Background/Objectives. Intracranial hemorrhages (ICHs) require immediate diagnosis for optimal clinical outcomes. Artificial intelligence (AI) is considered a potential solution for optimizing neuroimaging under conditions of radiologist shortage and increasing workload. This study aimed to directly compare diagnostic effectiveness between standalone AI services and AI-assisted radiologists in detecting ICHs on brain CT. Methods. A prospective, multicenter comparative study was conducted in 67 medical organizations in Moscow over 15+ months (April 2022–December 2024). We analyzed 3409 brain CT studies containing 1101 ICH cases (32.3%). Three commercial AI services with state registration were compared with radiologist conclusions formulated with access to AI results as auxiliary tools. Statistical analysis included McNemar’s test for paired data and Cohen’s h effect size analysis. Results. Radiologists with AI assistance statistically significantly outperformed AI services across all diagnostic metrics (p < 0.001): sensitivity 98.91% vs. 95.91%, specificity 99.83% vs. 87.35%, and accuracy 99.53% vs. 90.11%. The radiologists’ diagnostic odds ratio exceeded that of AI by 323-fold. The critical difference was in false-positive rates: 293 cases for AI vs. 4 for radiologists (73-fold increase). Complete complementarity of ICH misses was observed: all 12 cases undetected by radiologists were identified by AI, while all 45 cases missed by AI were diagnosed by radiologists. Agreement between methods was 89.6% (Cohen’s kappa 0.776). Conclusions. Radiologists maintain their role as the gold standard in ICH diagnosis, significantly outperforming AI services. Error complementarity indicates potential for improvement through systematic integration of AI as a “second reader” rather than a primary diagnostic tool. However, the high false-positive rate of standalone AI requires substantial algorithm refinement. The optimal implementation strategy involves using AI as an auxiliary tool within radiologist workflows rather than as an autonomous diagnostic system, with potential for delayed verification protocols to maximize diagnostic sensitivity while managing the false-positive burden.

BACKGROUND: Adrenal neoplasms are a common incidental finding on computed tomography, which remains the primary imaging method used to make a presumptive diagnosis of the lesion’s nosological type. Artificial intelligence-based software solutions for detecting adrenal neoplasms on computed tomography scans have been actively developed and implemented. AIM: This study aimed to assess the diagnostic effectiveness of artificial intelligence-based software in identifying adrenal neoplasms on computed tomography images of the chest and abdominal organs available as of the first quarter of 2024. METHODS: Artificial intelligence-based software was tested in two modifications: a single-purpose service designed to detect adrenal neoplasms and comprehensive artificial intelligence service for analyzing non-contrast computed tomography image series of the chest and abdominal organs (including contrast-enhanced studies). Two datasets were used: dataset 1 included abdominal computed tomography scans, and dataset 2 comprised chest computed tomography scans. Each dataset consisted of 100 anonymized computed tomography studies of patients with (n = 50) and without (n = 50) adrenal neoplasms. The diagnostic accuracy of artificial intelligence-based software was determined by calculating the following statistical metrics: area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, and specificity. RESULTS: Testing of the artificial intelligence-based software on datasets with signs of adrenal neoplasms demonstrated high diagnostic accuracy metrics that exceeded the declared performance values. AUC ranged from 0.858 to 0.995 (the highest value was achieved by the artificial intelligence single-purpose service-2 for analyzing abdominal computed tomography images); specificity ranged from 0.920 to 1.000 (the highest value was achieved by the artificial intelligence comprehensive service-2 for analyzing chest computed tomography images); and sensitivity ranged from 0.739 to 1.000 (the highest values were achieved by the artificial intelligence single-purpose and artificial intelligence comprehensive service-2 for analyzing abdominal computed tomography images). CONCLUSION: Artificial intelligence-based software for detecting adrenal neoplasms showed high diagnostic accuracy across all the evaluated metrics. Therefore, such systems may be effective for identifying adrenal neoplasms on chest and abdominal computed tomography scans of the chest and abdominal organs.

Lung cancer is one of the most common and socially significant cancers worldwide and consists of two main subtypes: small cell lung cancer (SCLC) and non-small cell lung cancer (NSCLC), which require different treatments. Computed tomography (CT) scans cannot reliably differentiate these subtypes, often necessitating invasive biopsies that carry significant risks. Radiomics offers a promising non-invasive alternative by quantitatively analyzing imaging data to extract detailed tissue characteristics beyond visual assessment. This pilot retrospective study analyzed 200 Moscow patients with histologically confirmed SCLC or NSCLC. Manual tumor segmentation on pretreatment CT scans allowed extraction of 107 radiomic features, from which 16 key features were selected to train four machine learning models. Models were evaluated using stratified 5-fold cross-validation, focusing on ROC AUC, accuracy, precision, and recall. All models demonstrated strong performance in distinguishing SCLC from NSCLC, with the gradient boosting model achieving the highest accuracy of 80.5% and ROC AUC of 0.888. These results highlight the potential of radiomics combined with machine learning to enable accurate, non-invasive differentiation of lung cancer subtypes. Further research is needed to expand feature sets, develop automated segmentation tools, and enhance clinical application of this approach.

X‑ray density of biological tissues is an important diagnostic parameter. Metal structures in the CT scanning area distort it, creating artifacts. Thus, hip joint endoprostheses (HJE) often complicate visualization of nearby soft tissue structures of the pelvic organs, which can interfere with the qualitative and quantitative analysis of changes when assessing the prevalence of the oncological process in this area. It is possible to correct these distortions using software methods, bringing the Hounsfield units (HU) values closer to the true ones.   Purpose of the study. To conduct a visual (qualitative) and quantitative assessment of metal artifacts in CT images using software methods for their reduction.   Materials and methods. A phantom was used for quantitative assessment: a plexiglass cylinder with a HJE in the center and test tubes with potassium hydrophosphate solution around it. The study was performed on a CT scanner with (FBP, iDose, iMR) reconstruction algorithms and O‑MAR technology for artifact suppression. The mean values and standard deviation of HU, the degree of susceptibility to artifacts were measured. Image quality was visually assessed using a five‑point Likert scale.   Results. The use of the O‑MAR algorithm does not distort HU in the absence of an HJE and smoothens the HU distribution in its presence. Deviation from the specified values at the level of the HJE neck decreased from 32–36 HU without O‑MAR to ‑1.5 – ‑4.7 HU with O‑MAR. The minimum noise was observed for iMR with O‑MAR at the level of the neck (31.6 HU) and stem (6.2 HU) of the HJE, the maximum – for FBP without O‑MAR (77.0 and 33.2 HU, respectively). The quality assessment was best for iMR with O‑MAR (3 points), the worst for FBP without O‑MAR (1.4 points). It was also shown that O‑MAR forms additional artifacts near the HJE.   Conclusion. Metal artifact reduction algorithms do not distort the X‑ray density without an artifact source. In the presence of metal structures, the algorithms reduce HU deviations and improve visualization, but they can form additional artifacts in the form of areas of increased and decreased density, so it is necessary to combine them with reconstruction without artifact reduction. To reduce the noise level, as well as to increase the contrast sensitivity, the use of model iterative reconstruction technology is optimal.

В экономически развитых странах растет нагрузка на системы здравоохранения из-за увеличения объемов медицинских услуг диагностической направленности. Однако, остается проблема дефицита кадров. Решением становится централизация лучевой диагностики в сочетании с автоматизацией процессов, при которой врачи-рентгенологи фокусируются только на анализе и интерпретацией результатов исследований. Необходимо провести оценку эффективность труда врачей-рентгенологов в условиях централизации в Московском референс-центр лучевой диагностики. Цель: комплексно оценить эффективность труда врачей-рентгенологов в условиях централизации лучевой диагностики. Материалы и методы: данные охватывали период с августа 2020 г. по июль 2024 г. Использована выгрузка из Единого радиологического информационного сервиса Единой медицинской информационно-аналитической системы г. Москвы. Структура полученной базы данных: деперсонализированный идентификатор врача-рентгенолога, дата начала работы, объем ставки, отработанные часы, фактически выполненные описания по модальностям, всего выполненных описаний, план и его выполнение в условных единицах и процентах, результаты аудита. Проведен расчет показателей интенсивности, трудоемкости, выработки, производительности труда. Результаты: в Московском референс-центр отмечается постоянное стабильное увеличение производительности труда. С 2020 по 2024 гг. произошел почти двукратный рост этого показателя: с 782,02 до 1325,20 соответственно. В 2020 г. трудоемкость работы врачей-рентгенологов в референс-центре в среднем составляла 12,57, к 2024 г. составила 8,77. Среднее значение интенсивности труда в 2022 г. составляло 92,34, в 2024 г. значение этого показателя составило 196,07. Проведен расчет показателя выработки. В 2020 г. он составил 8,57, в 2021 г. повысился до 11,07. В последующие периоды установилось «плато». Вывод: изменение условий труда и способов его осуществления повышает производительность труда врачей-рентгенологов с одновременным снижением его трудоёмкости. Модель организации лучевой диагностики на основе централизации более эффективна, чем распределение ресурсов по медицинским организациям в первичном звене здравоохранения. Economically developed countries face an increasing healthcare burden due to the growing number of diagnostic procedures. Compounding this issue, personnel shortages persist. Centralizing radiological care and transitioning to automation allow radiologists to focus on the analysis and interpretation of imaging findings. Assessing radiologist productivity within a centralized system, such as the Moscow Reference Center for Radiology, is necessary. Goal: to evaluate radiologist productivity within a centralized radiological care system. Materials and Methods: data from August 2020 to July 2024 were analyzed. The data were extracted from the Unified Radiological Information Service of the Unified Medical Information and Analytical System of Moscow. The database included anonymized radiologist identifiers, dates of employment, contractual working hours, hours worked, completed radiological reports (broken down by imaging modality), total completed reports, planned versus executed workload (in standardized units and percentages), and audit results. Labor effort, labor hours per unit, output, and productivity metrics were calculated. Results: the Moscow Reference Center demonstrated a consistent and stable increase in labor productivity. From 2020 to 2024, this rate nearly doubled, increasing from 782.02 to 1325.20. In 2020, the average labor hours per unit per radiologist was 12.57, decreasing to 8.77 by 2024. The average labor effort was 92.34 in 2022 and increased to 196.07 in 2024. In 2020, it was 8.57, increasing to 11.07 in 2021. The output indicator was calculated. In 2020, it was 8.57, increasing to 11.07 in 2021. A plateau was observed in subsequent periods. Conclusion: modifying working conditions and activities enhances radiologist productivity while reducing labor hours per unit. A centralized approach to radiological care proved more effective compared to distributing resources across primary care facilities.

Background: Existing literature primarily focuses on artificial intelligence (AI) ability to detect malignant breast tumors, often neglecting or limiting analysis to Breast Imaging-Reporting and Data System (BI-RADS) categories 4 and 5. The diagnostic performance of AI for other BI-RADS categories remains understudied. The objective of this study is to compare the diagnostic accuracy of three mammographic AI services in predicting individual BI-RADS categories and definition of opportunity integration of AI into routine clinical practice. Methods: Anonymized mammograms were obtained from the Unified Radiological Information Service of Moscow. Inclusion criteria: screening mammogram, radiology report from an AI and a human radiologist, age patients 40–75 years. Exclusion criteria: mammograms without BI-RADS categories, BI-RADS categories 0 and 6. The AI performance was assessed by calculating their diagnostic performance using the radiologists’ opinion as the ground truth together with the calibration tests. Results: The study sample consisted of 81,895 mammograms. Median accuracy was 76.9%, with a positive predictive value (PPV) of 11.8%. The highest negative predictive value (NPV) was observed for BI-RADS 2 (78.5–83.4%). The second highest NPVs were observed for BI-RADS 1, 3, 4, and 5 (over 84.7%). Binary classification yielded median accuracy and PPV values of 80.5% and 98.6% respectively, compared to the calibration testing (76.0% and 84.7%). Conclusions: Most AI service metrics were suboptimal for individual BI-RADS prediction, potentially due to reliance on variable radiologist conclusions and lack of histological calibration. Binary classification demonstrated higher performance metrics, and no significant differences in NPV were observed across AI applications, which means they can be recommended to confirm the absence of pathology. Successful integration of AI into routine clinical practice requires consideration of various diagnostic accuracy assessment methods, tailored to specific use cases.