Pub Date : 2025-06-01Epub Date: 2024-12-12DOI: 10.1016/j.ejro.2024.100622
Qinxuan Tan , Jingyu Miao , Leila Nitschke , Marcel Dominik Nickel , Markus Herbert Lerchbaumer , Tobias Penzkofer , Sebastian Hofbauer , Robert Peters , Bernd Hamm , Dominik Geisel , Moritz Wagner , Thula Cannon Walter-Rittel
Background
Deep learning (DL) accelerated controlled aliasing in parallel imaging results in higher acceleration (CAIPIRINHA)-volumetric interpolated breath-hold examination (VIBE), provides high spatial resolution T1-weighted imaging of the upper abdomen. We aimed to investigate whether DL-CAIPIRINHA-VIBE can improve image quality, vessel conspicuity, and lesion detectability compared to a standard CAIPIRINHA-VIBE in renal imaging at 3 Tesla.
Methods
In this prospective study, 50 patients with 23 solid and 45 cystic renal lesions underwent MRI with clinical MR sequences, including standard CAIPIRINHA-VIBE and DL-CAIPIRINHA-VIBE sequences in the nephrographic phase at 3 Tesla. Two experienced radiologists independently evaluated both sequences and multiplanar reconstructions (MPR) of the sagittal and coronal planes for image quality with a Likert scale ranging from 1 to 5 (5 =best). Quantitative measurements including the size of the largest lesion and renal lesion contrast ratios were evaluated.
Results
DL-CAIPIRINHA-VIBE compared to standard CAIPIRINHA-VIBE showed significantly improved overall image quality, higher scores for renal border delineation, renal sinuses, vessels, adrenal glands, reduced motion artifacts and reduced perceived noise in nephrographic phase images (all p < 0.001). DL-CAIPIRINHA-VIBE with MPR showed superior lesion conspicuity and diagnostic confidence compared to standard CAIPIRINHA-VIBE. However, DL-CAIPIRINHA-VIBE presented a more synthetic appearance and more aliasing artifacts (p < 0.023). The mean size and signal intensity of renal lesions for DL-CAIPIRINHA-VIBE showed no significant differences compared to standard CAIPIRINHA-VIBE (p > 0.9).
Conclusions
DL-CAIPIRINHA-VIBE is well suited for kidney imaging in the nephrographic phase, provides good image quality, improved delineation of anatomic structures and renal lesions.
{"title":"Deep learning enabled near-isotropic CAIPIRINHA VIBE in the nephrogenic phase improves image quality and renal lesion conspicuity","authors":"Qinxuan Tan , Jingyu Miao , Leila Nitschke , Marcel Dominik Nickel , Markus Herbert Lerchbaumer , Tobias Penzkofer , Sebastian Hofbauer , Robert Peters , Bernd Hamm , Dominik Geisel , Moritz Wagner , Thula Cannon Walter-Rittel","doi":"10.1016/j.ejro.2024.100622","DOIUrl":"10.1016/j.ejro.2024.100622","url":null,"abstract":"<div><h3>Background</h3><div>Deep learning (DL) accelerated controlled aliasing in parallel imaging results in higher acceleration (CAIPIRINHA)-volumetric interpolated breath-hold examination (VIBE), provides high spatial resolution T1-weighted imaging of the upper abdomen. We aimed to investigate whether DL-CAIPIRINHA-VIBE can improve image quality, vessel conspicuity, and lesion detectability compared to a standard CAIPIRINHA-VIBE in renal imaging at 3 Tesla.</div></div><div><h3>Methods</h3><div>In this prospective study, 50 patients with 23 solid and 45 cystic renal lesions underwent MRI with clinical MR sequences, including standard CAIPIRINHA-VIBE and DL-CAIPIRINHA-VIBE sequences in the nephrographic phase at 3 Tesla. Two experienced radiologists independently evaluated both sequences and multiplanar reconstructions (MPR) of the sagittal and coronal planes for image quality with a Likert scale ranging from 1 to 5 (5 =best). Quantitative measurements including the size of the largest lesion and renal lesion contrast ratios were evaluated.</div></div><div><h3>Results</h3><div>DL-CAIPIRINHA-VIBE compared to standard CAIPIRINHA-VIBE showed significantly improved overall image quality, higher scores for renal border delineation, renal sinuses, vessels, adrenal glands, reduced motion artifacts and reduced perceived noise in nephrographic phase images (all p < 0.001). DL-CAIPIRINHA-VIBE with MPR showed superior lesion conspicuity and diagnostic confidence compared to standard CAIPIRINHA-VIBE. However, DL-CAIPIRINHA-VIBE presented a more synthetic appearance and more aliasing artifacts (p < 0.023). The mean size and signal intensity of renal lesions for DL-CAIPIRINHA-VIBE showed no significant differences compared to standard CAIPIRINHA-VIBE (p > 0.9).</div></div><div><h3>Conclusions</h3><div>DL-CAIPIRINHA-VIBE is well suited for kidney imaging in the nephrographic phase, provides good image quality, improved delineation of anatomic structures and renal lesions.</div></div>","PeriodicalId":38076,"journal":{"name":"European Journal of Radiology Open","volume":"14 ","pages":"Article 100622"},"PeriodicalIF":1.8,"publicationDate":"2025-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11699112/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142932876","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-06-01Epub Date: 2025-04-08DOI: 10.1016/j.ejro.2025.100650
Liyan Li , Xueying Wang , Zeming Tan , Yipu Mao , Deyou Huang , Xiaoping Yi , Muliang Jiang , Bihong T. Chen
Objectives
To develop and validate a prediction model based on brain MRI features to predict disease-free survival (DFS) and overall survival (OS) for patients with intracranial extraventricular ependymoma (IEE).
Methods
The study included 114 patients with pathology-proven IEE, of whom 80 were randomly assigned to a training group and 34 to a validation group. Preoperative brain MRI images were assessed with the Visually AcceSAble Rembrandt Images (VASARI) feature set. Clinical variables were assessed including age, gender, KPS, pathological grade of the tumor and blood test data such as eosinophil, blood urea nitrogen and serum creatinine. Multivariate Cox proportional hazards regression analysis was performed to select the independent prognostic factors for DFS and OS. Three prediction models were built with clinical variables, MRI-VASARI features, and combined clinical and MRI-VASARI data, respectively. The predictive power of survival models was assessed using c-index and calibration curve.
Results
Clinical variables such as eosinophil, blood urea nitrogen and serum creatinine, and MRI-VASARI feature for definition of the non-enhancing margin (F13) were significantly correlated with the prognosis of DFS. Blood urea nitrogen, D-dimer, tumor location (F1), eloquent brain (F3), and T1/FLAIR ratio (F10) were independent predictors of OS. Based on these factors, prediction models were constructed. The concordance indices of the three survival models for OS were 0.732, 0.729, and 0.768, respectively. For DFS, the concordance indices were respectively 0.694, 0.576, and 0.714.
Conclusion
Predictive modelling combining both clinical and MRI-VASARI features is robust and may assist in the assessment of prognosis in patients with IEE.
{"title":"Preoperative MR - based model for predicting prognosis in patients with intracranial extraventricular ependymoma","authors":"Liyan Li , Xueying Wang , Zeming Tan , Yipu Mao , Deyou Huang , Xiaoping Yi , Muliang Jiang , Bihong T. Chen","doi":"10.1016/j.ejro.2025.100650","DOIUrl":"10.1016/j.ejro.2025.100650","url":null,"abstract":"<div><h3>Objectives</h3><div>To develop and validate a prediction model based on brain MRI features to predict disease-free survival (DFS) and overall survival (OS) for patients with intracranial extraventricular ependymoma (IEE).</div></div><div><h3>Methods</h3><div>The study included 114 patients with pathology-proven IEE, of whom 80 were randomly assigned to a training group and 34 to a validation group. Preoperative brain MRI images were assessed with the Visually AcceSAble Rembrandt Images (VASARI) feature set. Clinical variables were assessed including age, gender, KPS, pathological grade of the tumor and blood test data such as eosinophil, blood urea nitrogen and serum creatinine. Multivariate Cox proportional hazards regression analysis was performed to select the independent prognostic factors for DFS and OS. Three prediction models were built with clinical variables, MRI-VASARI features, and combined clinical and MRI-VASARI data, respectively. The predictive power of survival models was assessed using c-index and calibration curve.</div></div><div><h3>Results</h3><div>Clinical variables such as eosinophil, blood urea nitrogen and serum creatinine, and MRI-VASARI feature for definition of the non-enhancing margin (F13) were significantly correlated with the prognosis of DFS. Blood urea nitrogen, D-dimer, tumor location (F1), eloquent brain (F3), and T1/FLAIR ratio (F10) were independent predictors of OS. Based on these factors, prediction models were constructed. The concordance indices of the three survival models for OS were 0.732, 0.729, and 0.768, respectively. For DFS, the concordance indices were respectively 0.694, 0.576, and 0.714.</div></div><div><h3>Conclusion</h3><div>Predictive modelling combining both clinical and MRI-VASARI features is robust and may assist in the assessment of prognosis in patients with IEE.</div></div>","PeriodicalId":38076,"journal":{"name":"European Journal of Radiology Open","volume":"14 ","pages":"Article 100650"},"PeriodicalIF":1.8,"publicationDate":"2025-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143792716","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pulmonary embolism (PE) is the third most fatal cardiovascular disease in the United States. Currently, Computed Tomography Pulmonary Angiography (CTPA) serves as diagnostic gold standard for detecting PE. However, its efficacy is limited by factors such as contrast bolus timing, physician-dependent diagnostic accuracy, and time taken for scan interpretation. To address these limitations, we propose an AI-based PE triaging model (AID-PE) designed to predict the presence and key characteristics of PE on CTPA. This model aims to enhance diagnostic accuracy, efficiency, and the speed of PE identification.
Materials and methods
We trained AID-PE on the RSNA-STR PE CT (RSPECT) Dataset, N = 7279 and subsequently tested it on an in-house dataset (n = 106). We evaluated efficiency in a separate dataset (D4, n = 200) by comparing the time from scan to report in standard PE detection workflow versus AID-PE.
Results
A comparative analysis showed that AID-PE had an AUC/accuracy of 0.95/0.88. In contrast, a Convolutional Neural Network (CNN) classifier and a CNN-Long Short-Term Memory (LSTM) network without an attention module had an AUC/accuracy of 0.5/0.74 and 0.88/0.65, respectively. Our model achieved AUCs of 0.82 and 0.95 for detecting PE on the validation dataset and the independent test set, respectively. On D4, AID-PE took an average of 1.32 s to screen for PE across 148 CTPA studies, compared to an average of 40 min in contemporary workflow.
Conclusion
AID-PE outperformed a baseline CNN classifier and a single-stage CNN-LSTM network without an attention module. Additionally, its efficiency is comparable to the current radiological workflow.
{"title":"Comparing efficiency of an attention-based deep learning network with contemporary radiological workflow for pulmonary embolism detection on CTPA: A retrospective study","authors":"Gagandeep Singh , Annie Singh , Tejasvi Kainth , Sudhir Suman , Nicole Sakla , Luke Partyka , Tej Phatak , Prateek Prasanna","doi":"10.1016/j.ejro.2025.100657","DOIUrl":"10.1016/j.ejro.2025.100657","url":null,"abstract":"<div><h3>Rational and objectives</h3><div>Pulmonary embolism (PE) is the third most fatal cardiovascular disease in the United States. Currently, Computed Tomography Pulmonary Angiography (CTPA) serves as diagnostic gold standard for detecting PE. However, its efficacy is limited by factors such as contrast bolus timing, physician-dependent diagnostic accuracy, and time taken for scan interpretation. To address these limitations, we propose an AI-based PE triaging model (AID-PE) designed to predict the presence and key characteristics of PE on CTPA. This model aims to enhance diagnostic accuracy, efficiency, and the speed of PE identification.</div></div><div><h3>Materials and methods</h3><div>We trained AID-PE on the RSNA-STR PE CT (RSPECT) Dataset, N = 7279 and subsequently tested it on an in-house dataset (n = 106). We evaluated efficiency in a separate dataset (D<sub>4</sub>, n = 200) by comparing the time from scan to report in standard PE detection workflow versus AID-PE.</div></div><div><h3>Results</h3><div>A comparative analysis showed that AID-PE had an AUC/accuracy of 0.95/0.88. In contrast, a Convolutional Neural Network (CNN) classifier and a CNN-Long Short-Term Memory (LSTM) network without an attention module had an AUC/accuracy of 0.5/0.74 and 0.88/0.65, respectively. Our model achieved AUCs of 0.82 and 0.95 for detecting PE on the validation dataset and the independent test set, respectively. On D<sub>4</sub>, AID-PE took an average of 1.32 s to screen for PE across 148 CTPA studies, compared to an average of 40 min in contemporary workflow.</div></div><div><h3>Conclusion</h3><div>AID-PE outperformed a baseline CNN classifier and a single-stage CNN-LSTM network without an attention module. Additionally, its efficiency is comparable to the current radiological workflow.</div></div>","PeriodicalId":38076,"journal":{"name":"European Journal of Radiology Open","volume":"14 ","pages":"Article 100657"},"PeriodicalIF":1.8,"publicationDate":"2025-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143924087","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-06-01Epub Date: 2025-01-08DOI: 10.1016/j.ejro.2024.100633
Katarzyna Bokwa-Dąbrowska , Rafał Zych , Dan Mocanu , Michael Huuskonen , Dawid Dziedzic , Pawel Szaro
Diagnosing peroneus brevis split tears is a significant challenge, as many cases are missed both clinically and on imaging. Anatomical variations within the superior peroneal tunnel can contribute to peroneus brevis split tears or instability of the peroneal tendons. However, determining which anatomical variations predispose patients to these injuries remains challenging due to conflicting data in the literature. In this review, we present the current understanding of the role of anatomical variants in the development of peroneus brevis split tears. Many studies emphasize the significance of the retromalleolar groove and retromalleolar tubercle, the impact of a low-lying muscle belly, and the presence of accessory muscles within the superior peroneal tunnel as contributors to peroneal pathology. Hypertrophy of the peroneal tubercle or post-traumatic irregularities in the surface of the retromalleolar groove can accelerate degenerative changes in the peroneal tendons, potentially leading to peroneus brevis split tears. The topographic anatomy of the superior peroneal tunnel is essential for systematically performing ultrasound and interpreting magnetic resonance imaging of the ankle. The first part of this review focuses on the anatomical foundations of imaging diagnostics for peroneus brevis pathology. In the second part, we will examine the radiological spectrum of peroneal tendon injuries, offering a framework to enhance diagnostic confidence in this frequently underdiagnosed pathology.
{"title":"Peroneus brevis split tear – A challenging diagnosis: A pictorial review of magnetic resonance and ultrasound imaging. Part 1. Anatomical basis and clinical insights","authors":"Katarzyna Bokwa-Dąbrowska , Rafał Zych , Dan Mocanu , Michael Huuskonen , Dawid Dziedzic , Pawel Szaro","doi":"10.1016/j.ejro.2024.100633","DOIUrl":"10.1016/j.ejro.2024.100633","url":null,"abstract":"<div><div>Diagnosing peroneus brevis split tears is a significant challenge, as many cases are missed both clinically and on imaging. Anatomical variations within the superior peroneal tunnel can contribute to peroneus brevis split tears or instability of the peroneal tendons. However, determining which anatomical variations predispose patients to these injuries remains challenging due to conflicting data in the literature. In this review, we present the current understanding of the role of anatomical variants in the development of peroneus brevis split tears. Many studies emphasize the significance of the retromalleolar groove and retromalleolar tubercle, the impact of a low-lying muscle belly, and the presence of accessory muscles within the superior peroneal tunnel as contributors to peroneal pathology. Hypertrophy of the peroneal tubercle or post-traumatic irregularities in the surface of the retromalleolar groove can accelerate degenerative changes in the peroneal tendons, potentially leading to peroneus brevis split tears. The topographic anatomy of the superior peroneal tunnel is essential for systematically performing ultrasound and interpreting magnetic resonance imaging of the ankle. The first part of this review focuses on the anatomical foundations of imaging diagnostics for peroneus brevis pathology. In the second part, we will examine the radiological spectrum of peroneal tendon injuries, offering a framework to enhance diagnostic confidence in this frequently underdiagnosed pathology.</div></div>","PeriodicalId":38076,"journal":{"name":"European Journal of Radiology Open","volume":"14 ","pages":"Article 100633"},"PeriodicalIF":1.8,"publicationDate":"2025-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11764704/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143048214","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-06-01Epub Date: 2025-06-10DOI: 10.1016/j.ejro.2025.100664
Wen-Chang Tseng , Yung-Cheng Wang , Wei-Chi Chen , Kang-Ping Lin
Purpose
This study develops an AI-assisted pneumothorax diagnosis system using deep learning and chest X-ray images to enhance diagnostic efficiency and accuracy, reduce radiologists' workload, and provide timely treatment. The system addresses limitations of traditional methods, which rely on subjective interpretation and are vulnerable to fatigue or inexperience.
Methods
The DenseNet121 model was employed using a chest X-ray dataset from a medical center in northern Taiwan, with a total of 6888 images’ divided into training (64 %), validation (16 %), and testing (20 %) sets. Image preprocessing involved normalization, data augmentation (rotation, translation, scaling, brightness adjustment), and standardization. The model was trained using stochastic gradient descent with an initial learning rate of 0.0016 for 150 epochs. Performance evaluation included accuracy, sensitivity, specificity, and AUROC, integrating with the hospital's PACS for real-time analysis.
Results
Initial testing yielded AUROC values of 94.52 % and 97.21 % for pneumothorax and mild pneumothorax groups. However, when applied to 6888 clinical images, the AUROC dropped to 62.55 %, resulting in 4294 false positives. Adjusting the dataset split and retraining with 1000 false positive images improved the AUROC from 62.55 % to 85.53 %.
Conclusions
The AI model shows potential in pneumothorax detection, but performance is influenced by data diversity, image quality, and clinical complexity. The model struggles to identify key areas in complex cases, indicating a need for attention mechanisms or region proposal networks (RPN). Expanding the dataset, optimizing preprocessing, and training separate models for different image locations could enhance performance further.
{"title":"Development of an AI model for pneumothorax imaging: Dataset and model optimization strategies for real-world deployment","authors":"Wen-Chang Tseng , Yung-Cheng Wang , Wei-Chi Chen , Kang-Ping Lin","doi":"10.1016/j.ejro.2025.100664","DOIUrl":"10.1016/j.ejro.2025.100664","url":null,"abstract":"<div><h3>Purpose</h3><div>This study develops an AI-assisted pneumothorax diagnosis system using deep learning and chest X-ray images to enhance diagnostic efficiency and accuracy, reduce radiologists' workload, and provide timely treatment. The system addresses limitations of traditional methods, which rely on subjective interpretation and are vulnerable to fatigue or inexperience.</div></div><div><h3>Methods</h3><div>The DenseNet121 model was employed using a chest X-ray dataset from a medical center in northern Taiwan, with a total of 6888 images’ divided into training (64 %), validation (16 %), and testing (20 %) sets. Image preprocessing involved normalization, data augmentation (rotation, translation, scaling, brightness adjustment), and standardization. The model was trained using stochastic gradient descent with an initial learning rate of 0.0016 for 150 epochs. Performance evaluation included accuracy, sensitivity, specificity, and AUROC, integrating with the hospital's PACS for real-time analysis.</div></div><div><h3>Results</h3><div>Initial testing yielded AUROC values of 94.52 % and 97.21 % for pneumothorax and mild pneumothorax groups. However, when applied to 6888 clinical images, the AUROC dropped to 62.55 %, resulting in 4294 false positives. Adjusting the dataset split and retraining with 1000 false positive images improved the AUROC from 62.55 % to 85.53 %.</div></div><div><h3>Conclusions</h3><div>The AI model shows potential in pneumothorax detection, but performance is influenced by data diversity, image quality, and clinical complexity. The model struggles to identify key areas in complex cases, indicating a need for attention mechanisms or region proposal networks (RPN). Expanding the dataset, optimizing preprocessing, and training separate models for different image locations could enhance performance further.</div></div>","PeriodicalId":38076,"journal":{"name":"European Journal of Radiology Open","volume":"14 ","pages":"Article 100664"},"PeriodicalIF":1.8,"publicationDate":"2025-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144239690","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-06-01Epub Date: 2025-02-15DOI: 10.1016/j.ejro.2025.100638
Jingjing Pan , Qianyu Huang , Jiangming Zhu , Wencai Huang , Qian Wu , Tingting Fu , Shuhui Peng , Jiani Zou
Objectives
To develop and validate the value of different machine learning models of pericoronary adipose tissue (PCAT) radiomics based on coronary computed tomography angiography (CCTA) for predicting coronary plaque progression (PP).
Methods
This retrospective study evaluated 97 consecutive patients (with 127 plaques: 40 progressive and 87 nonprogressive) who underwent serial CCTA examinations. We analyzed conventional parameters and PCAT radiomics features. PCAT radiomics models were constructed using logistic regression (LR), K-nearest neighbors (KNN), and random forest (RF). Logistic regression analysis was applied to identify variables for developing conventional parameter models. Model performances were assessed by metrics including area under the curve (AUC), accuracy, sensitivity, and specificity.
Results
At baseline CCTA, 93 radiomics features were extracted from CCTA images. After dimensionality reduction and feature selection, two radiomics features were deemed valuable. Among radiomics models, we selected the RF as the optimal model in the training and validation sets (AUC = 0.971, 0.821). At follow-up CCTA, logistic regression analysis showed that increase in fat attenuation index (FAI) and decrease in PCAT volume were independent predictors of PP. The predictive capability of the combined model (increase in FAI + decrease in PCAT volume) was the best in the training and validation sets (AUC = 0.907, 0.882).
Conclusions
At baseline CCTA, the RF-based PCAT radiomics model demonstrated excellent predictive ability for PP. Furthermore, at follow-up CCTA, our results indicated that both increase in FAI and decrease in PCAT volume can independently predict PP, and their combination provided enhanced predictive ability.
{"title":"Prediction of plaque progression using different machine learning models of pericoronary adipose tissue radiomics based on coronary computed tomography angiography","authors":"Jingjing Pan , Qianyu Huang , Jiangming Zhu , Wencai Huang , Qian Wu , Tingting Fu , Shuhui Peng , Jiani Zou","doi":"10.1016/j.ejro.2025.100638","DOIUrl":"10.1016/j.ejro.2025.100638","url":null,"abstract":"<div><h3>Objectives</h3><div>To develop and validate the value of different machine learning models of pericoronary adipose tissue (PCAT) radiomics based on coronary computed tomography angiography (CCTA) for predicting coronary plaque progression (PP).</div></div><div><h3>Methods</h3><div>This retrospective study evaluated 97 consecutive patients (with 127 plaques: 40 progressive and 87 nonprogressive) who underwent serial CCTA examinations. We analyzed conventional parameters and PCAT radiomics features. PCAT radiomics models were constructed using logistic regression (LR), K-nearest neighbors (KNN), and random forest (RF). Logistic regression analysis was applied to identify variables for developing conventional parameter models. Model performances were assessed by metrics including area under the curve (AUC), accuracy, sensitivity, and specificity.</div></div><div><h3>Results</h3><div>At baseline CCTA, 93 radiomics features were extracted from CCTA images. After dimensionality reduction and feature selection, two radiomics features were deemed valuable. Among radiomics models, we selected the RF as the optimal model in the training and validation sets (AUC = 0.971, 0.821). At follow-up CCTA, logistic regression analysis showed that increase in fat attenuation index (FAI) and decrease in PCAT volume were independent predictors of PP. The predictive capability of the combined model (increase in FAI + decrease in PCAT volume) was the best in the training and validation sets (AUC = 0.907, 0.882).</div></div><div><h3>Conclusions</h3><div>At baseline CCTA, the RF-based PCAT radiomics model demonstrated excellent predictive ability for PP. Furthermore, at follow-up CCTA, our results indicated that both increase in FAI and decrease in PCAT volume can independently predict PP, and their combination provided enhanced predictive ability.</div></div>","PeriodicalId":38076,"journal":{"name":"European Journal of Radiology Open","volume":"14 ","pages":"Article 100638"},"PeriodicalIF":1.8,"publicationDate":"2025-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143420237","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-06-01Epub Date: 2024-12-17DOI: 10.1016/j.ejro.2024.100624
Yingqi Luo , Qingqi Yang , Jinglang Hu , Xiaowen Qin , Shengnan Jiang , Ying Liu
Objectives
To develop and validate a deep learning model using multimodal PET/CT imaging for detecting and classifying focal liver lesions (FLL).
Methods
This study included 185 patients who underwent 18F-FDG PET/CT imaging at our institution from March 2022 to February 2023. We analyzed serological data and imaging. Liver lesions were segmented on PET and CT, serving as the "reference standard". Deep learning models were trained using PET and CT images to generate predicted segmentations and classify lesion nature. Model performance was evaluated by comparing the predicted segmentations with the reference segmentations, using metrics such as Dice, Precision, Recall, F1-score, ROC, and AUC, and compared it with physician diagnoses.
Results
This study finally included 150 patients, comprising 46 patients with benign liver nodules, 51 patients with malignant liver nodules, and 53 patients with no FLLs. Significant differences were observed among groups for age, AST, ALP, GGT, AFP, CA19–9and CEA. On the validation set, the Dice coefficient of the model was 0.740. For the normal group, the recall was 0.918, precision was 0.904, F1-score was 0.909, and AUC was 0.976. For the benign group, the recall was 0.869, precision was 0.862, F1-score was 0.863, and AUC was 0.928. For the malignant group, the recall was 0.858, precision was 0.914, F1-score was 0.883, and AUC was 0.979. The model's overall diagnostic performance was between that of junior and senior physician.
Conclusion
This deep learning model demonstrated high sensitivity in detecting FLLs and effectively differentiated between benign and malignant lesions.
{"title":"Preliminary study on detection and diagnosis of focal liver lesions based on a deep learning model using multimodal PET/CT images","authors":"Yingqi Luo , Qingqi Yang , Jinglang Hu , Xiaowen Qin , Shengnan Jiang , Ying Liu","doi":"10.1016/j.ejro.2024.100624","DOIUrl":"10.1016/j.ejro.2024.100624","url":null,"abstract":"<div><h3>Objectives</h3><div>To develop and validate a deep learning model using multimodal PET/CT imaging for detecting and classifying focal liver lesions (FLL).</div></div><div><h3>Methods</h3><div>This study included 185 patients who underwent <sup>18</sup>F-FDG PET/CT imaging at our institution from March 2022 to February 2023. We analyzed serological data and imaging. Liver lesions were segmented on PET and CT, serving as the \"reference standard\". Deep learning models were trained using PET and CT images to generate predicted segmentations and classify lesion nature. Model performance was evaluated by comparing the predicted segmentations with the reference segmentations, using metrics such as Dice, Precision, Recall, F1-score, ROC, and AUC, and compared it with physician diagnoses.</div></div><div><h3>Results</h3><div>This study finally included 150 patients, comprising 46 patients with benign liver nodules, 51 patients with malignant liver nodules, and 53 patients with no FLLs. Significant differences were observed among groups for age, AST, ALP, GGT, AFP, CA19–9and CEA. On the validation set, the Dice coefficient of the model was 0.740. For the normal group, the recall was 0.918, precision was 0.904, F1-score was 0.909, and AUC was 0.976. For the benign group, the recall was 0.869, precision was 0.862, F1-score was 0.863, and AUC was 0.928. For the malignant group, the recall was 0.858, precision was 0.914, F1-score was 0.883, and AUC was 0.979. The model's overall diagnostic performance was between that of junior and senior physician.</div></div><div><h3>Conclusion</h3><div>This deep learning model demonstrated high sensitivity in detecting FLLs and effectively differentiated between benign and malignant lesions.</div></div>","PeriodicalId":38076,"journal":{"name":"European Journal of Radiology Open","volume":"14 ","pages":"Article 100624"},"PeriodicalIF":1.8,"publicationDate":"2025-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11720101/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142972520","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-06-01Epub Date: 2025-01-21DOI: 10.1016/j.ejro.2025.100635
Antonia M. Pausch , Vivien Filleböck , Clara Elsner , Niels J. Rupp , Daniel Eberli , Andreas M. Hötker
Purpose
To compare the diagnostic performance and image quality of a deep-learning-assisted ultra-fast biparametric MRI (bpMRI) with the conventional multiparametric MRI (mpMRI) for the diagnosis of clinically significant prostate cancer (csPCa).
Methods
This prospective single-center study enrolled 123 biopsy-naïve patients undergoing conventional mpMRI and additionally ultra-fast bpMRI at 3 T between 06/2023–02/2024. Two radiologists (R1: 4 years and R2: 3 years of experience) independently assigned PI-RADS scores (PI-RADS v2.1) and assessed image quality (mPI-QUAL score) in two blinded study readouts. Weighted Cohen’s Kappa (κ) was calculated to evaluate inter-reader agreement. Diagnostic performance was analyzed using clinical data and histopathological results from clinically indicated biopsies.
Results
Inter-reader agreement was good for both mpMRI (κ = 0.83) and ultra-fast bpMRI (κ = 0.87). Both readers demonstrated high sensitivity (≥94 %/≥91 %, R1/R2) and NPV (≥96 %/≥95 %) for csPCa detection using both protocols. The more experienced reader mostly showed notably higher specificity (≥77 %/≥53 %), PPV (≥62 %/≥45 %), and diagnostic accuracy (≥82 %/≥65 %) compared to the less experienced reader. There was no significant difference in the diagnostic performance of correctly identifying csPCa between both protocols (p > 0.05). The ultra-fast bpMRI protocol had significantly better image quality ratings (p < 0.001) and achieved a reduction in scan time of 80 % compared to conventional mpMRI.
Conclusion
Deep-learning-assisted ultra-fast bpMRI protocols offer a promising alternative to conventional mpMRI for diagnosing csPCa in biopsy-naïve patients with comparable inter-reader agreement and diagnostic performance at superior image quality. However, reader experience remains essential for diagnostic performance.
{"title":"Ultra-fast biparametric MRI in prostate cancer assessment: Diagnostic performance and image quality compared to conventional multiparametric MRI","authors":"Antonia M. Pausch , Vivien Filleböck , Clara Elsner , Niels J. Rupp , Daniel Eberli , Andreas M. Hötker","doi":"10.1016/j.ejro.2025.100635","DOIUrl":"10.1016/j.ejro.2025.100635","url":null,"abstract":"<div><h3>Purpose</h3><div>To compare the diagnostic performance and image quality of a deep-learning-assisted ultra-fast biparametric MRI (bpMRI) with the conventional multiparametric MRI (mpMRI) for the diagnosis of clinically significant prostate cancer (csPCa).</div></div><div><h3>Methods</h3><div>This prospective single-center study enrolled 123 biopsy-naïve patients undergoing conventional mpMRI and additionally ultra-fast bpMRI at 3 T between 06/2023–02/2024. Two radiologists (R1: 4 years and R2: 3 years of experience) independently assigned PI-RADS scores (PI-RADS v2.1) and assessed image quality (mPI-QUAL score) in two blinded study readouts. Weighted Cohen’s Kappa (κ) was calculated to evaluate inter-reader agreement. Diagnostic performance was analyzed using clinical data and histopathological results from clinically indicated biopsies.</div></div><div><h3>Results</h3><div>Inter-reader agreement was good for both mpMRI (κ = 0.83) and ultra-fast bpMRI (κ = 0.87). Both readers demonstrated high sensitivity (≥94 %/≥91 %, R1/R2) and NPV (≥96 %/≥95 %) for csPCa detection using both protocols. The more experienced reader mostly showed notably higher specificity (≥77 %/≥53 %), PPV (≥62 %/≥45 %), and diagnostic accuracy (≥82 %/≥65 %) compared to the less experienced reader. There was no significant difference in the diagnostic performance of correctly identifying csPCa between both protocols (p > 0.05). The ultra-fast bpMRI protocol had significantly better image quality ratings (p < 0.001) and achieved a reduction in scan time of 80 % compared to conventional mpMRI.</div></div><div><h3>Conclusion</h3><div>Deep-learning-assisted ultra-fast bpMRI protocols offer a promising alternative to conventional mpMRI for diagnosing csPCa in biopsy-naïve patients with comparable inter-reader agreement and diagnostic performance at superior image quality. However, reader experience remains essential for diagnostic performance.</div></div>","PeriodicalId":38076,"journal":{"name":"European Journal of Radiology Open","volume":"14 ","pages":"Article 100635"},"PeriodicalIF":1.8,"publicationDate":"2025-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143100624","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-06-01Epub Date: 2025-03-15DOI: 10.1016/j.ejro.2025.100643
Da Guo , Liping Liu , Yu Jin
<div><h3>Objectives</h3><div>This study aims to develop and validate a predictive nomogram for early recurrence in hepatocellular carcinoma (HCC), utilizing gadoxetic acid-enhanced MRI and intravoxel incoherent motion (IVIM) imaging to improve preoperative assessment and decision-making.</div></div><div><h3>Materials and methods</h3><div>From March 2018 and June 2022, a total of 245 patients with pathologically confirmed HCC, who underwent preoperative gadoxetic acid-enhanced MRI and IVIM, were retrospectively enrolled from two hospitals. These patients were divided into a training cohort (n = 160) and a validation cohort (n = 85). All patients were followed until death or the last follow-up date, with a minimum follow-up period of two years. Clinical indicators and pathologic information were compared between train cohort and validation cohort. Radiological features and diffusion parameters were compared between recurrence and non-recurrence groups using the chi-square test, Mann-Whitney U test and independent sample t test in training cohort. Univariate and multivariate analyses were performed to identify significant clinical-radiological variables associated with early recurrence in the training cohort. Based on these findings, a predictive nomogram integrating risk factors and diffusion parameters was developed. The predictive performance of the nomogram was evaluated in both the training and validation cohorts.</div></div><div><h3>Results</h3><div>No statistically significant difference in clinical and pathologic characteristics were observed between the training and validation cohorts. In training cohort, significant differences were identified between the recurrence and non-recurrence groups in tumor size, nodule-in-nodule architecture, mosaic architecture, non-smooth tumor margin, intratumor necrosis, satellite nodule, and peritumoral hypo-intensity in the hepatobiliary phase (HBP). The results of multivariate analysis identified tumor size (HR, 1.435; 95 % CI, 0.702–2.026; p < 0.05), mosaic architecture (HR, 0.790; 95 % CI, 0.421–1.480; p < 0.05), non-smooth tumor margin (HR, 1.775; 95 % CI, 0.941–3.273; p < 0.05), intratumor necrosis (HR, 1.414; 95 % CI, 0.807–2.476; p < 0.05), satellite nodule (HR, 0.648; 95 % CI, 0.352–1.191; p < 0.01), peritumoral hypo-intensity on HBP (HR, 2.786; 95 % CI, 1.141–6.802; p < 0.001) and D (HR, 0.658; 95 % CI,0.487–0.889; p < 0.01) were the independent risk factor for recurrence. The nomogram exhibited excellent predictive performance with C-index of 0.913 and 0.875 in the training cohort and validation cohort, respectively. Also, based on the nomogram score, the patients were classified according to risk factor and the Kaplan-Meier curve analysis also showed that the nomogram had a good predictive efficacy.</div></div><div><h3>Conclusion</h3><div>The nomogram, integrating radiological risk factors and diffusion parameters, offers a reliable tool for preoperative prediction of early
{"title":"Prediction early recurrence of hepatocellular carcinoma after hepatectomy using gadoxetic acid-enhanced MRI and IVIM","authors":"Da Guo , Liping Liu , Yu Jin","doi":"10.1016/j.ejro.2025.100643","DOIUrl":"10.1016/j.ejro.2025.100643","url":null,"abstract":"<div><h3>Objectives</h3><div>This study aims to develop and validate a predictive nomogram for early recurrence in hepatocellular carcinoma (HCC), utilizing gadoxetic acid-enhanced MRI and intravoxel incoherent motion (IVIM) imaging to improve preoperative assessment and decision-making.</div></div><div><h3>Materials and methods</h3><div>From March 2018 and June 2022, a total of 245 patients with pathologically confirmed HCC, who underwent preoperative gadoxetic acid-enhanced MRI and IVIM, were retrospectively enrolled from two hospitals. These patients were divided into a training cohort (n = 160) and a validation cohort (n = 85). All patients were followed until death or the last follow-up date, with a minimum follow-up period of two years. Clinical indicators and pathologic information were compared between train cohort and validation cohort. Radiological features and diffusion parameters were compared between recurrence and non-recurrence groups using the chi-square test, Mann-Whitney U test and independent sample t test in training cohort. Univariate and multivariate analyses were performed to identify significant clinical-radiological variables associated with early recurrence in the training cohort. Based on these findings, a predictive nomogram integrating risk factors and diffusion parameters was developed. The predictive performance of the nomogram was evaluated in both the training and validation cohorts.</div></div><div><h3>Results</h3><div>No statistically significant difference in clinical and pathologic characteristics were observed between the training and validation cohorts. In training cohort, significant differences were identified between the recurrence and non-recurrence groups in tumor size, nodule-in-nodule architecture, mosaic architecture, non-smooth tumor margin, intratumor necrosis, satellite nodule, and peritumoral hypo-intensity in the hepatobiliary phase (HBP). The results of multivariate analysis identified tumor size (HR, 1.435; 95 % CI, 0.702–2.026; p < 0.05), mosaic architecture (HR, 0.790; 95 % CI, 0.421–1.480; p < 0.05), non-smooth tumor margin (HR, 1.775; 95 % CI, 0.941–3.273; p < 0.05), intratumor necrosis (HR, 1.414; 95 % CI, 0.807–2.476; p < 0.05), satellite nodule (HR, 0.648; 95 % CI, 0.352–1.191; p < 0.01), peritumoral hypo-intensity on HBP (HR, 2.786; 95 % CI, 1.141–6.802; p < 0.001) and D (HR, 0.658; 95 % CI,0.487–0.889; p < 0.01) were the independent risk factor for recurrence. The nomogram exhibited excellent predictive performance with C-index of 0.913 and 0.875 in the training cohort and validation cohort, respectively. Also, based on the nomogram score, the patients were classified according to risk factor and the Kaplan-Meier curve analysis also showed that the nomogram had a good predictive efficacy.</div></div><div><h3>Conclusion</h3><div>The nomogram, integrating radiological risk factors and diffusion parameters, offers a reliable tool for preoperative prediction of early ","PeriodicalId":38076,"journal":{"name":"European Journal of Radiology Open","volume":"14 ","pages":"Article 100643"},"PeriodicalIF":1.8,"publicationDate":"2025-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143631869","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-06-01Epub Date: 2025-05-08DOI: 10.1016/j.ejro.2025.100658
Susan V. van Hees , Martin B. Schilder , Alexandra Keyser , Alessandro Sbrizzi , Jordi P.D. Kleinloog , Wouter P.C. Boon
Purpose
MRI waitlists and discomfort from long scanning sessions are significant problems in clinical radiology. Novel multiparametric quantitative MRI techniques (qMRI) for radiological imaging enable acquisition of full-brain data within minutes to address these problems. While technical and clinical work is advancing, there has been limited research on implementing fast qMRI. This paper aims to identify implementation factors and scenarios within a healthcare setting facing rising demand, staff shortages, and limited capacity of MRI systems.
Methods
The paper reports on data collected using qualitative methods: 1) Interviews and guided discussions, 2) co-creation workshop. Both steps involved key representatives with various backgrounds and expertise, such as radiologists, lab technicians, insurers, and patients.
Results
Workshop participants visualised current and future workflows, which helped articulate implementation factors for qMRI. Supply and demand in MRI will change with increased accessibility and shortened timeslots. Three implementation scenarios came forward: 1) stable deployment, 2) extension to conducting more complex diagnostic exams, and 3) (more) preventive screening.
Discussion and conclusions
This paper demonstrates challenges, solutions, and opportunities for successfully implementing fast qMRI in the clinic, and five lessons for adoption in the clinic: 1) importance of balancing perfectionism with confidence when it comes to clinicians’ expectations, 2) good use of Artificial Intelligence, 3) considering a learning curve associated with implementation, 4) regarding competing technologies, and 5) including patients’ experiences. Future research should investigate salient issues regarding future of AI in radiology and for moving imaging practices out of the clinic.
{"title":"Exploring scenarios for implementing fast quantitative MRI","authors":"Susan V. van Hees , Martin B. Schilder , Alexandra Keyser , Alessandro Sbrizzi , Jordi P.D. Kleinloog , Wouter P.C. Boon","doi":"10.1016/j.ejro.2025.100658","DOIUrl":"10.1016/j.ejro.2025.100658","url":null,"abstract":"<div><h3>Purpose</h3><div>MRI waitlists and discomfort from long scanning sessions are significant problems in clinical radiology. Novel multiparametric quantitative MRI techniques (qMRI) for radiological imaging enable acquisition of full-brain data within minutes to address these problems. While technical and clinical work is advancing, there has been limited research on implementing fast qMRI. This paper aims to identify implementation factors and scenarios within a healthcare setting facing rising demand, staff shortages, and limited capacity of MRI systems.</div></div><div><h3>Methods</h3><div>The paper reports on data collected using qualitative methods: 1) Interviews and guided discussions, 2) co-creation workshop. Both steps involved key representatives with various backgrounds and expertise, such as radiologists, lab technicians, insurers, and patients.</div></div><div><h3>Results</h3><div>Workshop participants visualised current and future workflows, which helped articulate implementation factors for qMRI. Supply and demand in MRI will change with increased accessibility and shortened timeslots. Three implementation scenarios came forward: 1) stable deployment, 2) extension to conducting more complex diagnostic exams, and 3) (more) preventive screening.</div></div><div><h3>Discussion and conclusions</h3><div>This paper demonstrates challenges, solutions, and opportunities for successfully implementing fast qMRI in the clinic, and five lessons for adoption in the clinic: 1) importance of balancing perfectionism with confidence when it comes to clinicians’ expectations, 2) good use of Artificial Intelligence, 3) considering a learning curve associated with implementation, 4) regarding competing technologies, and 5) including patients’ experiences. Future research should investigate salient issues regarding future of AI in radiology and for moving imaging practices out of the clinic.</div></div>","PeriodicalId":38076,"journal":{"name":"European Journal of Radiology Open","volume":"14 ","pages":"Article 100658"},"PeriodicalIF":1.8,"publicationDate":"2025-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143918099","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号