnoses. As a consequence of the training and implementation of the expansion, the system will be capable of covering approximately 85% of the most frequently established confirmed diagnoses. \u0000A considerable number of expert physicians were involved in the establishment and evaluation of the systems, with over 10,000 cases being handled. \u0000In December 2023, a pilot project was conducted at the City Polyclinic No. 64 (Moscow) with the involvement of almost 100 doctors of this medical institution to identify the possibility of improving the reliability of the model. According to its results, it was found that the diagnoses made by the doctor and the artificial intelligence system coincide by 89%. Despite the impressive achievements of technology, it is important to emphasize that the use of artificial intelligence is not intended to replace the doctor, but rather serves as a second opinion in the work of a specialist. \u0000CONCLUSIONS: The integration of artificial intelligence into the operations of Moscow’s polyclinics not only reduces the time re","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":"141681397","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}
Mariia Belsheva, Anastasia V. Guseva, Fedor A. Koleda, Polina V. Murlina, Larisa P. Safonova
BACKGROUND: Time-resolved spectrophotometry enables the contact probing of biological tissues at a depth of two millimeters to several centimeters, with a spatial resolution of one to five millimeters. This technique provides a quantitative assessment of optical parameters, concentrations of main chromophores, identification of tissue type and inclusions in the volume, which is relevant for intraoperative diagnostics [1–3]. The variability of optical properties during probe squeezing necessitates the implementation of force control of squeezing, which, like positioning, is used in robotic surgery and diagnostics [4–11]. A combined mechanical and spectrophotometric approach holds promise in this regard. However, further research is required concerning spectrophotometer setup, the development of test objects, and the determination of the possibilities of positioning-force-controlled spectrophotometry for the identification of tissues and inclusions. Development of approaches to active positional force control to study the functionality of spectrophotometry in identifying tissue structures. MATERIALS AND METHODS: An experimental bench was constructed based on a two-wavelength spectrophotometer with OxiplexTS frequency approach (ISS Inc., USA). This bench allows for the position control of the optical probe using a robotic mini-manipulator (U-Arm, China). Additionally, a software program was developed to record the pressing force of the fabricated probe in a customized nozzle for the manipulator. Finally, an algorithm was proposed for processing experimental data to estimate biomechanical, optical, and physiological parameters of the tissue. A single healthy subject participated in the experimental study. Measurements were conducted on the dorsal and ventral surfaces of the forearm and on the palmar surface of the hypotenar. RESULTS: The quantitative assessment of elastic properties of biological tissue can be achieved through the use of force-displacement data. The simultaneous registration of optical parameters, concentrations of hemoglobin fractions in a unit of the investigated volume, and tissue saturation in the dynamics of probe pressing allows for the estimation of microcirculatory blood flow, the revelation of the presence and type of large vessels. The standard silicone test objects used for spectrophotometer calibration do not align with the mechanical properties of biological tissues. Given the diminutive dimensions of the optical probe, this discrepancy introduces an additional degree of uncertainty in the quantitative assessment of tissue properties. CONCLUSIONS: The addition of active force control and automated positioning of the optical probe during spectrophotometry enhances its functional capabilities for identifying tissue structures and expands its applications in robotic pre-, intra- and post-operative diagnostics. For further studies on a larger number of tissues, tissue structures and mimicking tissue test objects, an impr
背景：时间分辨分光光度法可对生物组织进行深度为两毫米至几厘米的接触式探测，空间分辨率为一至五毫米。该技术可定量评估光学参数、主要发色团的浓度、组织类型的识别以及体积中的夹杂物，这与术中诊断息息相关 [1-3]。探针挤压过程中光学特性的变化要求对挤压进行力控制，这与定位一样，可用于机器人手术和诊断[4-11]。在这方面，机械和分光光度测量相结合的方法大有可为。不过，还需要进一步研究分光光度计的设置、测试对象的开发，以及确定定位-力控制分光光度计用于鉴定组织和内含物的可能性。开发主动定位力控制方法，以研究分光光度法在识别组织结构方面的功能。材料与方法：基于 OxiplexTS 频率方法的双波长分光光度计（美国 ISS 公司）建造了一个实验台。该实验台可使用微型机械手（U-Arm，中国）控制光学探针的位置。此外，还开发了一个软件程序，用于记录制造的探针在机械手定制喷嘴中的压力。最后，还提出了一种处理实验数据的算法，以估算组织的生物力学、光学和生理参数。一名健康受试者参与了实验研究。测量在前臂的背侧和腹侧表面以及下臂的掌侧表面进行。结果：利用力位移数据可对生物组织的弹性特性进行定量评估。在探针按压的动态过程中，同时记录光学参数、调查体积单位内的血红蛋白浓度和组织饱和度，可以估算微循环血流量，揭示大血管的存在和类型。用于分光光度计校准的标准硅胶测试对象与生物组织的机械特性不符。鉴于光学探头的尺寸较小，这种差异给组织特性的定量评估带来了额外的不确定性。结论：在分光光度测量过程中增加主动力控制和光学探针自动定位功能，可增强其识别组织结构的功能，扩大其在机器人术前、术中和术后诊断中的应用。为了进一步研究更多的组织、组织结构和模拟组织测试对象，需要对实验台进行改进：提高力传感器的灵敏度、定位过程中运动的平稳性和慎密性，例如用协作机器人取代微型机械手。软件部分的改进包括通过输入接口模块实现与 OxiplexTS 的同步，编写自动表面扫描程序。
{"title":"Position-force control in the identification of tissue structures using the spectrophotometric method","authors":"Mariia Belsheva, Anastasia V. Guseva, Fedor A. Koleda, Polina V. Murlina, Larisa P. Safonova","doi":"10.17816/dd626641","DOIUrl":"https://doi.org/10.17816/dd626641","url":null,"abstract":"BACKGROUND: Time-resolved spectrophotometry enables the contact probing of biological tissues at a depth of two millimeters to several centimeters, with a spatial resolution of one to five millimeters. This technique provides a quantitative assessment of optical parameters, concentrations of main chromophores, identification of tissue type and inclusions in the volume, which is relevant for intraoperative diagnostics [1–3]. The variability of optical properties during probe squeezing necessitates the implementation of force control of squeezing, which, like positioning, is used in robotic surgery and diagnostics [4–11]. A combined mechanical and spectrophotometric approach holds promise in this regard. However, further research is required concerning spectrophotometer setup, the development of test objects, and the determination of the possibilities of positioning-force-controlled spectrophotometry for the identification of tissues and inclusions. \u0000Development of approaches to active positional force control to study the functionality of spectrophotometry in identifying tissue structures. \u0000MATERIALS AND METHODS: An experimental bench was constructed based on a two-wavelength spectrophotometer with OxiplexTS frequency approach (ISS Inc., USA). This bench allows for the position control of the optical probe using a robotic mini-manipulator (U-Arm, China). Additionally, a software program was developed to record the pressing force of the fabricated probe in a customized nozzle for the manipulator. Finally, an algorithm was proposed for processing experimental data to estimate biomechanical, optical, and physiological parameters of the tissue. A single healthy subject participated in the experimental study. Measurements were conducted on the dorsal and ventral surfaces of the forearm and on the palmar surface of the hypotenar. \u0000RESULTS: The quantitative assessment of elastic properties of biological tissue can be achieved through the use of force-displacement data. The simultaneous registration of optical parameters, concentrations of hemoglobin fractions in a unit of the investigated volume, and tissue saturation in the dynamics of probe pressing allows for the estimation of microcirculatory blood flow, the revelation of the presence and type of large vessels. The standard silicone test objects used for spectrophotometer calibration do not align with the mechanical properties of biological tissues. Given the diminutive dimensions of the optical probe, this discrepancy introduces an additional degree of uncertainty in the quantitative assessment of tissue properties. \u0000CONCLUSIONS: The addition of active force control and automated positioning of the optical probe during spectrophotometry enhances its functional capabilities for identifying tissue structures and expands its applications in robotic pre-, intra- and post-operative diagnostics. For further studies on a larger number of tissues, tissue structures and mimicking tissue test objects, an impr","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":"141683990","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. V. Manaev, A. A. Trukhin, S. M. Zakharova, M. S. Sheremeta, E. A. Troshina
BACKGROUND: Thyroid nodules are a prevalent issue, with an estimated incidence of 19% to 35% based on ultrasound examination and 8% to 65% based on autopsy findings [1]. In some cases, Plummer’s disease is observed, and nodular masses may be observed in 10% to 35% of Graves’ disease cases, with iodine accumulation of a different nature [2, 3]. One of the principal treatments for Graves’ and Plummer’s diseases is radioiodine therapy, which serves to exclude the possibility of malignancy in nodules. Furthermore, the pharmacokinetics of iodine is investigated, which represents the most time-consuming and labor-intensive stage of preparation for radioiodine therapy. In clinical practice, ultrasound is performed in accordance with the TI-RADS system, followed (if necessary) by fine-needle aspiration puncture biopsy, stratified according to the Bethesda system. However, the interpretation of ultrasound examinations is inherently subjective, whereas the use of decision support systems can reduce the number of fine-needle aspiration puncture biopsies by 27% and the number of missed malignant neoplasms by 1.9%. Furthermore, the quantitative characterization of nodal ultrasound may enhance the investigation of the pharmacokinetics of I-131 [4, 5]. AIM: The study aimed to develop a method for quantitatively characterizing ultrasound images of thyroid nodular masses for predicting malignancy and I-131 accumulation by nodular masses. MATERIALS AND METHODS: The study included 125 nodules with pathomorphologic findings (65 benign, 60 malignant) and 25 benign nodules (established by cytologic examination) of patients who underwent radioiodotherapy as part of the Russian Science Foundation grant project No. 22-15-00135. Longitudinal and transverse projections of thyroid nodules were obtained using GE Voluson E8 (36% of all benign nodules and 27% of malignant nodules) and GE Logiq E (64% of benign and 73% of malignant nodules). A pharmacokinetics study was conducted on 25 nodes obtained on a GE Logiq V2 device. The accumulation index of I-131 was determined after 24 hours. A spatial adjacency matrix, gray level line length matrix, gray level zone size matrix, and histogram were employed to investigate features based on ultrasound images. RESULTS: The malignancy prediction model, developed on the basis of the most significant features and after KNN correlation analysis, exhibited a diagnostic accuracy value of 72±3%, a sensitivity of 73±5%, and a specificity of 73±5%. An investigation of I-131 pharmacokinetics revealed that the maximum histogram intensity gradient (r=–0.48, p=0.08) and intensity entropy (r=–0.51, p=0.06) exhibited the highest Spearman correlation coefficient modulus with I-131 accumulation after 24 hours. CONCLUSIONS: The present study demonstrates the feasibility of using quantitative characterization of ultrasound images of nodal masses as a tool to monitor nodules before radioiodotherapy. This is with a view to subsequent adjunctive fine-nee
{"title":"Artificial intelligence in ultrasound of thyroid nodules, prognosis of I-131 uptake","authors":"A. V. Manaev, A. A. Trukhin, S. M. Zakharova, M. S. Sheremeta, E. A. Troshina","doi":"10.17816/dd625986","DOIUrl":"https://doi.org/10.17816/dd625986","url":null,"abstract":"BACKGROUND: Thyroid nodules are a prevalent issue, with an estimated incidence of 19% to 35% based on ultrasound examination and 8% to 65% based on autopsy findings [1]. In some cases, Plummer’s disease is observed, and nodular masses may be observed in 10% to 35% of Graves’ disease cases, with iodine accumulation of a different nature [2, 3]. One of the principal treatments for Graves’ and Plummer’s diseases is radioiodine therapy, which serves to exclude the possibility of malignancy in nodules. Furthermore, the pharmacokinetics of iodine is investigated, which represents the most time-consuming and labor-intensive stage of preparation for radioiodine therapy. In clinical practice, ultrasound is performed in accordance with the TI-RADS system, followed (if necessary) by fine-needle aspiration puncture biopsy, stratified according to the Bethesda system. However, the interpretation of ultrasound examinations is inherently subjective, whereas the use of decision support systems can reduce the number of fine-needle aspiration puncture biopsies by 27% and the number of missed malignant neoplasms by 1.9%. Furthermore, the quantitative characterization of nodal ultrasound may enhance the investigation of the pharmacokinetics of I-131 [4, 5]. \u0000AIM: The study aimed to develop a method for quantitatively characterizing ultrasound images of thyroid nodular masses for predicting malignancy and I-131 accumulation by nodular masses. \u0000MATERIALS AND METHODS: The study included 125 nodules with pathomorphologic findings (65 benign, 60 malignant) and 25 benign nodules (established by cytologic examination) of patients who underwent radioiodotherapy as part of the Russian Science Foundation grant project No. 22-15-00135. Longitudinal and transverse projections of thyroid nodules were obtained using GE Voluson E8 (36% of all benign nodules and 27% of malignant nodules) and GE Logiq E (64% of benign and 73% of malignant nodules). A pharmacokinetics study was conducted on 25 nodes obtained on a GE Logiq V2 device. The accumulation index of I-131 was determined after 24 hours. A spatial adjacency matrix, gray level line length matrix, gray level zone size matrix, and histogram were employed to investigate features based on ultrasound images. \u0000RESULTS: The malignancy prediction model, developed on the basis of the most significant features and after KNN correlation analysis, exhibited a diagnostic accuracy value of 72±3%, a sensitivity of 73±5%, and a specificity of 73±5%. An investigation of I-131 pharmacokinetics revealed that the maximum histogram intensity gradient (r=–0.48, p=0.08) and intensity entropy (r=–0.51, p=0.06) exhibited the highest Spearman correlation coefficient modulus with I-131 accumulation after 24 hours. \u0000CONCLUSIONS: The present study demonstrates the feasibility of using quantitative characterization of ultrasound images of nodal masses as a tool to monitor nodules before radioiodotherapy. This is with a view to subsequent adjunctive fine-nee","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":"141683066","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}
Alexandra S. Tyan, Grigoriy G. Karmazanovskij, N. A. Karelskaya, Evgeniy V. Kondratyev, Alexander D. Kovalev
BACKGROUND: Prostate cancer is currently the second most commonly diagnosed cancer in men. The second edition of the Prostate Imaging Magnetic Resonance Imaging Data Assessment and Reporting System (PI-RADS) was released in 2019 to standardize the diagnostic process. Within this classification, the PI-RADS 3 category indicates an intermediate risk of clinically significant prostate cancer. There is currently no consensus in the literature regarding the optimal treatment for patients in this category. Some researchers advocate for biopsy as a means of further evaluation, while others propose a strategy of active surveillance for these patients. AIM: The aim of this study is to analyze and compare existing diagnostic models based on radiomics to differentiate and detect clinically significant prostate cancer in patients with a PI-RADS 3 category. MATERIALS AND METHODS: A comprehensive search of the PubMed, Scopus, and Web of Science databases was conducted using the following keywords: PI-RADS 3, radiomics, texture analysis, clinically significant prostate cancer, with additional emphasis on studies evaluated by Radiology Quality Score. The selected studies were required to meet the following criteria: (1) identification of PI-RADS 3 according to version 2.1 guidelines, (2) use of systemic biopsy as a control, (3) use of tools compatible with the IBSI standard for analyzing radiologic features, and (4) detailed description of methodology. Consequently, four meta-analyses and 12 original articles were selected. RESULTS: Radiomics-based diagnostic models have demonstrated considerable potential for enhancing the accuracy of detecting clinically significant prostate cancer in the PI-RADS 3 category using the PI-RADS V2.1 system. However, studies by A. Stanzione A. et al. and J. Bleker et al. have identified quality issues with such models, which constrains their clinical application based on low Radiology Quality Score values. In contrast, the works of T. Li et al. and Y. Hou et al. proposed innovative methods, including nomogram development and the application of machine learning, which demonstrated the potential of radiomics in improving diagnosis for this category. This indicates the potential for further development and application of radiomics in clinical practice. CONCLUSIONS: Although the models developed today cannot completely replace PI-RADS, the inclusion of radiomics can greatly enhance the efficiency of the diagnostic process by providing radiologists with quantitative and qualitative criteria that will enable the diagnosis of prostate cancer with greater confidence.
背景：前列腺癌是目前第二大最常诊断出的男性癌症。2019 年发布了第二版前列腺成像磁共振成像数据评估和报告系统（PI-RADS），以规范诊断过程。在这一分类中，PI-RADS 3 类表示临床意义重大的前列腺癌的中等风险。目前，文献中尚未就该类患者的最佳治疗方法达成共识。一些研究人员主张将活检作为进一步评估的一种手段，而另一些研究人员则建议对这些患者采取积极监测的策略。目的：本研究旨在分析和比较现有的基于放射组学的诊断模型，以区分和检测 PI-RADS 3 类患者中具有临床意义的前列腺癌。材料与方法：使用以下关键词对 PubMed、Scopus 和 Web of Science 数据库进行了全面搜索：PI-RADS 3、放射组学、纹理分析、有临床意义的前列腺癌，重点是通过放射学质量评分进行评估的研究。所选研究必须符合以下标准：(1) 根据 2.1 版指南确定 PI-RADS 3；(2) 使用全身活检作为对照；(3) 使用符合 IBSI 标准的工具分析放射学特征；(4) 详细描述研究方法。因此，共筛选出 4 篇荟萃分析和 12 篇原创文章。结果：基于放射组学的诊断模型已显示出相当大的潜力，可提高使用 PI-RADS V2.1 系统检测 PI-RADS 3 类别中具有临床意义的前列腺癌的准确性。然而，A. Stanzione A.等人和 J. Bleker 等人的研究发现了这些模型的质量问题，这限制了它们的临床应用，因为它们的放射质量评分值很低。相比之下，T. Li 等人和 Y. Hou 等人的研究提出了创新方法，包括提名图的开发和机器学习的应用，证明了放射组学在改善这类疾病诊断方面的潜力。这表明放射组学在临床实践中具有进一步发展和应用的潜力。结论：虽然目前开发的模型还不能完全取代 PI-RADS，但放射组学的加入可以为放射科医生提供定量和定性的标准，使他们在诊断前列腺癌时更有信心，从而大大提高诊断过程的效率。
{"title":"Radiomics for diagnosing clinically significant prostate cancer PI-RADS 3: what is already known and what to do next?","authors":"Alexandra S. Tyan, Grigoriy G. Karmazanovskij, N. A. Karelskaya, Evgeniy V. Kondratyev, Alexander D. Kovalev","doi":"10.17816/dd627093","DOIUrl":"https://doi.org/10.17816/dd627093","url":null,"abstract":"BACKGROUND: Prostate cancer is currently the second most commonly diagnosed cancer in men. The second edition of the Prostate Imaging Magnetic Resonance Imaging Data Assessment and Reporting System (PI-RADS) was released in 2019 to standardize the diagnostic process. Within this classification, the PI-RADS 3 category indicates an intermediate risk of clinically significant prostate cancer. There is currently no consensus in the literature regarding the optimal treatment for patients in this category. Some researchers advocate for biopsy as a means of further evaluation, while others propose a strategy of active surveillance for these patients. \u0000AIM: The aim of this study is to analyze and compare existing diagnostic models based on radiomics to differentiate and detect clinically significant prostate cancer in patients with a PI-RADS 3 category. \u0000MATERIALS AND METHODS: A comprehensive search of the PubMed, Scopus, and Web of Science databases was conducted using the following keywords: PI-RADS 3, radiomics, texture analysis, clinically significant prostate cancer, with additional emphasis on studies evaluated by Radiology Quality Score. The selected studies were required to meet the following criteria: (1) identification of PI-RADS 3 according to version 2.1 guidelines, (2) use of systemic biopsy as a control, (3) use of tools compatible with the IBSI standard for analyzing radiologic features, and (4) detailed description of methodology. Consequently, four meta-analyses and 12 original articles were selected. \u0000RESULTS: Radiomics-based diagnostic models have demonstrated considerable potential for enhancing the accuracy of detecting clinically significant prostate cancer in the PI-RADS 3 category using the PI-RADS V2.1 system. However, studies by A. Stanzione A. et al. and J. Bleker et al. have identified quality issues with such models, which constrains their clinical application based on low Radiology Quality Score values. In contrast, the works of T. Li et al. and Y. Hou et al. proposed innovative methods, including nomogram development and the application of machine learning, which demonstrated the potential of radiomics in improving diagnosis for this category. This indicates the potential for further development and application of radiomics in clinical practice. \u0000CONCLUSIONS: Although the models developed today cannot completely replace PI-RADS, the inclusion of radiomics can greatly enhance the efficiency of the diagnostic process by providing radiologists with quantitative and qualitative criteria that will enable the diagnosis of prostate cancer with greater confidence.","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":"141681319","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}
Alisher A. Ibragimov, Sofya A. Senotrusova, Arsenii A. Litvinov, Aleksandra A. Beliaeva, E. Ushakov, Yu. Markin
BACKGROUND: Breast cancer is one of the leading causes of cancer-related mortality in women [1]. Regular mass screening with mammography plays a critical role in the early detection of changes in breast tissue. However, the early stages of pathology often go undetected and are difficult to diagnose [2]. Despite the effectiveness of mammography in reducing breast cancer mortality, manual image analysis can be time consuming and labor intensive. Therefore, attempts to automate this process, for example using computer-aided diagnosis systems, are relevant [3]. In recent years, however, solutions based on neural networks have gained increasing interest, especially in biology and medicine [4-6]. Technological advances using artificial intelligence have already demonstrated their effectiveness in pathology detection [7, 8]. AIM: The study aimed to develop an automated solution to detect breast cancer on mammograms. MATERIALS AND METHODS: The solution is implemented as follows: a deep neural network-based tool has been developed to obtain the probability of malignancy from the input image. A combined dataset from public datasets such as MIAS, CBIS-DDSM, INbreast, CMMD, KAU-BCMD, and VinDr-Mammo [9–14] was used to train the model. RESULTS: The classification model, based on the EfficientNet-B3 architecture, achieved an area under the ROC curve of 0.95, a sensitivity of 0.88, and a specificity of 0.9 when tested on a sample from the combined dataset. The model’s high generalization ability, which is another advantage, was demonstrated by its ability to perform well on images from different datasets with varying data quality and acquisition regions. Furthermore, techniques such as image pre-cropping and augmentations during training were used to enhance the model's performance. CONCLUSIONS: The experimental results demonstrated that the model is capable of accurately detecting malignancies with a high degree of confidence. The obtained high-quality metrics offer a significant potential for implementing this method in automated diagnostics, for instance, as an additional opinion for medical specialists.
背景：乳腺癌是导致妇女癌症相关死亡的主要原因之一 [1]。定期进行乳腺 X 射线照相检查对早期发现乳腺组织的变化起着至关重要的作用。然而，病理的早期阶段往往未被发现，难以诊断[2]。尽管乳腺 X 射线照相术能有效降低乳腺癌死亡率，但人工图像分析耗时耗力。因此，尝试将这一过程自动化，例如使用计算机辅助诊断系统，具有重要意义[3]。然而，近年来，基于神经网络的解决方案越来越受到关注，尤其是在生物学和医学领域[4-6]。人工智能技术的进步已经证明了其在病理检测方面的有效性[7, 8]。目的：本研究旨在开发一种自动解决方案，用于检测乳房 X 光照片上的乳腺癌。材料与方法：该解决方案的实施过程如下：开发了一种基于深度神经网络的工具，用于从输入图像中获取恶性肿瘤的概率。该模型由 MIAS、CBIS-DDSM、INbreast、CMMD、KAU-BCMD 和 VinDr-Mammo [9-14] 等公共数据集组合而成。结果：基于 EfficientNet-B3 架构的分类模型在综合数据集样本上进行测试时，ROC 曲线下面积达到 0.95，灵敏度为 0.88，特异度为 0.9。该模型的另一个优势是具有很高的泛化能力，它能够在数据质量和采集区域各不相同的不同数据集的图像上表现出色。此外，在训练过程中还使用了图像预裁剪和增强等技术来提高模型的性能。结论实验结果表明，该模型能够以较高的置信度准确检测出恶性肿瘤。所获得的高质量指标为在自动诊断中应用该方法提供了巨大的潜力，例如，作为医学专家的补充意见。
{"title":"Classification of the presence of malignant lesions on mammogram using deep learning","authors":"Alisher A. Ibragimov, Sofya A. Senotrusova, Arsenii A. Litvinov, Aleksandra A. Beliaeva, E. Ushakov, Yu. Markin","doi":"10.17816/dd627019","DOIUrl":"https://doi.org/10.17816/dd627019","url":null,"abstract":"BACKGROUND: Breast cancer is one of the leading causes of cancer-related mortality in women [1]. Regular mass screening with mammography plays a critical role in the early detection of changes in breast tissue. However, the early stages of pathology often go undetected and are difficult to diagnose [2]. \u0000Despite the effectiveness of mammography in reducing breast cancer mortality, manual image analysis can be time consuming and labor intensive. Therefore, attempts to automate this process, for example using computer-aided diagnosis systems, are relevant [3]. In recent years, however, solutions based on neural networks have gained increasing interest, especially in biology and medicine [4-6]. Technological advances using artificial intelligence have already demonstrated their effectiveness in pathology detection [7, 8]. \u0000AIM: The study aimed to develop an automated solution to detect breast cancer on mammograms. \u0000MATERIALS AND METHODS: The solution is implemented as follows: a deep neural network-based tool has been developed to obtain the probability of malignancy from the input image. A combined dataset from public datasets such as MIAS, CBIS-DDSM, INbreast, CMMD, KAU-BCMD, and VinDr-Mammo [9–14] was used to train the model. \u0000RESULTS: The classification model, based on the EfficientNet-B3 architecture, achieved an area under the ROC curve of 0.95, a sensitivity of 0.88, and a specificity of 0.9 when tested on a sample from the combined dataset. The model’s high generalization ability, which is another advantage, was demonstrated by its ability to perform well on images from different datasets with varying data quality and acquisition regions. Furthermore, techniques such as image pre-cropping and augmentations during training were used to enhance the model's performance. \u0000CONCLUSIONS: The experimental results demonstrated that the model is capable of accurately detecting malignancies with a high degree of confidence. The obtained high-quality metrics offer a significant potential for implementing this method in automated diagnostics, for instance, as an additional opinion for medical specialists.","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":"141682848","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}
Pavel M. Ignatov, A. Oleynikov, Alexander V. Gus’kov, Alina L. Shlykova, Dmitrii A. Surov
BACKGROUND: Artificial intelligence software used in contemporary dentistry is capable of autonomously selecting prosthetic structures based on treatment conditions, establishing a diagnosis based on X-ray and intraoral jaw scanning data. A neural network in the field of machine learning is a mathematical model that employs the principles of a neural network found in living organisms. It is capable of processing input signals in accordance with weight coefficients, passing them through a specific number of layers, and forming the correct answer at the output. This answer corresponds to the neuron of the output layer with the highest value of the activation function. AIM: The aim of the study was to develop a neural network for clinical decision making in orthopedic treatment planning. MATERIALS AND METHODS: A neural network was constructed using the Processing programming environment and a C-like programming language. At the stage of network training, the number of hidden layers was determined, the training coefficient was selected, and the number of training epochs was determined. The network was trained using the backpropagation of error method, which involved calculating the root-mean-square error of the network, backpropagating the signal through the neural network, and adjusting the weighting coefficients in consideration of the learning coefficient. The input layer (vector) comprised clinical conditions [1, 2]: oral cavity condition, allergoanamnesis, and various manifestations of the clinical picture (index of destruction of tooth surfaces, vitality of teeth, etc.). The dimensionality of the output layer was dependent on the number of constructions used and amounted to 19 neurons (prostheses including burette, telescopic, cover, plate; microprostheses by type such as table-top, overlay, and inlay). The output layer consisted of removable and fixed prostheses, the selection of which was based on a pre-designed algorithm. This algorithm was based on the following clinical conditions: Condition and number of teeth retained Index of destruction of the occlusal surface of masticatory teeth Black’s classification of carious cavities Parafunctions, allergic history [3, 4]. RESULTS: A neural network algorithm was developed in which a physician was required to input clinical data following an oral examination. The neural network, which facilitates clinical decision-making assistance, performs mathematical calculations in each layer, multiplying the elements of the input vector (and subsequently, each layer) by weighting coefficients (obtained as a result of training the neural network), and adding a bias. In order to obtain the results in the area of the activation function calculation, the obtained result was conducted through the activation function (Sigmoid, ReLu), selecting the output neuron with the largest result and predicting the most appropriate design [5, 6]. CONCLUSIONS: Consequently, the developed neural network is capable
背景：当代牙科中使用的人工智能软件能够根据治疗条件自主选择修复结构，并根据 X 射线和口内颌骨扫描数据进行诊断。机器学习领域的神经网络是一种数学模型，它采用了生物体内神经网络的原理。它能够根据权重系数处理输入信号，通过特定的层数，并在输出端形成正确的答案。这个答案与输出层中激活函数值最大的神经元相对应。目的：本研究旨在开发一种用于骨科治疗计划临床决策的神经网络。材料与方法：使用 Processing 编程环境和类 C 编程语言构建了一个神经网络。在网络训练阶段，确定了隐藏层的数量，选择了训练系数，并确定了训练历元的数量。网络训练采用误差反向传播法，即计算网络的均方根误差，通过神经网络反向传播信号，并根据学习系数调整加权系数。输入层（向量）包括临床条件[1, 2]：口腔状况、过敏性鼻炎和各种临床表现（牙面破坏指数、牙齿活力等）。输出层的维度取决于所用结构的数量，共有 19 个神经元（修复体包括滴定管式、伸缩式、盖式、板式；微型修复体按类型分列，如台式、覆盖式和镶嵌式）。输出层包括活动和固定假体，根据预先设计的算法进行选择。该算法基于以下临床条件： 保留牙齿的状况和数量 咀嚼牙齿咬合面破坏指数 布莱克龋洞分类 副功能、过敏史[3, 4]。 结果：开发了一种神经网络算法，要求医生在口腔检查后输入临床数据。该神经网络可协助临床决策，在每一层进行数学计算，将输入向量（以及随后的每一层）的元素与加权系数（通过训练神经网络获得）相乘，并添加偏差。为了获得激活函数计算区域内的结果，通过激活函数（Sigmoid、ReLu）对获得的结果进行，选择结果最大的输出神经元，预测最合适的设计[5, 6]。结论：因此，考虑到不同假体的潜在用途，所开发的神经网络能够针对不同病例提出临床上合理的矫形治疗方案。
{"title":"A neural network for clinical decision support in orthopedic dentistry","authors":"Pavel M. Ignatov, A. Oleynikov, Alexander V. Gus’kov, Alina L. Shlykova, Dmitrii A. Surov","doi":"10.17816/dd627046","DOIUrl":"https://doi.org/10.17816/dd627046","url":null,"abstract":"BACKGROUND: Artificial intelligence software used in contemporary dentistry is capable of autonomously selecting prosthetic structures based on treatment conditions, establishing a diagnosis based on X-ray and intraoral jaw scanning data. A neural network in the field of machine learning is a mathematical model that employs the principles of a neural network found in living organisms. It is capable of processing input signals in accordance with weight coefficients, passing them through a specific number of layers, and forming the correct answer at the output. This answer corresponds to the neuron of the output layer with the highest value of the activation function. \u0000AIM: The aim of the study was to develop a neural network for clinical decision making in orthopedic treatment planning. \u0000MATERIALS AND METHODS: A neural network was constructed using the Processing programming environment and a C-like programming language. At the stage of network training, the number of hidden layers was determined, the training coefficient was selected, and the number of training epochs was determined. The network was trained using the backpropagation of error method, which involved calculating the root-mean-square error of the network, backpropagating the signal through the neural network, and adjusting the weighting coefficients in consideration of the learning coefficient. \u0000The input layer (vector) comprised clinical conditions [1, 2]: oral cavity condition, allergoanamnesis, and various manifestations of the clinical picture (index of destruction of tooth surfaces, vitality of teeth, etc.). The dimensionality of the output layer was dependent on the number of constructions used and amounted to 19 neurons (prostheses including burette, telescopic, cover, plate; microprostheses by type such as table-top, overlay, and inlay). \u0000The output layer consisted of removable and fixed prostheses, the selection of which was based on a pre-designed algorithm. This algorithm was based on the following clinical conditions: \u0000 \u0000Condition and number of teeth retained \u0000Index of destruction of the occlusal surface of masticatory teeth \u0000Black’s classification of carious cavities \u0000Parafunctions, allergic history [3, 4]. \u0000 \u0000RESULTS: A neural network algorithm was developed in which a physician was required to input clinical data following an oral examination. The neural network, which facilitates clinical decision-making assistance, performs mathematical calculations in each layer, multiplying the elements of the input vector (and subsequently, each layer) by weighting coefficients (obtained as a result of training the neural network), and adding a bias. In order to obtain the results in the area of the activation function calculation, the obtained result was conducted through the activation function (Sigmoid, ReLu), selecting the output neuron with the largest result and predicting the most appropriate design [5, 6]. \u0000CONCLUSIONS: Consequently, the developed neural network is capable","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":"141683993","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. A. Martirosyan, G. G. Karmazanovsky, Evgeniy V. Kondratyev, E. A. Sokolova
. BACKGROUND: A limited number of studies have been conducted in Russian and world literature on the differential diagnosis of gastrointestinal stromal tumors with other intra-abdominal tumors. The treatment of gastric non-epithelial tumors is dependent on the histologic type. The standard treatment for localized forms of gastrointestinal stromal tumors is surgery. For subepithelial masses up to 2 cm in size, in the absence of endoscopic signs of high risk, a strategy of active surveillance with mandatory endoscopic ultrasound examination and compliance with short-term intervals may be considered. Leiomyomas, benign masses, do not typically necessitate surgical intervention in the absence of clinical symptoms. Therefore, preoperative determination of the tumor type may help to avoid unwarranted surgical intervention. However, the ability of computed tomography to differentiate these tumor types is limited due to the similar radiological picture. Therefore, new scientific and clinical methods are needed. One of the possible techniques is texture analysis (radiomics). AIM: The study aims to investigate the potential of texture analysis (radiomics) in the diagnosis and differential diagnosis of gastrointestinal stromal tumors and gastric leiomyomas by analyzing the available world scientific literature. MATERIALS AND METHODS: A search was conducted in PubMed, Scopus, and Web of Science databases for published articles using the following keywords gastrointestinal stromal tumors, leiomyomas, and radiomics. The review included 4 meta-analyses and 16 original articles. RESULTS: Texture analysis represents a promising tool for quantifying the heterogeneity of masses on radiologic images, thereby enabling the extraction of additional data that cannot be assessed by imaging analysis. The potential applications of texture analysis for differential diagnosis of gastrointestinal stromal tumors with other gastrointestinal neoplasms, risk stratification, and prediction of outcome after surgical treatment, as well as assessment of the mutational status of tumors, were explored. A differential diagnosis of gastrointestinal stromal tumors should be made with other mesenchymal tumors of the stomach (schwannoma, leiomyoma), as well as with malignant tumors (adenocarcinoma, lymphoma), although the number of such publications is limited. Some published studies on texture analysis of gastrointestinal stromal tumors have demonstrated excellent reproducibility of the obtained models. CONCLUSIONS: The lack of standardization and differences in study methodology present significant challenges to the clinical application of radiomics. Texture analysis may offer a valuable tool for the initial evaluation of gastric tumors, reducing the time required for diagnosis and determining patient management before biopsy. This approach could help to prevent inappropriate treatment.
.背景：俄罗斯和世界文献中关于胃肠道间质瘤与其他腹腔内肿瘤鉴别诊断的研究数量有限。胃非上皮性肿瘤的治疗取决于组织学类型。局部胃肠道间质瘤的标准治疗方法是手术。对于大小不超过 2 厘米的上皮下肿块，如果没有内镜下的高危迹象，可以考虑采取积极的监测策略，强制进行内镜超声检查，并遵守短期间隔期的规定。作为良性肿块的子宫肌瘤，在没有临床症状的情况下，一般不需要进行手术治疗。因此，术前确定肿瘤类型有助于避免不必要的手术干预。然而，由于放射学表现相似，计算机断层扫描区分这些肿瘤类型的能力有限。因此，需要新的科学和临床方法。纹理分析（放射组学）就是可行的技术之一。目的：本研究旨在通过分析现有的世界科学文献，研究纹理分析（放射组学）在诊断和鉴别诊断胃肠道间质瘤和胃癌中的潜力。材料与方法：使用以下关键词在 PubMed、Scopus 和 Web of Science 数据库中搜索已发表的文章：胃肠道间质瘤、胃癌和放射组学。综述包括 4 项荟萃分析和 16 篇原创文章。结果：纹理分析是量化放射影像上肿块异质性的一种很有前途的工具，可提取影像分析无法评估的其他数据。我们探讨了纹理分析在胃肠道间质瘤与其他胃肠道肿瘤的鉴别诊断、风险分层、手术治疗后疗效预测以及肿瘤突变状态评估方面的潜在应用。胃肠道间质瘤应与胃部其他间质瘤（裂孔瘤、子宫肌瘤）以及恶性肿瘤（腺癌、淋巴瘤）进行鉴别诊断，但此类出版物数量有限。一些已发表的胃肠道间质瘤纹理分析研究表明，所获得的模型具有极佳的再现性。结论：缺乏标准化和研究方法的差异给放射组学的临床应用带来了巨大挑战。纹理分析可为胃肿瘤的初步评估提供有价值的工具，缩短诊断所需的时间，并在活检前确定患者的治疗方案。这种方法有助于防止不适当的治疗。
{"title":"Radiomics in the differential diagnosis of gastrointestinal stromal tumors and leiomyomas. A literature review","authors":"E. A. Martirosyan, G. G. Karmazanovsky, Evgeniy V. Kondratyev, E. A. Sokolova","doi":"10.17816/dd627088","DOIUrl":"https://doi.org/10.17816/dd627088","url":null,"abstract":". \u0000BACKGROUND: A limited number of studies have been conducted in Russian and world literature on the differential diagnosis of gastrointestinal stromal tumors with other intra-abdominal tumors. The treatment of gastric non-epithelial tumors is dependent on the histologic type. The standard treatment for localized forms of gastrointestinal stromal tumors is surgery. For subepithelial masses up to 2 cm in size, in the absence of endoscopic signs of high risk, a strategy of active surveillance with mandatory endoscopic ultrasound examination and compliance with short-term intervals may be considered. Leiomyomas, benign masses, do not typically necessitate surgical intervention in the absence of clinical symptoms. Therefore, preoperative determination of the tumor type may help to avoid unwarranted surgical intervention. However, the ability of computed tomography to differentiate these tumor types is limited due to the similar radiological picture. Therefore, new scientific and clinical methods are needed. One of the possible techniques is texture analysis (radiomics). \u0000AIM: The study aims to investigate the potential of texture analysis (radiomics) in the diagnosis and differential diagnosis of gastrointestinal stromal tumors and gastric leiomyomas by analyzing the available world scientific literature. \u0000MATERIALS AND METHODS: A search was conducted in PubMed, Scopus, and Web of Science databases for published articles using the following keywords gastrointestinal stromal tumors, leiomyomas, and radiomics. The review included 4 meta-analyses and 16 original articles. \u0000RESULTS: Texture analysis represents a promising tool for quantifying the heterogeneity of masses on radiologic images, thereby enabling the extraction of additional data that cannot be assessed by imaging analysis. The potential applications of texture analysis for differential diagnosis of gastrointestinal stromal tumors with other gastrointestinal neoplasms, risk stratification, and prediction of outcome after surgical treatment, as well as assessment of the mutational status of tumors, were explored. A differential diagnosis of gastrointestinal stromal tumors should be made with other mesenchymal tumors of the stomach (schwannoma, leiomyoma), as well as with malignant tumors (adenocarcinoma, lymphoma), although the number of such publications is limited. Some published studies on texture analysis of gastrointestinal stromal tumors have demonstrated excellent reproducibility of the obtained models. \u0000CONCLUSIONS: The lack of standardization and differences in study methodology present significant challenges to the clinical application of radiomics. Texture analysis may offer a valuable tool for the initial evaluation of gastric tumors, reducing the time required for diagnosis and determining patient management before biopsy. This approach could help to prevent inappropriate treatment.","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":"141683209","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.M. Barskova, A.D. Kuklev, Nikolay V. Polukarov, E. Achkasov
BACKGROUND: The process of acquiring visual data from microelectromechanical sensors currently requires significant time and effort on the part of the clinician. The use of artificial intelligence algorithms to approximate data could potentially reduce the time required and increase the amount of work performed. AIM: The aim of this study is to approximate the data generated by sensors located in the shoe insole of basketball athletes and to compare the change in movement parameters of athletes when using CAD/CAM insoles. MATERIALS AND METHODS: Prior to the commencement of the study, permission was obtained from the local ethical committee of Sechenov University (protocol No. 19–23). The main cohort consisted of 39 athletes, comprising 21 men (53%) and 18 women (47%). The mean age of the athletes was 22.4 ± 7.54 years. The athletes were divided into three equal comparison groups according to the type of insoles they were wearing. Throughout the study period, all athletes remained healthy and free from injuries. The assessment of movement in space was conducted using a three-test system. This involved the use of microelectromechanical system sensors with an artificial intelligence algorithm, which facilitated the construction of visually clear and well-interpreted median lines (data approximation). RESULTS: For objective assessment of jumping characteristics, angular changes, velocity movements in space, and a comparison of all parameters on days 0 and 21, we developed and used our own software system, which was based on mathematical algorithmization and transformation formulas on specific axes. All data were entered into a neural network to construct averaged values of the parameters of movement in space. This approach allows the doctor to evaluate the changes of each peak movement on three different axes. Furthermore, it is possible to summarize the athlete's movement parameters with the aid of artificial intelligence, thereby enabling the detection of changes in different axes on days 0 and 21. Insole model C-1 exhibited the following improvements: X-axis movement speed (+7.7%), Y-axis jump height (+17.3%), endurance (+3.1%), and a 1.43-fold enhancement in shock absorption. Insole model C-2 exhibited an 8.4% increase in X-axis travel speed, a 20.8% enhancement in Y-axis jump height, a 6.6% improvement in endurance, and a 1.48-fold enhancement in shock absorption. Insole model C-3 demonstrated an 13.5% surge in X-axis travel speed, a 22.4% surge in Y-axis jump height, a 9.5% surge in endurance, and a 1.53-fold enhancement in shock absorption. CONCLUSIONS: The approximation of the data (median lines using an artificial intelligence algorithm) allows for the straightforward interpretation and comparison of various parameters, as well as the drawing of conclusions regarding the efficacy of individual sports CAD/CAM insoles. Additionally, it enables the assessment of changes in endurance, speed of movement during prolonged and intensive movement
{"title":"Using artificial intelligence algorithms to approximate data from inertial measurement unit sensors and strain gauges in basketball players","authors":"E.M. Barskova, A.D. Kuklev, Nikolay V. Polukarov, E. Achkasov","doi":"10.17816/dd626858","DOIUrl":"https://doi.org/10.17816/dd626858","url":null,"abstract":"BACKGROUND: The process of acquiring visual data from microelectromechanical sensors currently requires significant time and effort on the part of the clinician. The use of artificial intelligence algorithms to approximate data could potentially reduce the time required and increase the amount of work performed. \u0000AIM: The aim of this study is to approximate the data generated by sensors located in the shoe insole of basketball athletes and to compare the change in movement parameters of athletes when using CAD/CAM insoles. \u0000MATERIALS AND METHODS: Prior to the commencement of the study, permission was obtained from the local ethical committee of Sechenov University (protocol No. 19–23). The main cohort consisted of 39 athletes, comprising 21 men (53%) and 18 women (47%). The mean age of the athletes was 22.4 ± 7.54 years. The athletes were divided into three equal comparison groups according to the type of insoles they were wearing. Throughout the study period, all athletes remained healthy and free from injuries. The assessment of movement in space was conducted using a three-test system. This involved the use of microelectromechanical system sensors with an artificial intelligence algorithm, which facilitated the construction of visually clear and well-interpreted median lines (data approximation). \u0000RESULTS: For objective assessment of jumping characteristics, angular changes, velocity movements in space, and a comparison of all parameters on days 0 and 21, we developed and used our own software system, which was based on mathematical algorithmization and transformation formulas on specific axes. All data were entered into a neural network to construct averaged values of the parameters of movement in space. This approach allows the doctor to evaluate the changes of each peak movement on three different axes. Furthermore, it is possible to summarize the athlete's movement parameters with the aid of artificial intelligence, thereby enabling the detection of changes in different axes on days 0 and 21. Insole model C-1 exhibited the following improvements: X-axis movement speed (+7.7%), Y-axis jump height (+17.3%), endurance (+3.1%), and a 1.43-fold enhancement in shock absorption. Insole model C-2 exhibited an 8.4% increase in X-axis travel speed, a 20.8% enhancement in Y-axis jump height, a 6.6% improvement in endurance, and a 1.48-fold enhancement in shock absorption. Insole model C-3 demonstrated an 13.5% surge in X-axis travel speed, a 22.4% surge in Y-axis jump height, a 9.5% surge in endurance, and a 1.53-fold enhancement in shock absorption. \u0000CONCLUSIONS: The approximation of the data (median lines using an artificial intelligence algorithm) allows for the straightforward interpretation and comparison of various parameters, as well as the drawing of conclusions regarding the efficacy of individual sports CAD/CAM insoles. Additionally, it enables the assessment of changes in endurance, speed of movement during prolonged and intensive movement","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":"141681342","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：