Volume 6, Nº 2 (2025)

BACKGROUND: The accuracy of quantitative parameters obtained using magnetic resonance imaging is of scientific and practical interest. Monitoring of scan parameters and standardization of commonly used approaches to assess fat fraction remain challenging in radiology.

AIM: This study aimed to evaluate the possibility of fat fraction quantification using standard Dixon pulse sequences through phantom modeling.

METHODS: This multicenter, cross-sectional, nonblinded experimental study used direct oil-in-water emulsions to model substances with varying fat concentrations. Test tubes containing these emulsions were placed in a cylindrical phantom. The emulsions were prepared with mixtures of vegetable oils, with fat fraction values of 10%–60%. Several tests were conducted using scanners from different manufacturers and with varying magnetic field strengths: Optima MR450w, 1.5 T; MAGNETOM Skyra, 3 T; Ingenia, 1.5 T; and Ingenia Achieva dStream, 3.0 T at different medical centers. Fat fraction was obtained using standard formulas based on signal intensity measurements. A regression analysis was conducted to assess the linear relationship between the measured and predefined fat fraction concentrations and an F-test to evaluate variability.

RESULTS: Phantom modeling was employed to determine the performance of Dixon pulse sequences across different magnetic resonance imaging scanners for quantitative fat fraction estimation using relevant formulas. In assessing the accuracy of fat fraction quantification, a weak linear correlation was found between the obtained values and predefined fat fraction concentrations. Additionally, significant deviations >5% were observed for certain scanners. Reproducibility analysis demonstrated variability in fat fraction concentration across different scanner models and within the same model.

CONCLUSION: Obtained results confirm that fat fraction quantification using Dixon pulse sequences and relevant formulas should be performed only after preliminary phantom scanning. The use of a phantom ensures adequate quality control and calibration of the magnetic resonance imaging scanner, making accurate quantitative fat measurement more reliable and widely accessible.

BACKGROUND: Intracranial hemorrhages are associated with high mortality and risk of disability, requiring prompt and accurate diagnosis, particularly within the first 24 hours. The use of artificial intelligence technologies in analyzing brain computed tomography scans can shorten diagnostic time and improve diagnostic quality. The relevance of this study is emphasized by the limited number of certified artificial intelligence services for detecting intracranial hemorrhages in Russia and lacking data on their long-term effectiveness, highlighting the need for multicenter monitoring to assess the stability and accuracy of such systems in clinical practice.

AIM: The study aimed to assess the diagnostic accuracy and stability of an artificial intelligence service in detecting intracranial hemorrhages on non-contrast brain computed tomography scans in a multicenter clinical monitoring setting for 18 months.

METHODS: Anonymized brain computed tomography scans were used. The artificial intelligence service underwent a three-phase evaluation to evaluate its diagnostic accuracy and clinical performance using limited datasets. Two radiologists specializing in neuroimaging examined 80 brain computed tomography scans each month for 18 months, which had been preprocessed by the artificial intelligence service and randomly selected from the clinical workflow. The results were analyzed using ROC analysis with sensitivity, specificity, accuracy, and area under the curve.

RESULTS: During clinical monitoring, 1200 brain computed tomography scans were analyzed, with signs of intracranial hemorrhage detected in 48.3% of the scans. Based on the binary classification of intracranial hemorrhage presence or absence performed by the artificial intelligence service, the following diagnostic metrics were obtained: sensitivity, 97.4% (95.8–98.5); specificity, 75.4% (71.8–78.7); accuracy, 86.0% (83.9–87.9); and area under the curve, 94% (92.6–95.3). Eventually, a significant moderate positive correlation was observed in most diagnostic metrics and the time variable, except for sensitivity, which was affected by an update to the service version. However, full concordance between artificial intelligence-based markings and radiologist conclusions was noted in 28.5% of cases of identified intracranial hemorrhage, whereas discrepancies were found in 71.5%. The refined diagnostic metrics for cases with complete agreement with the radiologists’ report were as follows: sensitivity, 26.6%; specificity, 73.8%; accuracy, 50.1%; and area under the curve, 49.6%.

CONCLUSION: The current configuration of the artificial intelligence service allows ruling out intracranial hemorrhage with very high probability, which may be useful in the initial triaging of patients in emergency settings. However, low values of refined metrics indicate considerable discrepancies between radiologist reports and service-generated results regarding the interpretation of pathological findings.

BACKGROUND: The increasing availability of ¹⁸F-prostate-specific membrane antigen-1007 (¹⁸F-PSMA-1007) for prostate cancer staging highlighted its advantages, particularly its higher spatial resolution compared to analogs. Moreover, accumulating scientific data indicate an increase in false-positive findings, predominantly in bones, which may lead to unwarranted upstaging of the disease. Diffusion-weighted imaging may be used for the early detection of bone metastases.

AIM: This study aimed to assess and compare the diagnostic accuracy of whole-body ¹⁸F-PSMA-1007 positron emission tomography combined with computed tomography and whole-body and pelvic bone diffusion-weighted imaging in patients with prostate cancer.

METHODS: A retrospective single-center selective study was conducted. The imaging results of 119 patients with prostate cancer were divided into two groups: group 1 comprised 40 pairs of ¹⁸F-PSMA-1007 positron emission tomography combined with computed tomography and whole-body diffusion-weighted magnetic resonance imaging scans, and group 2 included 79 pairs of similar studies, with magnetic resonance imaging performed only for the pelvic bones. The diagnostic studies were performed at an inter-study interval ≤14 days. The metastatic bone lesions detected in different anatomical regions was counted using data from ¹⁸F-PSMA-1007 positron emission tomography combined with computed tomography and magnetic resonance imaging. Lesions were considered true positives if confirmed by additional magnetic resonance imaging pulse sequences and/or follow-up observation.

RESULTS: Whole-body diffusion-weighted imaging demonstrated higher specificity (58.1%) for detecting bone metastases than ¹⁸F-PSMA-1007 positron emission tomography combined with computed tomography (51.06%). However, its sensitivity was lower: 93.22% versus 97.55%.

CONCLUSION: Despite its advantages, ¹⁸F-PSMA-1007 positron emission tomography combined with computed tomography shows a high rate of false-positive bone findings. These are most commonly noted in the ribs, vertebrae, and pelvic bones. Suspicious bone lesions should be further evaluated to avoid unjustified disease upstaging. Thus, whole-body magnetic resonance imaging with diffusion-weighted sequences and selective fat signal suppression can be used.

BACKGROUND: With the increasing volume of data, laboratory medicine requires automation and standardization of routine processes to reduce workload on healthcare professionals and clear their time for more specialized tasks. Machine learning models and artificial neural networks support image recognition and analysis of large data sets, which allows their integration into laboratory workflows to solve routine tasks.

AIM: This study aimed to analyze global scientific publications on the application of artificial intelligence technologies in laboratory medicine and their potential to address current challenges and identify barriers in their integration into laboratory workflows.

METHODS: A search for publications was conducted using PubMed, manufacturer websites offering ready-to-use laboratory solutions, and reference lists from other reviews. The Mendeley software was utilized for bibliographic data management. The search covered the time interval 2019–2024. Obtained data included bibliometric indicators, research areas, key methodological characteristics, diagnostic effectiveness values for artificial intelligence systems and healthcare professionals, the number and experience of involved healthcare professionals, and validated outcomes of artificial intelligence implementation. The study quality was assessed using a modified QUADAS-CAD checklist.

RESULTS: Twenty-three publications presenting studies at the pre-analytical (n = 1), analytical (n = 19), and post-analytical (n = 3) stages of laboratory analysis were included. Most studies focused on cytology and microbiology, accounting for 48% and 35% of the studies, respectively. Artificial intelligence demonstrated high effectiveness in solving tasks across all stages of the laboratory process. Moreover, its diagnostic accuracy was comparable to that of healthcare professionals; however, decision-making speed was higher. All studies demonstrated a risk of systematic bias, which was associated with unbalanced samples, lacking external data validation, and incomplete description of datasets and analytical methods.

CONCLUSION: Artificial intelligence demonstrates high potential in diagnostic accuracy and processing speed, making it a promising tool to be integrated into laboratory practice and automation of routine processes. However, to achieve this, research methodologies for artificial intelligence should be standardized to reduce the risk of systematic bias, establish reference values for laboratories to ensure the reproducibility and generalizability of results, raise awareness among healthcare professionals and patients on how artificial intelligence works to overcome prejudices, and develop reliable mechanisms for protecting personal data when using artificial intelligence.

Early diagnosis of thoracic aortic aneurysms and pathological pulmonary trunk dilation is crucial to prevent severe complications, including vascular wall rupture and acute right ventricular failure, and reduce cardiovascular mortality. This review examines contemporary imaging approaches for these conditions, focusing on computed tomography as the gold standard modality. Emphasis was placed on the implementation of artificial intelligence technologies, which enable automatic segmentation of vascular structures, measurement of their diameter, and opportunistic screening, allowing early detection of asymptomatic conditions without additional diagnostic procedures, thereby reducing radiologist workload and improving medical care quality. The study comprehensively analyzed the Moscow Experiment, wherein the application of artificial intelligence in medical image analysis showed high sensitivity, reproducibility, and reduced reporting time. Despite these significant advantages, the need for expert supervision of artificial intelligence-generated results to ensure diagnostic accuracy and reliability is emphasized. Moreover, the review highlights the importance of adapting algorithms to different scanning protocols and population-specific features. Additionally, the importance of interdisciplinary collaboration among cardiologists, radiologists, data scientists, and software developers for the effective integration into routine clinical practice is pointed out. Therefore, the review outlines the potential of artificial intelligence technologies to enhance diagnostic quality and underscores the need for further clinical research and standardization of methods for successful integration into daily practice.

Pancreatic ductal adenocarcinoma is the most common pancreatic cancer. It is characterized by a progressive course or distant metastases in 80%–85% of cases. Despite advances in understanding of pancreatic ductal adenocarcinoma, the disease is consistently linked to poor prognosis due to late diagnosis and limited treatment options in advanced stages. Recently, image processing using artificial intelligence has been introduced for pancreatic ductal adenocarcinoma diagnosis and demonstrated promising results. This review summarizes current scientific data, evaluates the role of artificial intelligence in imaging and early detection of pancreatic ductal adenocarcinoma, and identifies issues that warrant further investigation. The search for publications was conducted using PubMed, Google Scholar, and eLibrary. The following Russian and English search keywords were used: ранняя диагностика рака поджелудочной железы (early diagnosis of pancreatic cancer), искусственный интеллект (artificial intelligence), протоковая аденокарцинома поджелудочной железы (pancreatic ductal adenocarcinoma), медицинская визуализация (medical visualization), наночастицы (nanoparticles), pancreatic cancer, artificial intelligence, early diagnosis pancreatic ductal adenocarcinoma, and pancreatic cancer imaging. Significant progress in early detection of pancreatic ductal adenocarcinoma using artificial intelligence technologies was observed. Current approaches include pre-imaging risk stratification and increased data volume by analyzing electronic medical records. Despite substantial achievements, the clinical implementation of artificial intelligence technologies remains challenging. The use of artificial intelligence along with biomarkers is a promising direction and may enhance theranostics of various malignancies, including pancreatic ductal adenocarcinoma.

This article introduces a new submission policy for original research manuscripts. Starting next year, such manuscripts will be considered only if they comply with the reporting guidelines recommended by the EQUATOR (Enhancing the QUAlity and Transparency Of health Research) Network and are approved by an ethics committee during the planning stage. The information to be included in a manuscript upon submission has been expanded. When these rules are fully implemented, prior study registration and submission of primary research data to the editorial office (with subsequent publication upon article acceptance) will become mandatory for manuscript consideration. These changes aim to transition the journal from organic growth to controlled development according to the principles of scientific integrity.