Digital Diagnostics

BACKGROUND: Non-compacted left ventricular myocardium is a phenotypic variant of the myocardium structure with increased trabecularity and the formation of two distinct layers: compacted and non-compacted. It is known that patients with non-compacted left ventricular myocardium gradually develop contractile dysfunction, leading to chronic heart failure. Increased trabecularity can be observed both in healthy individuals and in certain non-ischemic heart diseases, particularly in dilated cardiomyopathy. In some cases, their differential diagnosis may be complicated, especially when the left ventricle ejection fraction is not reduced or is moderately reduced, and left ventricular dilation is minimal. Recently, publications have reported that myocardial strain parameters assessed using echocardiography and/or cardiac magnetic resonance imaging can detect subclinical disorders of myocardial contractility.

AIM: To assess the myocardial strain parameters using cardiac magnetic resonance imaging in patients with non-compacted left ventricular myocardium and preserved left ventricle ejection fraction compared with patients with dilated cardiomyopathy and healthy controls, and to determine the diagnostic value of these parameters for detecting subclinical disorders of myocardial contractility.

METHODS: This multicenter, observational, retrospective, cross-sectional study included patients with non-compacted left ventricular myocardium with preserved left ventricle ejection fraction (group 1), healthy volunteers (group 2), and patients with dilated cardiomyopathy (group 3). All patients underwent CMR with intravenous contrast administration.

RESULTS: A total of 112 participants were enrolled: 16 patients with non-compacted left ventricular myocardium with preserved left ventricle ejection fraction (group 1), 51 patients in the control group without pathological changes in heart chambers according to cardiac magnetic resonance imaging (group 2), and 45 patients with dilated cardiomyopathy (group 3). Mutations in the MYH7, MYBPC3, TTN, and DES genes had been previously detected in all patients with non-compacted left ventricular myocardium with preserved left ventricle ejection fraction. Analysis of the myocardial strain parameters revealed a decrease in the global and segmental strain parameters in patients with dilated cardiomyopathy; a decrease in longitudinal and circumferential strain in patients with non-compacted left ventricular myocardium compared with the control group in the middle and apical segments, and a decrease in radial strain in all segments.

CONCLUSION: Myocardial strain assessment can serve as a valuable diagnostic tool for detecting early contractility disorders in patients with non-compacted left ventricular myocardium with preserved left ventricle ejection fraction.

BACKGROUND: Surgical site infections are common postoperative complications that often develop after hospital discharge. Timely diagnosis and optimal treatment choice are crucial for clinical success and cost-effectiveness in surgical site infection treatment. Computer vision and artificial intelligence methods have demonstrated their effectiveness in analyzing chronic wounds, but their applicability to postoperative wound assessment remains poorly understood.

AIM: To compare the diagnostic accuracy of various machine learning models for the classification of surgical wound images to determine the presence or absence of surgical site infections in trauma and orthopedic patients.

METHODS: The study included patients aged ≥18 years who were hospitalized at two clinical centers for joint replacement, metal osteosynthesis, spine decompression-stabilization, or other interventions, or for the treatment of surgical site infections following these procedures. The following machine learning algorithms were used for infection recognition: support vector machines (SVMs), logistic regression (LR), random forests (RF), convolutional neural networks (VGG16 + CNN), and a model with an attention mechanism (YOLO 11s-cls).

RESULTS: The study group included 183 patients with surgical site infections, and the control group consisted of 115 patients without surgical site infections. A total of 512 surgical wound images (292 with surgical site infection and 220 without infection) obtained from 298 patients were included in the study. After augmentation, the training set comprised approximately 2500 images. The YOLO 11s-cls model demonstrated the best metrics on the test set: sensitivity 91.2%, accuracy 91%; F1-score 89%. However, the differences between this model and the VGG16 + CNN neural network in terms of sensitivity, specificity, and accuracy were not significant. For the remainig models, sensitivity ranged from 69.6% (RF) to 87% (VGG16 + CNN), accuracy ranged from 68% (RF) to 85% (VGG16 + CNN), and F1-score ranged from 66% (RF) to 83% (VGG16 + CNN).

CONCLUSION: The potential of using machine learning methods for remote monitoring of surgical wounds in trauma and orthopedic patients was confirmed. The developed models can be used to create a multimodal system for assessing and monitoring wound infection after surgical interventions.

BACKGROUND: The regulated quality control procedure for radiography involves the analysis of test objects without simulating anatomical features. However, subsequent assessment of radiologists’ performance or artificial intelligence-based decision support systems requires anonymized data samples from real patients annotated by multiple specialists. Collecting such datasets is a labor-intensive and time-consuming process, especially when we need a sample containing actual pathologies. Therefore, it is advisable to use anthropomorphic test objects (phantoms) which simulate pathological findings, so that the X-ray image will be as close as possible to the analyzed images of patients.

AIM: To develop and test an anthropomorphic aluminum phantom for digital radiography.

METHODS: A digital radiograph of the chest of a healthy person was used as a reference standard during phantom design. A sample of the designed phantom was made using layer-by-layer milling of an aluminum blank. Validation was carried out by comparing phantom images and the reference image. In addition, the effect of the phantom sample on the spatial resolution and contrast sensitivity was assessed for different acquisition modes of the X-ray unit. The applicability of the phantom for simulating focal lung lesions was studied, with the results evaluated by experts.

RESULTS: The phantom was found to reproduce the reference image with a high fidelity and allows simulation of pathological changes in chest X-rays by placing flexible pads in the required positions. The ratios of attenuation coefficients for different anatomical areas were identified during X-ray imaging of the phantom and three patients.

CONCLUSION: This work presents a method for developing a high-fidelity anthropomorphic phantom made of D16T aluminum alloy, designed for calibrating X-ray systems and testing radiologists. The phantom imitates the anatomical structures of the chest organs and pathological changes, such as focal lesions. The main advantages of this solution include reduced production costs, compact design, and high accuracy of reproduction of radiographic characteristics.

BACKGROUND: Chest X-ray is a widely used and accessible screening method for lung pathology. However, its diagnostic capabilities, particularly for visualizing pulmonary nodules, are limited. Suboptimal detection of pulmonary nodules by radiologists on chest X-rays affects the timeliness of diagnosis and therapeutic outcomes. One approach to improving lung nodule detection in chest X-rays involves adopting novel techniques, particularly those based on artificial intelligence. Nevertheless, the diagnostic value of these technologies in clinical practice remains largely unresolved due to limited data on radiologists’ performance conventional metrics.

AIM: To evaluate the diagnostic performance of 100 radiologists in identifying pulmonary nodules on chest X-rays.

METHODS: Each of the 100 radiologists was asked to evaluate 100 chest radiographs, of which 50 showed abnormal findings and 50 were normal. The presence of pulmonary nodules was assessed using the following scale: Absent (0 on the probability scale), likely absent (0.25), uncertain (0.50), likely present (0.75), present (1.00). Validation of nodule presence or absence was performed by three expert physicians using a binary scale (0/1) based on chest CT performed within 14 days after chest X-ray. We assessed image interpretation time, the difference in performance between radiologists and expert physicians (expressed in absolute units as D), and the primary diagnostic accuracy metrics of the radiologists.

RESULTS: The study yielded the following: ROC AUC: 0.858 ± 0.059, accuracy: 0.822 ± 0.048, sensitivity: 0.779 ± 0.097, and specificity: 0.864 ± 0.095. A negligible positive correlation was observed between expert accuracy and mean image interpretation time (Spearman correlation coefficient r_s = 0.189) and a low positive correlation (r_s = 0.344) between image interpretation time and the D value.

CONCLUSION: These results can be used to assess the quality of automated detection systems under development and to evaluate the efficacy of alternative methods and approaches for pulmonary nodule detection.

BACKGROUND: Diagnosis of head and neck malignancies and prediction of lymph node metastases are essential for determining the appropriate treatment strategy. Artificial intelligence-based systems for automated lymph node analysis have been developed for this purpose. The predictive accuracy of such systems can be enhanced by training on both radiological and clinical data.

AIM: To prepare a head and neck computed tomography dataset for the development of artificial intelligence systems designed to detect lymph node metastases.

METHODS: An anonymized dataset comprising contrast-enhanced head and neck computed tomography scans and associated clinical information was prepared. Computed tomography examinations were performed in patients aged >18 years with histologically confirmed malignancies, free of dental metallic hardware artifacts at the level of the target lymph nodes classified as Node-RADS categories 1 and 5, and without motion artifacts. The data were obtained from a single clinical center and extracted from the Unified Radiological Information Service / Unified Medical Information and Analytical System of Moscow. Computed tomography examinations were conducted between 2020 and 2023 using Toshiba Aquilion scanners. The dataset includes images acquired during the venous phase of contrast enhancement. Scanning was performed 70 seconds after peak aortic lumen attenuation (130 HU) was reached. A total of 75 lymph nodes classified as Node-RADS category 5 were annotated by three radiologists, each with >3 years of clinical experience.

RESULTS: The dataset contains 82 DICOM files with a total volume of 18.6 GB. To construct the dataset, head and neck computed tomography scans of outpatients from Oncology Center No. 1 of the Moscow City Hospital named after S.S. Yudin were selected between April 2024 and May 2025. The mean patient age was 62 ± 11.1 years (range, 33–84 years).

CONCLUSION: A publicly available dataset of contrast-enhanced head and neck computed tomography scans, along with clinical data from patients with malignancies, has been prepared and released. The dataset includes cervical lymph node annotations according to the Node-RADS classification (categories 1 and 5). It is intended for the development and validation of artificial intelligence algorithms for the detection and assessment of lymph nodes in head and neck oncology.

Timely detection of congenital malformations of the fetus remains one of the urgent problems of modern prenatal diagnostics. The survival rate of children, the volume and quality of medical care during treatment and rehabilitation directly depend on early and reliable diagnosis. In modern medicine, prenatal diagnosis is an obligatory complex of medical manipulations, various methods of examining patients to monitor the health of a pregnant woman and fetus. Ultrasound is one of the main methods of medical imaging, as it is non-invasive, safe and informative in the examination of pregnant women. Recently, technologies for processing video files and static images using artificial intelligence have been actively used in ultrasound diagnostics.

This review collects and analyzes 52 sources by both foreign and domestic authors. The list of sources used includes domestic and foreign original research in the field of the use of artificial intelligence in prenatal diagnostics, systematic reviews, methodological manuals, practical and clinical recommendations, and monographs. PubMed, Google Scholar, and eLibrary were selected as search engines. A comprehensive search was performed using keywords in Russian and English: искусственный интеллект / artificial intelligence, ультразвуковая диагностика / ultrasound diagnostics, нейросеть / neural network, плод / fetus, and врождённые пороки развития / congenital malformations. The search depth was 6 years (from 2020 to 2025).

The review revealed the obvious advantages of using neural network systems in prenatal diagnostics. Automation and standardization of fetal ultrasound examination make it possible to create a real-time neural network analysis algorithm and ensure quality control of the resulting echographic images. The undoubted advantages of artificial intelligence technologies are minimizing the variability of instrumental diagnosis between different specialists and reducing the time to obtain the “correct” echographic section. In addition, artificial intelligence enables the automatic identification of standard scanning planes, “recognition” of anatomical structures, and biometric measurements in the fetus.

There are problems associated with the introduction of modern intelligent decision support systems in healthcare around the world and in Russia in particular. The most pressing issues are medical, legal, and ethical issues, the problem of lack of transparency in decision-making (“black box”), leading to skepticism among specialists, and poor effectiveness in diagnosing rare anomalies due to the small amount of training material.

Today, modern computer technologies with the function of neural network analysis in prenatal diagnostics should be considered as a powerful auxiliary tool for doctors.

Neurofibromatosis type 1 is an autosomal dominant disorder characterized by the development of benign and, occasionally, malignant peripheral nerve sheath tumors. Although cutaneous manifestations are well described, breast involvement is rare and may mimic other benign or malignant breast lesions, posing a significant diagnostic challenge.

We report the case of a 55-year-old woman with a known history of neurofibromatosis type 1 who presented with a progressively enlarging, palpable mass in her left breast. Ultrasound imaging demonstrated a well-circumscribed, heterogeneous hypoechoic lesion, whereas mammography revealed a poorly defined mass within dense breast parenchyma. Histopathological examination following surgical excision confirmed the diagnosis of a neurofibroma composed of spindle-shaped Schwann cells with diffuse S-100 protein positivity and no evidence of cytological atypia.

Surgical excision was performed for symptom relief and cosmetic reasons, resulting in complete resolution of symptoms and a satisfactory aesthetic outcome. This case highlights the importance of considering neurofibromatosis during the differential diagnosis of breast masses, particularly in patients with verified neurofibromatosis type 1. Increased awareness of this rare manifestation allows timely diagnosis, appropriate surgical planning, and adequate genetic counseling.

Updated recommendations for breast surveillance—such as annual contrast-enhanced magnetic resonance imaging between the ages of 30 and 50 according to National Comprehensive Cancer Network 2025 guidelines—and new therapeutic options, including MEK inhibitors for symptomatic neurofibromas, should be considered in the comprehensive management of neurofibromatosis type 1 patients.

Diagnosis of post-traumatic peripheral neuropathy relies on a combination of methods, including clinical examination, electroneuromyography, ultrasound, and magnetic resonance imaging. However, ultrasound is operator-dependent, and magnetic resonance imaging is often unavailable or contraindicated, as in patients with metallic foreign bodies. Computed tomography has traditionally not been considered a method of choice for visualizing nerve structures due to its low soft-tissue contrast.

This clinical case describes a 35-year-old male with post-traumatic neuropathy of the left sciatic nerve following a shrapnel wound to the thigh. Conservative therapy for >6 months was ineffective. Computed tomography performed to rule out osteomyelitis allowed detailed visualization of the sciatic nerve along its course, revealing scar tissue compressing the nerve in the middle third of the thigh and a metal fragment in its bifurcation zone. Quantitative analysis of nerve tissue density revealed a pattern: the density of the healthy nerve on the contralateral side was in the negative Hounsfield unit range, which is characteristic of normal nerve tissue, whereas the affected nerve had higher values, reaching a maximum in the compression zone. These findings were fully confirmed intraoperatively during microsurgical neurolysis.

This case demonstrates that native computed tomography may in some cases serve as an additional imaging method in patients with post-traumatic sciatic neuropathy, especially in clinically challenging situations where magnetic resonance imaging is contraindicated or unavailable.