Polish journal of radiology最新文献

Introduction: Neurofibromatosis type 1 (NF-1) is an autosomal dominant genetic disorder notably linked to the development of central nervous system neoplasms - predominantly low-grade glial tumours like pilocytic astrocytoma. High-grade glial neoplasms (HGG) are rarer and more prevalent in adults, with very few comprehensive studies on imaging features of the same. This study aims to investigate the imaging characteristics of HGG in patients with NF-1, to identify alarming imaging features that potentially indicate higher-grade tumours.

Material and methods: Conducting a retrospective analysis, we examined histologically confirmed cases of HGG within clinically diagnosed NF-1 patients over 8 years. Our analysis scrutinised various imaging parameters, and histopathological and molecular data.

Results: Eight cases of NF-1-associated HGG were identified. Predominant features included large tumour size (> 5 cm) in most (77.8%), intra-tumoral necrosis (77.8%), and moderate to marked perilesional oedema (55.55%). Notably, more than half were centred in midline structures. Molecular analysis highlighted diverse statuses of ATRX, IDH1R132H, and P53.

Conclusions: This retrospective analysis of the largest single-centre dataset on imaging of HGG in NF-1 patients reported in the literature underscores that it may be more common than previously surmised. The need to look for alarming imaging indicators and raise suspicion in atypical locations like midline structures is essential for early detection and appropriate treatment.

Purpose: Accurate geometrical measurements of ankle joint (AJ) X-rays are essential for planning and executing orthopaedic procedures like alloplasty. Reliable assessment of the projection correctness of the AJ radiograms has to precede such measurements, and it is thus a vital step in the process. To create an artificial intelligence-based tool for automatic assessment of the correctness of the X-ray image projection of AJ.

Material and methods: 1062 antero-posterior and lateral AJ X-rays were categorized into correct and rotated groups based on the literature. The database was split with an 80 : 10 : 10 ratio for training, validation, and test sets, respectively. Data analysis was conducted using 32 targeted neural networks, evaluating with binary metrics: accuracy, precision, recall, and F1 score.

Results: The Xception neural network yielded the best results. Accuracies of 1.0, 0.849, and 0.888 were obtained for the training, validation, and test sets, respectively. The test set metrics achieved by Xception were as follows: precision - 0.935, recall - 0.879, and F1 score - 0.906.

Conclusions: The model achieved high accuracy in recognizing the projection correctness compared to literature reports, which can directly result in a reduction in the workload for radiologists or orthopaedic specialists, as well as a reduced risk of misdiagnosis.

Purpose: Hepatic fibrosis can be predicted using fibrosis-4 (FIB-4) and fibrosis-5 (FIB-5) scores. Functional liver imaging score (FLIS) provides valuable information regarding hepatic function. We aimed to assess whether easily obtained blood parameters (FIB-4 and FIB-5) may be used to discriminate preserved and impaired hepatic function based on FLIS.

Material and methods: Patients who underwent dynamic upper abdominal MRI with gadoxetic acid were retrospectively reviewed (n = 101, mean age 61.0 ± 11.3 years). FLIS values were categorized as FLIS < 4 (impaired hepatic function) and FLIS ≥ 4 (preserved hepatic function). The discriminative potential of FIB-4 and FIB-5 was assessed by plotting ROC curves.

Results: While FIB-4 was significantly higher, FIB-5 was significantly lower in patients with FLIS < 4. They had significant discriminative value in distinguishing patients with preserved and impaired hepatic function (area under the ROC curves 0.794 for FIB-4 and 0.748 for FIB-5, p-values < 0.001). Comparison of the area under the 2 ROC curves revealed that FIB-4 and FIB-5 had similar discriminative values (p = 0.405). For FIB-4, a cut-off value of 4.2 had a sensitivity of 88.9%, specificity of 66.3%, PPV of 20.5%, and NPV of 98.4%, meaning that FIB-4 values of < 4.2 are valuable in ruling out FLIS < 4 (poor hepatic function). For FIB-5, a cut-off value of 36.2 had a sensitivity of 88.9%, specificity of 60.9%, PPV of 18.2%, and NPV of 98.2%, meaning that FIB-5 values of > 36.2 are valuable in ruling out FLIS < 4.

Conclusions: FIB-4 and FIB-5 are valuable in discriminating preserved and impaired hepatic function based on FLIS scoring with similar diagnostic performance.

Purpose: The aim of the study was to evaluate the interventions in endoAVF created with the WavelinQ 4-F EndoAVF system to clarify the role of this method in personalised vascular access strategy for end-stage renal disease patients.

Material and methods: Endovascular fistula creation was performed in 16 patients. The type of additional endovascular procedures during fistula creation was evaluated. Postoperative surgical and endovascular intervention was divided into maintenance and enhanced maturation procedures. The necessity of using assisted maturation was examined by Kaplan-Meier analysis, the rate of interventions per patients per year and evaluation of the type, time, and the relationship between intraoperative and postoperative interventions.

Results: Access primary patency was significantly lower than access cumulative and functional patency (p < 0.001). During endoAVF creation 73.3% patients required 16 additional endovascular procedures. After endoAVF creation 73.3% patients needed endovascular 22 procedures and 20% underwent surgical interventions. Nine (81.8%) out of 11 patients required intravascular procedures due to lack of fistula maturation - 9 (45%) angioplasties and 11 (55%) vein embolisations. The postoperative venous embolisation rate was significantly dependent on vein embolisation during endoAVF creation (p < 0.04). After endoAVF creation significantly more patients - 11 (73.3%) - required endovascular interventions compared to 3 (20%) with surgical interventions (p < 0.01). Postoperative endovascular and surgical interventions was 0.09 and 0.02 per patients per year, respectively.

Conclusions: Maturation of endoAVFs required significantly more endovascular than surgical interventions. Venous embolisation combined with fistula creation reduced postoperative embolisation.

Purpose: Metabolic dysfunction-associated steatotic liver disease is currently one of the most common forms of chronic liver disease. This study aimed to assess whether extending the standard abdominal ultrasound protocol with quantitative liver evaluation increased the number of detected cases of liver steatosis.

Material and methods: This study was a retrospective cross-sectional comparison of the detectability of liver steatosis in a study group of 108 patients analysed using the attenuation coefficient, in relation to a matched control group assessed qualitatively with B-mode.

Results: Quantitative assessment based on the attenuation coefficient detected more patients with liver steatosis than qualitative assessment based on B-mode. With visual assessment in B-mode, we missed a significant number of patients, mainly those with an S1 steatosis grade.

Conclusions: The inclusion of quantitative liver evaluation in everyday practice seems justified, despite current problems with selecting the optimal assessment method and the lack of population-specific cut-off values.

Purpose: Cervical cancer continues to be one of the leading causes of death among females worldwide, and thus early diagnosis by using more advanced diagnostic procedures is crucial. The conventional Pap-smear procedure is accurate but subject to human error; thus, computerised, standardised, and automated diagnosis becomes imperative. Herein we present a novel framework of a fuzzy distance-based ensemble of convolutional neural networks (CNNs) for efficient cervical cancer classification from Pap-smear images.

Material and methods: The proposed approach integrates 5 models of CNN - Simple CNN, InceptionV3, Xception, Xception with Attention, and Inception Attention - via attention mechanisms to advance feature learning. A fuzzy distance-based aggregator function is introduced to fuse the predictions of these models optimally as per Euclidean, Manhattan, and cosine distance measures. Four advanced pre-processing techniques - wavelet denoising, contrast-limited adaptive histogram equalisation (CLAHE), background correction, and Laplacian sharpening - are employed to construct a cleaner dataset with enhanced image sharpness and segmentation.

Results: Experimental outcomes prove that the model is significantly better than state-of-the-art approaches, with an accuracy of 94% on the original dataset and 98.3% on the pre-processed dataset.

Conclusions: The method suggested herein has better noise robustness, interpretability through fuzzy logic, and automatic adaptation to various CNN frameworks without fine-tuning. These results acknowledge the promise of fuzzy logic-based CNN ensembles to improve machine-based cervical cancer diagnosis, which could be mapped to better and scalable diagnostic instruments in medical imaging.

Purpose: This study aimed to optimise the dynamic coronary computed tomography perfusion (CTP) protocol, focusing on patient preparation, scanning parameters, and image acquisition, reconstruction, and interpretation. Future phases will evaluate the diagnostic accuracy of dynamic CTP in detecting haemodynamically significant coronary artery disease (CAD), using invasive coronary angiography (ICA) and fractional flow reserve (FFR) as reference standards.

Material and methods: Thirty-six symptomatic patients with confirmed or suspected CAD underwent dynamic CTP using a whole-heart coverage CT scanner (Revolution Apex CT, GE Healthcare). Two patients were excluded due to non-diagnostic CTP results. Of the remaining 34 patients, 24 underwent both cardiac CT angiography (CCTA) and CTP, while 19 underwent CCTA, CTP, and ICA. Preliminary data were analysed by comparing CTP findings with CCTA and ICA/FFR when available.

Results: Among 578 myocardial segments, 424 (73.3%) showed normal perfusion and 154 (26.6%) exhibited hypoperfusion. Of the 17 cases with perfusion deficits, ICA confirmed significant stenosis in 10, resulting in 100% sensitivity and 22% specificity for detecting haemodynamically significant stenosis. FFR assessment in 10 patients demonstrated 60% concordance between CTP, ICA, and FFR. Incorporating CTP into the diagnostic pathway led to a 29.4% reclassification in management strategies.

Conclusions: The low specificity observed for detecting significant CAD underscores the need for further refinement of the CTP protocol. Future research should aim to optimise myocardial blood flow thresholds to improve diagnostic specificity and clinical applicability.

Purpose: Differentiating active from non-active multiple sclerosis (MS) lesions is critical for disease management but often relies on gadolinium-enhanced magnetic resonance imaging (MRI), raising concerns about retention risks and costs. This study introduces a contrast-free, multi-sequence MRI approach using radiomics and machine learning to classify MS lesion activity.

Material and methods: A total of 187 lesions from 31 MS patients (mean age 42.5 ± 11.3 years; 64.5% female) at Amin Hospital (November 2024 - February 2025) were retrospectively analysed using a 1.5 T MRI scanner. Five sequences - T1-weighted (T1W), T2-weighted (T2W), fluid-attenuated inversion recovery (FLAIR), diffusion-weighted imaging (DWI), and susceptibility-weighted imaging (SWI) - were processed to extract 8905 radiomic features, refined to 127 via correlation and recursive feature elimination. XGBoost classified lesions as active or non-active, validated on an internal test set (n = 28 lesions), with performance assessed by area under the receiver operating characteristic curve (AUC-ROC).

Results: The XGBoost model achieved an AUC-ROC of 0.87 (95% CI: 0.82-0.92), sensitivity of 0.85, and specificity of 0.83, outperforming other classifiers (SVM AUC 0.84). FLAIR (35.4%) and T2W (28.3%) dominated feature contributions, with SWI (12.6%) enhancing accuracy (AUC dropped to 0.84 without SWI). Noise simulation (Gaussian σ = 0.1) confirmed robustness (AUC = 0.86).

Conclusions: This integration of SWI with conventional sequences in a unified radiomic model offers a promising contrast-free alternative for MS lesion classification, achieving promising accuracy comparable to radiologist performance on an internal test set (n = 28 lesions), pending external validation. External validation is needed to confirm the generalisability, but this approach could reduce gadolinium reliance in clinical practice.

Purpose: To evaluate how different artificial intelligence (AI)-powered approaches affect human performance in a demanding chest computed tomography (CT) task, such as distinguishing between viral pneumonias.

Material and methods: Three radiologists blindly evaluated 220 chest CT scans of viral pneumonia cases (n = 151 COVID-19; n = 69 other viruses), classifying them with a probabilistic scoring system (COVID-19 Reporting and Data System - CO-RADS) in 2 phases: before (S1) and after (S2) receiving AI classifier results. Two S2 scenarios were investigated: a default approach, with AI predictions available for all cases, and a selective approach, with AI limited to equivocal S1 cases (CO-RADS = 3). Inter-reader agreement (Gwet's AC2) and diagnostic performance were analysed.

Results: Radiologists demonstrated good-to-excellent agreement across all scenarios (AC2 = 0.77-0.81). Evaluation changes between S1 and S2 occurred in 18% of cases, with 29% of cases initially classified as CO-RADS = 3. In these equivocal cases, AI led to an average correct classification rate of 85%. Conversely, when radiologists were confident in their S1 diagnoses (CO-RADS ≠ 3), classification changes in S2 occurred in 7% of cases, preventing incorrect diagnoses in 45% of patients but resulting in missed correct classifications in 55%. Regarding diagnostic performance, S1 accuracy was 78%, with 15% of CO-RADS = 3 cases. In S2, under the default approach, accuracy increased to 81%, with 16% of CO-RADS = 3 cases, whereas the selective approach achieved 79% accuracy with only 10% of CO-RADS = 3 cases. Only the selective approach significantly reduced the proportion of equivocal cases (p < 0.009).

Conclusions: A selective AI approach effectively reduces diagnostic uncertainty without introducing unnecessary complexity, emphasising its potential to optimise radiological workflows in challenging diagnostic scenarios.

Purpose: The purpose of this study was to determine the effectiveness of ADC histogram analysis in diagnosing and determining the aggressiveness of peripheral zone (PZ) prostate cancer, and to reveal the relationship between Gleason and PI-RADS scores.Material and method: 61 patients who underwent standard 12-core and cognitive prostate biopsy and multiparametric prostate magnetic resonance imaging before biopsy were included in the study. According to the pathology results, patients were classified as either having clinically significant cancer with malignancy (n = 35) or as clinically insignificant - benign (n = 26). The effectiveness of ADC histogram parameters to distinguish between benign and malignant lesions was investigated. Subsequently, 35 patients in the malignant group were grouped according to their Gleason scores, and the relationship between ADC histogram parameters and Gleason scores was examined.

Results: ADC max, standard deviation, entropy, voxel count, and volume were found to be significantly different between the benign and malignant groups (p < 0.05; p < 0.05; p < 0.01; p < 0.01; p < 0.01). According to the ROC curve: entropy (AUC = 0.75; 95% CI: 0.63-0.87), voxel count (AUC = 0.83; 95% CI: 0.73-0.93), and volume values (AUC = 0.83; 95% CI: 0.73-0.93) were statistically significant in the diagnosis of benign and malignant lesions in the prostate gland (area under the ROC curves). In the logistic regression analysis models (backward), it was found that an increase in volume increased the risk of malignant tumours by 1.75 times (p = 0.04; OR = 1.75; 95% CI: 1.00-3.04).

Conclusions: ADC histogram data contribute to the diagnosis of benign-malignant differentiation in PZ prostate lesions and predict the Gleason score in malignant lesions.