Polish journal of radiology最新文献

Purpose: Synthetic magnetic resonance imaging (MRI) allows reconstruction of multiple contrast-weighted images from a single acquisition of multiple delay multiple echo (MDME) sequence with quantitative relaxometry (longitudinal relaxation rate [R1], transverse relaxation rate [R2], and proton density [PD]) in a shorter acquisition time. We tried to explore synthetic MR-based relaxometry to differentiate central nervous system (CNS) tuberculomas from primary CNS neoplasm like glioblastoma.

Material and methods: Ten cases of CNS tuberculoma and 14 cases of glioblastoma underwent pre- and post-contrast synthetic MRI. R1, R2, and PD values were calculated from lesion core, wall, and perilesional oedema using free-hand region of interest and compared across the 2 groups.

Results: Both pre- and post-contrast R1 and R2 relaxation rates from core were significantly higher in tuberculoma (mean pre-contrast R1 - 0.93, R2 - 15.02; post-contrast R1 - 1.51, R2 - 15.48) from glioblastoma (mean pre-contrast R1 - 0.36, R2 - 4.58; post-contrast R1 - 0.43, R2 - 4.78). The same values were higher in perilesional oedema of glioblastoma (mean pre-contrast R1 - 0.75, R2 - 9.9; post-contrast R1 - 0.78, R2 - 10.48) compared to tuberculoma (mean pre-contrast R1 - 0.68, R2 - 8.57; post-contrast R1 - 0.72, R2 - 8.67). No significant difference was seen between relaxometry parameters from the walls of lesions.

Conclusions: Synthetic MR-based relaxometry can be useful in distinguishing CNS tuberculomas from glioblastoma. R1 and R2 relaxation rates from core of the lesions are most important in differentiating the two with R1 value > 0.852 and R2 value > 11.565 from core strongly suggests tuberculoma over glioblastoma.

Purpose: Extracranial internal carotid artery dissections (EICAD) remain a relatively common cause of ischaemic events in young patients. Currently, there is no consensus on standardised use of endovascular therapy in the treatment of these patients, but available data suggest that conservative treatment is not sufficient in 15% of cases. The aim of our study was to evaluate if endovascular stent placement was safe and effective for the treatment of extracranial internal carotid artery dissection, and whether it should be considered in properly selected patients.

Material and methods: This single-centre, retrospective study aimed to evaluate procedural and clinical outcomes of patients with EICAD who underwent endovascular stenting between 2015 and 2024. Procedural and clinical efficacy and safety, the rate of complications, and long-term outcomes were noted.

Results: A total of 21 patients (10 females) with an average age of 53 years underwent stenting for EICAD. Technical success was achieved in all cases. Perioperative complications were noted in 2 cases. Neurological evaluation performed at 6-month follow-up showed very good clinical results in the majority of cases (mRS 0 and mRS 1 were 76% and 19%, respectively). Control imaging examinations confirmed stent patency in all cases. No long-term mortality was observed.

Conclusions: This retrospective study demonstrated procedural and clinical safety and efficacy of endovascular stenting in patients with extracranial internal carotid artery dissection. That is why endovascular therapy should be proposed to individuals with unsatisfactory response to medical treatment and in cases of disease progression.

Purpose: We aim to evaluate the reproducibility of these features and apply machine learning algorithms to predict cancer diagnosis.

Material and methods: We analyzed magnetic resonance (MR) images from a cohort of 82 individuals, split between 41 prostate cancer patients and 41 healthy controls. A total of 215 radiomic features were extracted from T2-weighted and ADC images using the Software Environment for Radiomic Analysis (SERA). Intraclass correlation coefficient (ICC) analysis was used to assess the reproducibility of features, and Pearson's correlation was applied to remove redundant features. After feature selection, seven dimensionality reduction techniques, including principal component analysis (PCA), kernel PCA, linear discriminant analysis, and locally linear embedding, were applied to preprocess the radiomic features. Ten machine learning algorithms, including support vector machines (SVM), random forests, neural networks, logistic regression, and ensemble methods such as CatBoost and AdaBoost, were utilized to classify cancerous versus non-cancerous tissues. Model performance was evaluated using accuracy and AUC-ROC metrics.

Results: The results showed that features with high reproducibility (ICC > 0.75) contributed significantly to the performance of machine learning models. SVM, neural networks, and logistic regression achieved the highest accuracy (0.88-0.9) and AUC (up to 0.93) when using features from the good and excellent reproducibility categories. PCA emerged as the most effective dimensionality reduction method, preserving the discriminative power of reproducible features across all models.

Conclusion: The results indicate that radiomic feature extraction from MR images, combined with dimensionality reduction and machine learning algorithms, provides a robust approach for prostate cancer diagnosis.

Purpose: This study aims to assess the accuracy of artificial intelligence (AI) in mandibular canal (MC) segmentation on cone-beam computed tomography (CBCT) compared to semi-automatic segmentation. The impact of third molar status (absent, erupted, impacted) on AI performance was also evaluated.

Material and methods: A total of 150 CBCT scans (300 MCs) were retrospectively analysed. Semi-automatic MC segmentation was performed by experts using Romexis software, serving as the reference standard. AI-based segmentation was conducted using Diagnocat, an AI-driven cloud-based platform. Three-dimensional segmentation accuracy was assessed by comparing AI and semi-automatic segmentations through surface-to-surface distance metrics in Cloud Compare software. Statistical analyses included the intraclass correlation coefficient (ICC) for inter- and intra-rater reliability, Kruskal-Wallis tests for group comparisons, and Mann-Whitney U tests for post-hoc analyses.

Results: The median deviation between AI and semi-automatic MC segmentation was 0.29 mm (SD: 0.25-0.37 mm), with 88% of cases within the clinically acceptable limit (≤ 0.50 mm). Inter-rater reliability for semi-automatic segmentation was 84.5%, while intra-rater reliability reached 95.5%. AI segmentation demonstrated the highest accuracy in scans without third molars (median deviation: 0.27 mm), followed by erupted third molars (0.28 mm) and impacted third molars (0.32 mm).

Conclusions: AI demonstrated high accuracy in MC segmentation, closely matching expert-guided semi-automatic segmentation. However, segmentation errors were more frequent in cases with impacted third molars, probably due to anatomical complexity. Further optimisation of AI models using diverse training datasets and multi-centre validation is recommended to enhance reliability in complex cases.

Brucellosis is a zoonotic disease caused by Gram-negative bacteria of the Brucella genus that can be acquired through contact with a contaminated animal or its secretions. The course of the disease can be acute, chronic, or persistent. Axial skeleton and central nervous system (CNS) are among the most common affected locations and may be involved in each of the forms. Due to the varying clinical picture of the disease, diagnosis is made mainly on the basis of laboratory examinations that detect specific IgM and IgG antibodies in blood or other biological material and/or cultures. Imaging methods, especially magnetic resonance imaging, can aid in establishing proper diagnosis, monitoring of the disease and, to some extent, enable differential diagnosis before obtaining the laboratory tests results. The aim of this review is to present imaging features of Brucella infection of the spine and CNS and provide the recent advancements in the field.

Purpose: To compare the image quality in single-pass split-bolus abdominal computed tomography (CT) and conventional biphasic CT in abdominal trauma patients.

Material and methods: Sixty-six consecutive abdominal trauma patients referred for CT were randomised into 2 groups: the study group (n = 33), scanned using the split-bolus technique; and the control group (n = 33), scanned using the conventional biphasic technique. CT image quality was analysed subjectively by 2 observers based on a 5-point Likert scale. The images were also analysed quantitatively for attenuation values achieved by region of interest (ROI) placements in major arteries, veins, and solid organs. In addition, the radiation dose in terms of the dose length product (DLP) was compared between the 2 groups.

Results: The image quality in both groups ranged from good to excellent in most cases. There was no statistically significant difference in subjective image quality in both the groups as assessed by Likert score. Attenuation values in solid organs and major venous structures were significantly higher in the split-bolus group (p < 0.001). Arterial attenuation values were significantly higher in the control group (p < 0.001), but diagnostic levels were achieved in all patients. There was a reduction of 31.1% in DLP in the split-bolus group.

Conclusions: The split-bolus technique offers comparable image quality and higher solid organ and venous enhancement than conventional biphasic protocol at a reduced radiation dose.

Purpose: To evaluate the performance of a combined clinical-radiomics model using multiple machine learning approaches for predicting pathological differentiation in hepatocellular carcinoma (HCC).

Material and methods: A total of 196 patients with pathologically confirmed HCC, who underwent preoperative computed tomography (CT) were retrospectively enrolled (training: n = 156; validation: n = 40). The modelling process included the folowing: (1) clinical model construction through logistic regression analysis of risk factors; (2) radiomics model development by comparing 6 machine learning classifiers; and (3) integration of optimal clinical and radiomic features into a combined model. Model performance was assessed using the area under the curve (AUC), calibration curves, and decision curve analysis (DCA). A nomogram was constructed for clinical implementation.

Results: Two clinical risk factors (BMI and CA153) were identified as independent predictors of differentiated HCC. The clinical model showed moderate performance (AUC: training = 0.705, validation = 0.658). The radiomics model demonstrated improved prediction capability (AUC: training = 0.840, validation = 0.716). The combined model achieved the best performance in differentiating HCC pathological grades (AUC: training = 0.878, validation = 0.747).

Conclusions: The integration of CT radiomics features with clinical parameters through machine learning provides a promising non-invasive approach for predicting HCC pathological differentiation. This combined model could serve as a valuable tool for preoperative treatment planning.