Tomography, Volume 11, Issue 8 (August 2025) – 12 articles

Background: Accurate preoperative tumor sizing is critical for optimal surgical planning in breast cancer. Contrast-enhanced mammography (CEM) has emerged as a promising modality, yet its accuracy relative to conventional imaging and pathology requires further validation. Objective: To prospectively evaluate the dimensional accuracy and reproducibility of CEM compared to mammography and ultrasound, using surgical pathology as the reference standard. Methods: A total of 205 patients with 267 breast lesions underwent preoperative CEM, mammography, and ultrasound. Tumor sizes were measured independently by two radiologists. Accuracy was assessed via mean absolute error (MAE), Pearson and Spearman correlations, and inter-reader agreement evaluated by intraclass correlation coefficient (ICC) and Gwet’s AC1. Sensitivity analyses included bootstrap confidence intervals and log-transformed data. The surgical impact of additional lesions detected by CEM was also analyzed. Results: CEM showed superior accuracy with a mean absolute error of 0.46 mm (95% CI: 0.24–0.68) compared to mammography (4.06 mm) and ultrasound (3.52 mm) (p < 0.00001). Pearson’s correlation between CEM and pathology was exceptionally high (r = 0.995; 95% CI: 0.994–0.996), with similar robustness after log transformation. Inter-reader agreement for CEM was excellent (ICC 0.93; Gwet’s AC1 ~0.96, 95% CI: 0.93–0.98). CEM detected additional lesions in 13.1% of patients, leading to altered surgical management in 6.4%. Background parenchymal enhancement was independently associated with measurement error. Conclusions: CEM provides highly accurate and reproducible tumor size estimation superior to conventional imaging modalities, with potential clinical impact through detection of additional lesions. Its ability to detect additional lesions not seen on mammography or ultrasound has direct implications for surgical decision making, with the potential to reduce reoperations and improve oncologic and cosmetic outcomes. However, high correlation values and selective patient cohorts warrant cautious interpretation. Further multicenter studies are needed to confirm these findings and define CEM’s role in clinical practice. Full article

Objectives: This study aimed to investigate the consistency of lesion identification by Prostate Imaging Reporting and Data System (PI-RADS) and the related clinical and histological characteristics in a high-volume tertiary care center. Materials and methods: The analysis used real-world data from 111 patients between 2018 and 2022. Each patient underwent two multiparametric magnetic resonance imaging (MRI) scans of the prostate at different institutions with a median interval of 42 days between the scans, followed by an MRI-fused biopsy conducted 7 days after the second MRI. Results: The PI-RADS classifications assigned to the index lesions in the in-house prostate MRI were as follows: PI-RADS V, 33.3% (n = 37); PI-RADS IV, 49.5% (n = 55); PI-RADS III, 12.6% (n = 14); and PI-RADS II, 4.5% (n = 5). Cancer detection rates for randomized and/or targeted biopsies were 91.9% (n = 34) for PI-RADS V, 65.5% (n = 36) for PI-RADS IV, 21.4% (n = 3) for PI-RADS III, and 20% (n = 1) for PI-RADS II. Overall, malignant histology was observed in 64.9% (n = 72) of the targeted lesions and 57.7% (n = 64) of the randomized biopsies. In the first performed, external MRI, 18% (n = 20) and 10.8% (n = 12) of the patients were classified in the higher and lower PI-RADS categories, respectively. The biopsy plan was adjusted for 57 patients (51.4%); nevertheless, any cancer could have possibly been identified regardless of the adjustments. Conclusion: The 6-week interval between the MRI scans did not affect the quality of the biopsy results significantly. Full article

Background: Accurate subtype identification of lymphoma cancer is crucial for effective diagnosis and treatment planning. Although standard deep learning algorithms have demonstrated robustness, they are still prone to overfitting and limited generalization, necessitating more reliable and robust methods. Objectives: This study presents an autoencoder-augmented stacked ensemble learning (SEL) framework integrating deep feature extraction (DFE) and ensembles of machine learning classifiers to improve lymphoma subtype identification. Methods: Convolutional autoencoder (CAE) was utilized to obtain high-level feature representations of histopathological images, followed by dimensionality reduction via Principal Component Analysis (PCA). Various models were utilized for classifying extracted features, i.e., Random Forest (RF), Support Vector Machine (SVM), Multi-Layer Perceptron (MLP), AdaBoost, and Extra Trees classifiers. A Gradient Boosting Machine (GBM) meta-classifier was utilized in an SEL approach to further fine-tune final predictions. Results: All the models were tested using accuracy, area under the curve (AUC), and Average Precision (AP) metrics. The stacked ensemble classifier performed better than all the individual models with a 99.04% accuracy, 0.9998 AUC, and 0.9996 AP, far exceeding what regular deep learning (DL) methods would achieve. Of standalone classifiers, MLP (97.71% accuracy, 0.9986 AUC, 0.9973 AP) and Random Forest (96.71% accuracy, 0.9977 AUC, 0.9953 AP) provided the best prediction performance, while AdaBoost was the poorest performer (68.25% accuracy, 0.8194 AUC, 0.6424 AP). PCA and t-SNE plots confirmed that DFE effectively enhances class discrimination. Conclusion: This study demonstrates a highly accurate and reliable approach to lymphoma classification by using autoencoder-assisted ensemble learning, reducing the misclassification rate and significantly enhancing the accuracy of diagnosis. AI-based models are designed to assist pathologists by providing interpretable outputs such as class probabilities and visualizations (e.g., Grad-CAM), enabling them to understand and validate predictions in the diagnostic workflow. Future studies should enhance computational efficacy and conduct multi-centre validation studies to confirm the model’s generalizability on extensive collections of histopathological datasets. Full article

Background/Objectives: Accurate prediction of pediatric appendicitis diagnosis, management, and severity is critical for clinical decision-making. We aimed to evaluate the predictive performance of a wide range of machine learning models, combined with various feature selection techniques, on a pediatric appendicitis dataset. A particular focus was placed on the role of ultrasound (US) image-descriptive features in model performance and explainability. Methods: We conducted a retrospective cohort study on a dataset of 781 pediatric patients aged 0–18 presenting to Children’s Hospital St. Hedwig in Regensburg, Germany, between January 2016 and February 2023. We developed and validated predictive models; machine learning algorithms included the random forest, logistic regression, stochastic gradient descent, and the light gradient boosting machine (LGBM). These were paired exhaustively with feature selection methods spanning filter-based (association and prediction), embedded (LGBM and linear), and a novel redundancy-aware step-up wrapper approach. We employed a machine learning benchmarking study design where AI models were trained to predict diagnosis, management, and severity outcomes, both with and without US image-descriptive features, and evaluated on held-out testing samples. Model performance was assessed using overall accuracy and area under the receiver operating characteristic curve (AUROC). A deep learner optimized for tabular data, GANDALF, was also evaluated in these applications. Results: US features significantly improved diagnostic accuracy, supporting their use in reducing model bias. However, they were not essential for maximizing accuracy in predicting management or severity. In summary, our best-performing models were, for diagnosis, the random forest with embedded LGBM feature selection (98.1% accuracy, AUROC: 0.993), for management, the random forest without feature selection (93.9% accuracy, AUROC: 0.980), and for severity, the LGBM with filter-based association feature selection (90.1% accuracy, AUROC: 0.931). Conclusions: Our results demonstrate that high-performing, interpretable machine learning models can predict key clinical outcomes in pediatric appendicitis. US image features improve diagnostic accuracy but are not critical for predicting management or severity. Full article

Background/Objectives: Fat and inflammation confound current magnetic resonance imaging (MRI) methods for assessing fibrosis in liver disease. Sodium or amide proton transfer-weighted MRI methods may be more specific for assessing liver fibrosis. The purpose of this study was to determine the feasibility of sodium and amide proton transfer-weighted MRI in individuals with liver disease and to determine if either method correlated with clinical markers of fibrosis. Methods: T₁ and T₂ relaxation maps, proton density fat fraction maps, liver shear stiffness maps, amide proton transfer-weighted (APTw) images, and sodium images were acquired at 3T. Image data were extracted from regions of interest placed in the liver. ANOVA tests were run with disease status, age, and body mass index as independent factors; significance was set to p < 0.05. Post-hoc t-tests were run when the ANOVA showed significance. Results: A total of 36 participants were enrolled, 34 of whom were included in the final APTw analysis and 24 in the sodium analysis. Estimated liver tissue sodium concentration differentiated participants with liver disease from those without, whereas amide proton transfer-weighted MRI did not. Estimated liver tissue sodium concentration negatively correlated with the Fibrosis-4 score, but amide proton transfer-weighted MRI did not correlate with any clinical marker of disease. Conclusions: Amide proton-weighted imaging was not different between groups. Estimated liver tissue sodium concentrations did differ between groups but did not provide additional information over conventional methods. Full article

Objectives: This study investigates whether scan time in the high-resolution magnetic resonance imaging (MRI) of human embryos can be reduced without compromising spatial resolution by applying zero-shot self-supervised learning (ZS-SSL), a deep-learning-based reconstruction method. Methods: Simulations using a numerical phantom were conducted to evaluate spatial resolution across various acceleration factors (AF = 2, 4, 6, and 8) and signal-to-noise ratio (SNR) levels. Resolution was quantified using a blur-based estimation method based on the Sparrow criterion. ZS-SSL was compared to conventional compressed sensing (CS). Experimental imaging of a human embryo at Carnegie stage 21 was performed at a spatial resolution of (30 μm)³ using both retrospective and prospective undersampling at AF = 4 and 8. Results: ZS-SSL preserved spatial resolution more effectively than CS at low SNRs. At AF = 4, image quality was comparable to that of fully sampled data, while noticeable degradation occurred at AF = 8. Experimental validation confirmed these findings, with clear visualization of anatomical structures—such as the accessory nerve—at AF = 4; there was reduced structural clarity at AF = 8. Conclusions: ZS-SSL enables significant scan time reduction in high-resolution MRI of human embryos while maintaining spatial resolution at AF = 4, assuming an SNR above approximately 15. This trade-off between acceleration and image quality is particularly beneficial in studies with limited imaging time or specimen availability. The method facilitates the efficient acquisition of ultra-high-resolution data and supports future efforts to construct detailed developmental atlases. Full article

Background/Objectives: There is significant interest in using texture features to extract hidden image-based information. In medical imaging applications using radiomics, AI, or personalized medicine, the quest is to extract patient or disease specific information while being insensitive to other system or processing variables. While we use digital breast tomosynthesis (DBT) to show these effects, our results would be generally applicable to a wider range of other imaging modalities and applications. Methods: We examine factors in texture estimation methods, such as quantization, pixel distance offset, and region of interest (ROI) size, that influence the magnitudes of these readily computable and widely used image texture features (specifically Haralick’s gray level co-occurrence matrix (GLCM) textural features). Results: Our results indicate that quantization is the most influential of these parameters, as it controls the size of the GLCM and range of values. We propose a new multi-resolution normalization (by either fixing ROI size or pixel offset) that can significantly reduce quantization magnitude disparities. We show reduction in mean differences in feature values by orders of magnitude; for example, reducing it to 7.34% between quantizations of 8–128, while preserving trends. Conclusions: When combining images from multiple vendors in a common analysis, large variations in texture magnitudes can arise due to differences in post-processing methods like filters. We show that significant changes in GLCM magnitude variations may arise simply due to the filter type or strength. These trends can also vary based on estimation variables (like offset distance or ROI) that can further complicate analysis and robustness. We show pathways to reduce sensitivity to such variations due to estimation methods while increasing the desired sensitivity to patient-specific information such as breast density. Finally, we show that our results obtained from simulated DBT images are consistent with what we see when applied to clinical DBT images. Full article

Background: Despite the growing demand for amyloid PET quantification, practical challenges remain. As automated software platforms are increasingly adopted to address these limitations, we evaluated the reliability of commercial tools for Centiloid quantification against the original Centiloid Project method. Methods: This retrospective study included 332 amyloid PET scans (165 [¹⁸F]Florbetaben; 167 [¹⁸F]Flutemetamol) performed for suspected mild cognitive impairments or dementia, paired with T1-weighted MRI within one year. Centiloid values were calculated using three automated software platforms, BTXBrain, MIMneuro, and SCALE PET, and compared with the original Centiloid method. The agreement was assessed using Pearson’s correlation coefficient, the intraclass correlation coefficient (ICC), a Passing–Bablok regression, and Bland–Altman plots. The concordance with the visual interpretation was evaluated using receiver operating characteristic (ROC) curves. Results: BTXBrain (R = 0.993; ICC = 0.986) and SCALE PET (R = 0.992; ICC = 0.991) demonstrated an excellent correlation with the reference, while MIMneuro showed a slightly lower agreement (R = 0.974; ICC = 0.966). BTXBrain exhibited a proportional underestimation (slope = 0.872 [0.860–0.885]), MIMneuro showed a significant overestimation (slope = 1.053 [1.026–1.081]), and SCALE PET demonstrated a minimal bias (slope = 1.014 [0.999–1.029]). The bias pattern was particularly noted for FMM. All platforms maintained their trends for correlations and biases when focusing on subthreshold-to-low-positive ranges (0–50 Centiloid units). However, all platforms showed an excellent agreement with the visual interpretation (areas under ROC curves > 0.996 for all). Conclusions: Three automated platforms demonstrated an acceptable reliability for Centiloid quantification, although software-specific biases were observed. These differences did not impair their feasibility in aiding the image interpretation, as supported by the concordance with visual readings. Nevertheless, users should recognize the platform-specific characteristics when applying diagnostic thresholds or interpreting longitudinal changes. Full article

Advances in plaque imaging have transformed cardiovascular diagnostics through detailed characterization of atherosclerotic plaques beyond traditional stenosis assessment. This review outlines the clinical applications of varying modalities, including dual-layer spectral CT, photon-counting CT, dual-energy CT, and CT-derived fractional flow reserve (CT-FFR). These technologies offer improved spatial resolution, tissue differentiation, and functional assessment of coronary lesions. Additionally, artificial intelligence has emerged as a powerful tool to automate plaque detection, quantify burden, and refine risk prediction. Collectively, these innovations provide a more comprehensive approach to coronary artery disease evaluation and support personalized management strategies. Full article

Background/Objectives: X-ray computed tomography (XCT) is a powerful tool for volumetric imaging, where three-dimensional (3D) images are generated from a large number of individual X-ray projection images. However, collecting the required number of low-noise projection images is time-consuming, limiting its applicability to scenarios requiring high temporal resolution, such as the study of dynamic processes. Inspired by stereo vision, we previously developed stereo X-ray imaging methods that operate with only two X-ray projections, enabling the 3D reconstruction of point and line fiducial markers at significantly faster temporal resolutions. Methods: Building on our prior work, this paper demonstrates the use of stereo X-ray techniques for 3D reconstruction of sharp object corners, eliminating the need for internal fiducial markers. This is particularly relevant for deformation measurement of manufactured components under load. Additionally, we explore model training using synthetic data when annotated real data is unavailable. Results: We show that the proposed method can reliably reconstruct sharp corners in 3D using only two X-ray projections. The results confirm the method’s applicability to real-world stereo X-ray images without relying on annotated real training datasets. Conclusions: Our approach enables stereo X-ray 3D reconstruction using synthetic training data that mimics key characteristics of real data, thereby expanding the method’s applicability in scenarios with limited training resources. Full article

Targeted radioligand therapy (RLT) is an emerging field in anticancer therapeutics with great potential across tumor types and stages of disease. While much progress has focused on agents targeting somatostatin receptors and prostate-specific membrane antigen (PSMA), the same advanced radioconjugation methods and molecular targeting have spurred the development of numerous theranostic combinations for other targets. A number of the most promising agents have progressed to clinical trials and are poised to change the landscape of positron emission tomography (PET) imaging. Here, we present recent data on some of the most important emerging molecular targeted agents with their exemplar clinical images, including agents targeting fibroblast activation protein (FAP), hypoxia markers, gastrin-releasing peptide receptors (GRPrs), and integrins. These radiopharmaceuticals share the promising characteristic of being able to image multiple types of cancer. Early clinical trials have already demonstrated superiority to ¹⁸F-fluorodeoxyglucose (¹⁸F-FDG) for some, suggesting the potential to supplant this longstanding PET radiotracer. Here, we provide a primer for practicing radiologists, particularly nuclear medicine clinicians, to understand novel PET imaging agents and their clinical applications, as well as the availability of companion targeted radiotherapeutics, the status of their regulatory approval, the potential challenges associated with their use, and the future opportunities and perspectives. Full article

Background/Objectives: Myelofibrosis (MF) is a myeloproliferative neoplasm characterized by the replacement of healthy bone marrow (BM) with malignant and fibrotic tissue. In a healthy state, bone marrow is composed of approximately 60–70% fat cells, which are replaced as disease progresses. Proton density fat fraction (PDFF), a non-invasive and quantitative MRI metric, enables analysis of BM architecture by measuring the percentage of fat versus cells in the environment. Our objective is to investigate variance in quantitative PDFF-MRI values over time as a marker of disease progression and response to treatment. Methods: We analyzed existing data from three cohorts of mice: two groups with MF that failed to respond to therapy with approved drugs for MF (ruxolitinib, fedratinib), investigational compounds (navitoclax, balixafortide), or vehicle and monitored over time by MRI; the third group consisted of healthy controls imaged at a single time point. Using in-house MATLAB programs, we performed a voxel-wise analysis of PDFF values in lower extremity bone marrow, specifically comparing the variance of each voxel within and among mice. Results: Our findings revealed a significant difference in PDFF values between healthy and diseased BM. With progressive disease non-responsive to therapy, the expansion of hematopoietic cells in BM nearly completely replaced normal fat, as determined by a markedly reduced PDFF and notable reduction in the variance in PDFF values in bone marrow over time. Conclusions: This study validated our hypothesis that the variance in PDFF in BM decreases with disease progression, indicating pathologic expansion of hematopoietic cells. We can conclude that disease progression can be tracked by a decrease in PDFF values. Analyzing variance in PDFF may improve the assessment of disease progression in pre-clinical models and ultimately patients with MF. Full article