Aleksandr A. Borisov has a total of 41 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Evaluation of an artificial intelligence model for quality assessment of chest radiograph in public healthcare settings: a pilot cross-sectional study

    by Aleksandr A. Borisov, Kirill M. Arzamasov, Olga V. Omelyanskaya, Anton V. Vladzymyrskyy, Yuriy A. Vasilev

    BACKGROUND: Suboptimal quality of chest radiographs is a common problem in the diagnostic imaging of thoracic diseases. Quality assessment of routine chest radiographs is performed manually, which is complicated by remote reporting and the high workload of radiologists. We previously developed an artificial intelligence-powered tool for quality control of these images. However, its performance in real-world clinical practice has not been investigated. AIM: To evaluate the effectiveness of the automated artificial intelligence tool for detecting defects in chest radiographs in routine clinical practice. METHODS: The tool was validated using data from three Moscow outpatient hospitals over a two-month period. Chest radiographs were assessed using an automated quality assessment tool based on ensemble learning that sequentially processes chest radiographs in DICOM (Digital Imaging and Communications in Medicine) format. The tool is integrated into the routine chest radiography workflow at the Moscow Reference Center for Radiology. Following automated assessment, image quality was analyzed by radiologists (n = 10). Based on the automated assessment, a report was generated that included information on incomplete visualization of the lungs and costodiaphragmatic recess in frontal and lateral view, chest rotation, incorrect image orientation, and incorrect information (metadata) about the anatomical region, view, and photometric interpretation of the study. Radiologists assessed radiographic image defects using similar criteria. Each image was evaluated by a single radiologist. The radiologists were blinded to the automated image evaluation results. RESULTS: A total of 9642 radiographs were processed. The mean processing time per image was 14 seconds. In total, 3386 distinct patient positioning errors were identified, with each error detected in 14%–45% of images. The accuracy of metadata completion for radiographs ranged from 14% to 100%. The accuracy, expressed as the ROC AUC, for detecting patient positioning errors using the quality control tool ranged from 0.782 to 0.947, while the accuracy for detecting metadata completion errors ranged from 0.985 to 1.0. CONCLUSION: Automated quality control of radiographs is an accurate and rapid method for detecting patient positioning errors and errors in metadata completion.

  • Open Access

    Key aspects of fine-tuning and applying LLM-as-a-judge for clinical data summaries in the radiological workflow

    by Aleksandr Borisov, Tikhon Burtsev, Pavel Kosov, Tatiana Bobrovskaya, Yuri Vasilev, Anton Vladzymyrskyy, Olga Omelyanskaya, Anastasia Pamova, Kirill Arzamasov

    Background: This study aims to describe our experience in fine-tuning an LLM-as-a-Judge to evaluate the quality of clinical text summarization in the field of radiology and to formalize the main problems we encountered in solving this task. Methods: In this study, information from the Russian language electronic medical records of 30 patients was used. Patients who underwent abdominal computed tomography were selected. Anonymized information about complaints, disease history, medical history, and laboratory and instrumental findings was obtained from the electronic medical records of patients. This information was summarized by six large language models. The resulting summarizations were then evaluated by experts and six different LLMs-as-a-Judges. Kendall’s coefficient of concordance was employed to measure consistency. Results: The primary difficulties that we encountered in the development of LLM-as-a-Judge included the selection of the rating scale, evaluation criteria, various categories of members included in the expert team, and prompt granularity. No definitive association was identified between scale size and the consistency of ratings between radiologist experts and LLMs-as-a-Judges. Across different evaluation criteria, the highest level of consistency was achieved with varying scale sizes. Our results indicate that criteria effective for human text evaluation are not always suitable for assessment via an LLM-as-a-Judge. For the majority of the criteria, the highest consistency was observed when all LLMs-as-a-Judges operated with a detailed description of extreme scale values or without a detailed scale description in the prompt. For the effective development of an LLM judge, it is necessary to involve a diverse team of experts. Conclusion: For the proper configuration of an LLM-as-a-Judge, numerous factors should be considered, the number of which varies depending on the specific task. To achieve optimal results, additional experiments should be conducted to fine-tune the prompt and other model hyperparameters, comparing their responses against the desired output. Clinical trial registration: ClinicalTrials.gov , identifier NCT07057830.


  • Open Access

    Application of radiomics in the diagnosis of cervical cancer: systematic review

    by V. A. Solodky, N. V. Nudnov, P. N. Sultanova, S. P. Aksenova, A. A. Borisov, E. S.-A. Shakhvalieva, D. G. Karelidze, M. E. Ivannikov, A. I. Makovetskaya, S. R. Semenova

    Objective: To analyze the results of a study on the effectiveness of radiomic analysis in the interpretation of radiation images in clarifying the diagnosis of cervical cancer. Materials and Methods. A systematic literature search was conducted in the PubMed/MEDLINE, eLibrary, and Scopus databases, as well as in NCCN, ESUR, and ACR resources. Results. When selecting medical articles, a total of 289 unique publications were identified, 218 of which met the exclusion criteria. The final review included 71 articles. The average accuracy characteristics of the models were estimated based on the area under the ROC curve (AUC), including accuracy, sensitivity, specificity, and C-index. Conclusion. The main key aspects and advantages of the use of radiomics and textural image analysis in the diagnosis of cervical cancer are considered. The introduction of radiomic analysis has led to a renewed perception of medical image analysis. The results of a number of studies demonstrate that the data extracted using radiomic analysis have significant diagnostic and prognostic value, allowing an objective assessment of tumor characteristics, its stage and prevalence, and differential diagnosis of neoplasms.

  • Quality control system for patient positioning and filling in meta-information for chest X-ray examinations

    by A. A. Borisov, S. S. Semenov, Yu. S. Kirpichev, K. M. Arzamasov, O. V. Omelyanskaya, A. V. Vladzymyrskyy, Yu. A. Vasilev

    Purpose During radiography, irregularities occur, leading to decrease in the diagnostic value of the images obtained. The purpose of this work was to develop a system for automated quality assurance of patient positioning in chest radiographs, with detection of suboptimal contrast, brightness, and metadata errors. Methods The quality assurance system was trained and tested using more than 69,000 X-rays of the chest and other anatomical areas from the Unified Radiological Information Service (URIS) and several open datasets. Our dataset included studies regardless of a patient's gender and race, while the sole exclusion criterion being age below 18 years. A training dataset of radiographs labeled by expert radiologists was used to train an ensemble of modified deep convolutional neural networks architectures ResNet152V2 and VGG19 to identify various quality deficiencies. Model performance was accessed using area under the receiver operating characteristic curve (ROC-AUC), precision, recall, F1-score, and accuracy metrics. Results Seven neural network models were trained to classify radiographs by the following quality deficiencies: failure to capture the target anatomic region, chest rotation, suboptimal brightness, incorrect anatomical area, projection errors, and improper photometric interpretation. All metrics for each model exceed 95%, indicating high predictive value. All models were combined into a unified system for evaluating radiograph quality. The processing time per image is approximately 3 s. Conclusion The system supports multiple use cases: integration into an automated radiographic workstations, external quality assurance system for radiology departments, acquisition quality audits for municipal health systems, and routing of studies to diagnostic AI models.


  • Open Access

    Determination of the histological type of lung cancer based on radiomic analysis of computed tomography chest images

    by V. A. Solodkiy, N. V. Nudnov, D. G. Karelidze, A. A. Borisov, P. N. Sultanova, M. E. Ivannikov, E. S.-A. Shakhvalieva

    Lung cancer is one of the most common cancers. Bronchoscopy and transthoracic lung biopsy under the control of computed tomography (CT) are used for morphological verification of the tumor. Both of these technologies are invasive with certain risks and high costs. The accuracy of the morphologically verified diagnosis of lung cancer in Russia reaches an average of 88.2%. Treatment tactics, progression and prognosis of the disease depends on the histological type of lung cancer. The gold standard for lung cancer diagnosis is computed tomography of the chest. A developing area of CT image processing is radiomics, a mathematical analysis of data from radiation research methods that allows the detection of tissue texture features at a level inaccessible to the eye of a radiologist. The use of radiomics methods can contribute to the determination of the histotype of lung cancer even at the stage of diagnostic search. Objective: to determine the most common histological types of lung cancer based on the textural analysis of CT-scans of the chest organs. Materials and methods. The study included data from 200 patients treated at the RSCRR with histologically confirmed lung cancer, including 100 patients with small-cell lung cancer, 100 patients with non-small cell lung cancer (50 of them with adenocarcinoma and 50 with squamous cell carcinoma). 107 radiomic features were calculated for each tumor. Machine learning models were built in the Python 3.10 programming language using specialized libraries. To select the most effective models, standard machine learning metrics were used: precision, recall, accuracy, f1-measure and the area under the receiver operating characteristic curve (ROC-AUC). Results. Several machine learning models were developed, the best metrics were gradient boosting for differentiating non-small cell lung cancer and small-cell lung cancer with ROC-AUC 0.973 and a random forest based on three trees for differentiating adenocarcinoma and squamous cell carcinoma with ROC-AUC 0.833. Conclusion. Classification models developed by us have high metrics of diagnostic accuracy, which allows us to discourse about the applicability of radiomics features for differentiating various types of lung cancer at the stage of diagnostic search, as well as in situations where it is impossible to obtain material for histological examination.

  • Open Access

    Application of Radiomics for Differentiating Lung Neuroendocrine Neoplasms

    by Aleksandr Borisov, David Karelidze, Mikhail Ivannikov, Elina Shakhvalieva, Peri Sultanova, Kirill Arzamasov, Nikolai Nudnov, Yuriy Vasilev

    Background/Objectives: Lung neuroendocrine neoplasms (NENs) are a heterogeneous group of tumors requiring accurate differentiation from non-small cell lung cancer (NSCLC) for effective treatment. Conventional computed tomography (CT) lacks pathognomonic features to distinguish these subtypes. Radiomics, which extracts quantitative imaging features, offers a potential solution. Methods: This retrospective multicenter study included 301 patients with histologically confirmed lung cancer who underwent native CT scans. The dataset comprised 150 NSCLC cases (75 adenocarcinomas, 75 squamous cell carcinomas) and 151 NENs (75 SCLC, 60 carcinoids, 16 large cell neuroendocrine carcinomas). Tumors were manually segmented, and 107 radiomics features were extracted. Dimensionality reduction and feature selection were performed using Pearson correlation analysis and LASSO regression. Decision tree and random forest classifiers were trained and evaluated using a 70:30 training–testing split. Model performance was assessed using the area under the receiver operating characteristic curve (AUC), accuracy, precision, recall, and F1-score. Results: The model differentiating NENs from NSCLC achieved an AUC of 0.988 on the test set, with an accuracy of 97.8%. The model distinguishing SCLC from other NENs attained an AUC of 0.860 and an accuracy of 82.6%. First-order and textural radiomics features were key discriminators. Conclusions: Radiomics-based machine learning models demonstrated high diagnostic accuracy in differentiating lung NENs from NSCLC and in subclassifying NENs. These findings highlight the potential of radiomics as a non-invasive, quantitative tool for lung cancer diagnosis, warranting further validation in larger multicenter studies.


  • Open Access

    Use of radiomics and dosiomics to identify predictors of radiation induced lung injury

    by Nikolay V. Nudnov, Vladimir M. Sotnikov, Mikhail E. Ivannikov, Elina S.-A. Shakhvalieva, Aleksandr A. Borisov, Vasiliy V. Ledenev, Aleksei Yu. Smyslov, Alina V. Anan'Ina

    BACKGROUND: Radiomics is a machine learning based technology that extracts, analyzes, and interprets quantitative features from digital medical images. In recent years, dosiomics has become an increasingly common term in the literature to describe a new radiomics method. Dosiomics is a texture analysis method for evaluating radiotherapy dose distribution patterns. Most of the published research in dosiomics evaluates its use in predicting radiation induced lung injury. AIM: The aim of the study was to identify predictors (biomarkers) of radiation induced lung injury using texture analysis of computed tomography (CT) images of lungs and chest soft tissues using radiomics and dosiomics. MATERIALS AND METHODS: The study used data from 36 women with breast cancer who received postoperative conformal radiation therapy. Retrospectively, the patients were divided into two groups according to the severity of post radiation lung lesions. 3D Slicer was used to evaluate CT results of all patients obtained during radiation treatment planning and radiation dose distribution patterns. The software was able to unload radiomic and dosiomic features from regions of interest. The regions of interest included chest soft tissue and lung areas on the irradiated side where the dose burden exceeded 3 and 10 Gy. RESULTS: The first group included 13 patients with minimal radiation induced lung lesions, and the second group included 23 patients with post radiation pneumofibrosis. In the lung area on the side irradiated with more than 3 Gy, statistically significant differences between the patient groups were obtained for three radiomic features and one dosiomic feature. In the lung area on the side irradiated with more than 10 Gy, statistically significant differences were obtained for 12 radiomic features and 1 dosiomic feature. In the area of chest soft tissues on the irradiated side, significant differences were obtained for 18 radiomic features and 4 dosiomic features. CONCLUSIONS: As a result, a number of radiomic and dosiomic features were identified which were statistically different in patients with minimal lesions and pulmonary pneumofibrosis following radiation therapy for breast cancer. Based on texture analysis, predictors (biomarkers) were identified to predict post radiation lung injury and identify higher risk patients.


  • Open Access

    Neuroendocrine tumors of stomach and pancreas: diagnostic potential of radiomics, issues, and solutions

    by Nikolay V. Nudnov, Elina S-A. Shakhvalieva, David G. Karelidze, Aleksandr A. Borisov, Mikhail E. Ivannikov, Шахвалиева Элина Саид-Аминовна

    BACKGROUND: Radiomics is currently a promising and prospective tool for diagnosing and treating neuroendocrine neoplasms at various sites. This method is often used for differential diagnosis of gastrointestinal neuroendocrine tumors with other neoplasms at this site. AIM: The aim of the study was to evaluate the potential of radiomics for differential diagnosis of neuroendocrine tumors of stomach and pancreas. MATERIALS AND METHODS: The study included data of 12 patients with morphologically proven neoplasms of the stomach (6 with neuroendocrine tumors and 6 with adenocarcinomas) and data of 22 patients with morphologically proven neoplasms of the pancreas (11 with neuroendocrine tumors and 11 with adenocarcinomas). All patients underwent abdominal computed tomography (CT) with intravenous contrast enhancement prior to treatment at the Russian Scientific Center of Roentgenology and Radiology. Radiomics parameters were calculated for the area of gastric and pancreatic tumor manually segmented in the native phase of the CT scan. The results were processed and statistically analyzed using Microsoft Office Excel and R-Studio, a free, open-source software development environment for the R programming language. RESULTS: CT scan examples demonstrate typical and atypical visual signs of neuroendocrine tumors of stomach and pancreas, contrast enhancement characteristics, location and structure of neoplasms. Fifteen radiomics parameters were identified that were statistically significantly different between gastric neuroendocrine tumor and gastric adenocarcinoma. In pancreas, neuroendocrine tumors differed significantly from adenocarcinomas in 14 radiomics parameters. CONCLUSIONS: Neuroendocrine tumors of stomach and pancreas are rare neoplasms that are mostly asymptomatic and difficult to visualize due to their small size and contrast enhancement characteristics. Texture analysis may be a promising approach to differentiate gastrointestinal neuroendocrine tumors from other neoplasms at these sites, especially in the view of the difficulty in obtaining a biopsy.


  • Open Access

    Radiomics in the differential diagnosis of focal brain lesions: a retrospective study

    by N. V. Nudnov, E. V. Bit-Yunan, E. S.-A. Shakhvalieva, A. A. Borisov, P. N. Sultanova, M. E. Ivannikov, D. G. Karelidze, P. I. Bochkova

    INTRODUCTION: Glioblastoma and solitary metastases are the most common malignant neoplasms of the brain, characterized by high mortality and severe disability in patients. The method of choice for neuroimaging glioblastomas and metastases is contrast-enhanced magnetic resonance imaging. However, differentiation between the two is often difficult due to similar radiological features on MRI. Radiomics and machine learning can differentiate the primary origin of brain metastases and identify pathological tumor types noninvasively.OBJECTIVE: Application of texture analysis for differential diagnosis of glioblastomas and metastases of different etiologies.MATERIALS AND METHODS: 169 MRI studies from the RSCRR database were used in the study, 11 of which visualized morphologically differentiated glioblastoma of the brain, 55 lung cancer metastases and 103 breast cancer metastases. Segmentation of the regions of interest was performed semi-automatically in the free 3D-Slicer software with the ability to upload radiomic features from the regions of interest. For each lesion, 107 radiomic features were calculated from T1 and T2 sequences. Statistics: The calculation of statistical indicators was performed in a computer program for statistical data processing IBMSPSS Statistics 23. In statistical data processing, the Mann-Whitney statistical criterion for quantitative indicators and correlation analysis using the Pearson criterion were used to reduce the feature space. The reduction of the feature space and the selection of predictors by the feature_importances measure based on decision forests were carried out. Machine learning models were built in Python 3.10 using specialized libraries.RESULTS: For the model based on radiomic features extracted from T1 sequence, random forest showed the most efficient result, ROC-AUC=0.815 [0.749; 0.874]. For the model based on the radiomic features extracted from the T2 sequence, random forest showed the most effective result, ROC-AUC=0.817 [0.743; 0.873]. For the complex model based on radiomic features extracted from T1 and T2 sequences, random forest showed the most efficient result, ROC-AUC=0.855 [0.789; 0.906].DISCUSSION: The classification models and their metrics obtained by us indicate that the radiomic features extracted from T2 weighted MR images make it possible to differentiate breast cancer metastases from lung cancer metastases with higher sensitivity than the features extracted from T1 weighted MR images. We also identified a large number of significantly different indicators in the construction of models for the differentiation of glioblastomas from metastases, which demonstrates the prospects of this direction. It is planned to continue the study with the expansion of samples. Our conclusions are also confirmed by the research results of our foreign colleagues.CONCLUSION: The models we have obtained are highly accurate and sensitive to the differentiation of metastases of various etiologies and demonstrate significant potential in continuing this study with an expansion of samples.


  • Open Access

    Knowledge Base Prototype Creating with Using Interdisciplinary Metathesaurus

    by Pavel A. Astanin, Lev V. Ronzhin, Aleksandr A. Borisov, Anastasia V. Usova, Irina I. Potapova, Anna V. Budykina

    This article presents our experience in development an ontological model can be used in clinical decision support systems (CDSS) creating. We have used the largest international biomedical terminological metathesaurus the Unified Medical Language System (UMLS) as the basis of our model. This metathesaurus has been adapted into Russian using an automated hybrid translation system with expert control. The product we have created was named as the National Unified Terminological System (NUTS). We have added more than 33 million scientific and clinical relationships between NUTS terms, extracted from the texts of scientific articles and electronic health records. We have also computed weights for each relationship, standardized their values and created symptom checker in preliminary diagnostics based on this. We expect, that the NUTS allow solving task of named entity recognition (NER) and increasing terms interoperability in different CDSS.