Tomography

Background: The discovery of novel antibody–drug conjugates for low-expression human epidermal growth factor receptor 2 (HER2-low) breast cancer highlights the inadequacy of the conventional binary classification of HER2 status as either negative or positive. Identification of HER2-low breast cancer is crucial for selecting patients who may benefit from targeted therapies. This study aims to determine whether qualitative and quantitative magnetic resonance imaging (MRI) features can effectively reflect low-HER2-expression breast cancer. Methods: Pre-treatment breast MRI images from 232 patients with pathologically confirmed breast cancer were retrospectively analyzed. Both clinicopathologic and MRI features were recorded. Qualitative MRI features included Breast Imaging Reporting and Data System (BI-RADS) descriptors from dynamic contrast-enhanced MRI (DCE-MRI), as well as intratumoral T2 hyperintensity and peritumoral edema observed in T2-weighted imaging (T2WI). Quantitative features were derived from diffusion kurtosis imaging (DKI) using multiple b-values and included statistics such as mean, median, 5th and 95th percentiles, skewness, kurtosis, and entropy from apparent diffusion coefficient (ADC), D_app, and K_app histograms. Differences in clinicopathologic, qualitative, and quantitative MRI features were compared across groups, with multivariable logistic regression used to identify significant independent predictors of HER2-low breast cancer. The discriminative power of MRI features was assessed using receiver operating characteristic (ROC) curves. Results: HER2 status was categorized as HER2-zero (n = 60), HER2-low (n = 91), and HER2-overexpressed (n = 81). Clinically, estrogen receptor (ER), progesterone receptor (PR), hormone receptor (HR), and Ki-67 levels significantly differed between the HER2-low group and others (all p < 0.001). In MRI analyses, intratumoral T2 hyperintensity was more prevalent in HER2-low cases (p = 0.009, p = 0.008). Mass lesions were more common in the HER2-zero group than in the HER2-low group (p = 0.038), and mass shape (p < 0.001) and margin (p < 0.001) significantly varied between the HER2 groups, with mass shape emerging as an independent predictive factor (HER2-low vs. HER2-zero: p = 0.010, HER2-low vs. HER2-over: p = 0.012). Qualitative MRI features demonstrated an area under the curve (AUC) of 0.763 (95% confidence interval [CI]: 0.667–0.859) for distinguishing HER2-low from HER2-zero status. Quantitative features showed distinct differences between HER2-low and HER2-overexpression groups, particularly in non-mass enhancement (NME) lesions. Combined variables achieved the highest predictive accuracy for HER2-low status, with an AUC of 0.802 (95% CI: 0.701–0.903). Conclusions: Qualitative and quantitative MRI features offer valuable insights into low-HER2-expression breast cancer. While qualitative features are more effective for mass lesions, quantitative features are more suitable for NME lesions. These findings provide a more accessible and cost-effective approach to noninvasively identifying patients who may benefit from targeted therapy. Full article

Background/Objectives: While ancillary tests for brain death diagnosis are not routinely recommended in guidelines, they may be necessary in specific clinical scenarios. Computed tomography angiography (CTA) is particularly advantageous in pediatric patients due to its noninvasive nature, accessibility, and rapid provision of anatomical information. This study aims to assess the diagnostic sensitivity of a revised venous system (ICV-SPV) utilizing a 4-point scoring system in children clinically diagnosed with brain death. Materials and Methods: A total of 43 pediatric patients clinically diagnosed with brain death who underwent CTA were retrospectively analyzed. Imaging was performed using a standardized brain death protocol. Three distinct 4-point scoring systems (A20-V60, A60-V60, ICV-SPV) were utilized to assess vessel opacification in different imaging phases. To evaluate age-dependent sensitivity, patients were categorized into three age groups: 26 days–1 year, 2–6 years, and 6–18 years. The sensitivity of each 4-point scoring system in diagnosing brain death was calculated for all age groups. Results: The revised venous scoring system (ICV-SPV) demonstrated the highest overall sensitivity in confirming brain death across all age groups, significantly outperforming the reference 4-point scoring systems. Furthermore, the ICV-SPV system exhibited the greatest sensitivity in patients with cranial defects. Conclusions: The revised 4-point venous CTA scoring system, which relies on the absence of ICV and SPV opacification, is a reliable tool for confirming cerebral circulatory arrest in pediatric patients with clinical brain death. Full article

Objectives: Prostate-specific membrane antigen (PSMA)-targeted radioligands are promising diagnostic tools for the targeted positron emission tomography (PET) imaging of prostate cancer (PCa). In present work, we aimed to develop a novel PSMA tracer to provide an additional option for prostate cancer diagnosis. Methods: Our team designed a new structure of the PSMA tracer and evaluated it with cellular experiments in vitro to preliminarily verify the targeting and specificity of ⁶⁸Ga-NOTA-GC-PSMA. PET/CT imaging of PSMA-positive xenograft-bearing models in vivo to further validate the in vivo specificity and targeting of the radiotracer. Pathological tissue sections from prostate cancer patients were compared with pathological immunohistochemistry and pathological tissue staining results by radioautography experiments to assess the targeting-PSMA of ⁶⁸Ga-NOTA-GC-PSMA on human prostate cancer pathological tissues. Results: The novel tracer showed high hydrophilicity and rapid clearance rate. Specific cell binding and micro-PET imaging experiments showed that ⁶⁸Ga-NOTA-GC-PSMA displayed a high specific LNCaP tumor cell uptake (1.70% ± 0.13% at 120 min) and tumor-to-muscle (T/M) and tumor-to-kidney (T/K) ratio (13.87 ± 11.20 and 0.20 ± 0.08 at 60 min, respectively). Conclusions: The novel tracer ⁶⁸Ga-NOTA-GC-PSMA is promising radionuclide imaging of PCa. Full article

Calcified lung nodules are frequently encountered on chest imaging, often as incidental findings. While calcifications are typically associated with benign conditions, they do not inherently exclude malignancy, making accurate differentiation essential. The primary diagnostic challenge lies in distinguishing benign from malignant nodules based solely on imaging features. Various calcification patterns, including diffuse, popcorn, lamellated and eccentric, provide important diagnostic clues, though overlap among different conditions may persist. A comprehensive diagnostic approach integrates clinical history with multimodal imaging, including magnetic resonance and nuclear medicine, when necessary, to improve accuracy. When imaging findings remain inconclusive, tissue sampling through biopsy may be required for definitive characterization. This review provides an overview of the imaging features of calcified lung nodules, emphasizing key diagnostic challenges and their clinical implications. Full article

Background: Temporomandibular joint (TMJ) disorders are a significant cause of orofacial pain. Artificial intelligence (AI) has been successfully applied to other imaging modalities but remains underexplored in ultrasonographic evaluations of TMJ. Objective: This study aimed to develop and validate an AI-driven method for the automatic and reproducible measurement of TMJ space width from ultrasonographic images. Methods: A total of 142 TMJ ultrasonographic images were segmented into three anatomical components: the mandibular condyle, joint space, and glenoid fossa. State-of-the-art architectures were tested, and the best-performing 2D Residual U-Net was trained and validated against expert annotations. The algorithm for joint space width measurement based on TMJ segmentation was proposed, calculating the vertical distance between the superior-most point of the mandibular condyle and its corresponding point on the glenoid fossa. Results: The segmentation model achieved high performance for the mandibular condyle (Dice: 0.91 ± 0.08) and joint space (Dice: 0.86 ± 0.09), with notably lower performance for the glenoid fossa (Dice: 0.60 ± 0.24), highlighting variability due to its complex geometry. The TMJ space width measurement algorithm demonstrated minimal bias, with a mean difference of 0.08 mm and a mean absolute error of 0.18 mm compared to reference measurements. Conclusions: The model exhibited potential as a reliable tool for clinical use, demonstrating accuracy in TMJ ultrasonographic analysis. This study underscores the ability of AI-driven segmentation and measurement algorithms to bridge existing gaps in ultrasonographic imaging and lays the foundation for broader clinical applications. Full article

Background: Functional tumor volume (FTV), measured from dynamic contrast-enhanced MRI, is an imaging biomarker that can predict treatment response in breast cancer patients undergoing neoadjuvant chemotherapy (NAC). The FTV-based predictive model, combined with core biopsy, informed treatment decisions of recommending patients with excellent responses to proceed to surgery early in a large NAC clinical trial. Methods: In this retrospective study, we constructed models using FTV measurements. We analyzed performance tradeoffs when a probability threshold was used to identify excellent responders through the prediction of pathology complete response (pCR). Individual models were developed within cohorts defined by the hormone receptor and human epidermal growth factor receptor 2 (HR/HER2) subtype. Results: A total of 814 patients enrolled in the I-SPY 2 trial between 2010 and 2016 were included with a mean age of 49 years (range: 24 to 77). Among these patients, 289 (36%) achieved pCR. The area under the ROC curve (AUC) ranged from 0.68 to 0.74 for individual HR/HER2 subtypes. When probability thresholds were chosen based on minimum positive predictive value (PPV) levels of 50%, 70%, and 90%, the PPV-sensitivity tradeoff varied among subtypes. The highest sensitivities (100%, 87%, 45%) were found in the HR−/HER2+ sub-cohort for probability thresholds of 0, 0.62, and 0.72; followed by the triple-negative sub-cohort (98%, 52%, 4%) at thresholds of 0.13, 0.58, and 0.67; and HR+/HER2+ (78%, 16%, 8%) at thresholds of 0.34, 0.57, and 0.60. The lowest sensitivities (20%, 0%, 0%) occurred in the HR+/HER2− sub-cohort. Conclusions: Predictive models developed using imaging biomarkers, alongside clinically validated probability thresholds, can be incorporated into decision-making for precision oncology. Full article

Background/Objectives: The accurate detection of microcalcifications in mammograms is critical for the early detection of breast cancer. However, the variability between different manufacturers is significant, particularly with digital breast tomosynthesis (DBT). Manufacturers have many design differences, including sweep angles, detector types, reconstruction techniques, filters, and focal spot construction. This study outlined the development of an innovative phantom model using crystallizations to improve the accuracy of imaging microcalcifications in DBT. The goal of these models was to achieve consistent evaluations, thereby reducing the variability between different scanners. Methods: We created a novel phantom model that simulates different types of breast tissue densities with calcifications. Furthermore, these crystalline-grown phantoms can more accurately represent the physiological shapes and compositions of microcalcifications than do other available phantoms for calcifications and can be evaluated on different systems. Microcalcification patterns were generated using the evaporation of sodium chloride, transplantation of calcium carbonate crystals, and/or injection of hydroxyapatite. These patterns were embedded in multiple layers within the wax to simulate various depths and distributions of calcifications with the ability to generate a large variety of patterns. Results: The tomosynthesis imaging revealed phantoms that utilized calcium carbonate crystals showed demonstrable visualization differences between the 3D DBT reconstructions and the magnification/2D view, illustrating the model’s value. The phantom was able to highlight changes in the contrast and resolution, which is crucial for accurate microcalcification evaluation. Conclusions: Based on the crystalline growth, this phantom model offers an important new standardized target for evaluating DBT systems. By promoting standardization, especially through the development of advanced breast calcification phantoms, this work and design aimed to contribute to improving earlier and more accurate breast cancer detection. Full article

Lung cancer is the leading cause of cancer-related deaths globally and ranks among the most common cancer types. Given its low overall five-year survival rate, early diagnosis and timely treatment are essential to improving patient outcomes. In recent years, advances in computer technology have enabled artificial intelligence to make groundbreaking progress in imaging-based lung cancer diagnosis. The primary aim of this study is to develop a computer-aided diagnosis (CAD) system for lung cancer using endobronchial ultrasonography (EBUS) images and deep learning algorithms to facilitate early detection and improve patient survival rates. We propose ${\mathrm{M}}^3\text{-}\mathrm{Net}$, which is a multi-branch framework that integrates multiple features through an attention-based mechanism, enhancing diagnostic performance by providing more comprehensive information for lung cancer assessment. The framework was validated on a dataset of 95 patient cases, including 13 benign and 82 malignant cases. The dataset comprises 1140 EBUS images, with 540 images used for training, and 300 images each for the validation and test sets. The evaluation yielded the following results: accuracy of 0.76, F1-score of 0.75, AUC of 0.83, PPV of 0.80, NPV of 0.75, sensitivity of 0.72, and specificity of 0.80. These findings indicate that the proposed attention-based multi-feature fusion framework holds significant potential in assisting with lung cancer diagnosis. Full article

Background: X-ray computed tomography (CT) imaging technology provides high-precision anatomical visualization of patients and has become a standard modality in clinical diagnostics. A widely adopted strategy to mitigate radiation exposure is sparse-view scanning. However, traditional iterative approaches require manual design of regularization priors and laborious parameter tuning, while deep learning methods either heavily depend on large datasets or fail to capture global image correlations. Methods: Therefore, this paper proposes a combination of model-driven and data-driven methods, using the ADMM iterative algorithm framework to constrain the network to reduce its dependence on data samples and introducing the CNN and Transformer model to increase the ability to learn the global and local representation of images, further improving the accuracy of the reconstructed image. Results: The quantitative and qualitative results show the effectiveness of our method for sparse-view reconstruction compared with the current most advanced reconstruction algorithms, achieving a PSNR of 42.036 dB, SSIM of 0.979, and MAE of 0.011 at 32 views. Conclusions: The proposed algorithm has effective capability in sparse-view CT reconstruction. Compared with other deep learning algorithms, the proposed algorithm has better generalization and higher reconstruction accuracy. Full article

Purpose: Dual-energy computed tomography (DECT) allows for the measurement of iodine concentration, a component for the synthesis of thyroid hormones. DECT can create virtual non-contrast (VNC) images, potentially reducing radiation exposure. This study explores the correlations between thyroid function and iodine concentration, as well as the relationship between thyroid densities in true non-contrast (TNC) and virtual non-contrast (VNC) images and thyroid function. Methods: The study involved 87 patients undergoing 4D-CT imaging with single and dual-energy scans for diagnosing primary hyperparathyroidism. Thyroid densities and iodine concentrations were measured across all scanning phases. These measurements were correlated with thyroid function, indicated by TSH and FT4 levels. Differences in thyroid density between post-contrast phases and TNC phases (ΔHU) were analyzed for correlations with thyroid function and iodine concentrations. Results: Positive correlations between iodine concentrations and TSH were found, with Spearman’s coefficients (R) of 0.414, 0.361, and 0.349 for non-contrast, arterial, and venous phases, respectively. Thyroid density on TNC showed significant positive correlations with TSH levels (R = 0.436), consistently across both single- (R = 0.435) and dual-energy (R = 0.422) scans. Thyroid densities on VNC images did not correlate with TSH or FT4. Differences in density between contrast and non-contrast scans (ΔHU) negatively correlated with TSH (p = 0.002). Conclusions: DECT-derived iodine concentrations and thyroid densities in non-contrast CT scans demonstrated positive correlations with thyroid function, in contrast to thyroid densities on VNC scans. This indicates that VNC images are unsuitable for this purpose. Correlations between ΔHU and TSH suggest a potential link between the thyroid’s structural properties to capture iodine and its hormonal function. This study underscores the potential value of (DE-) CT imaging for evaluating thyroid function as an additional benefit in head and neck scans. Full article

Background/Objective: Longitudinal in vivo studies of murine xenograft models are widely utilized in oncology to study cancer biology and develop therapies. Magnetic resonance imaging (MRI) of these tumors is an invaluable tool for monitoring tumor growth and characterizing the tumors as well. Methods: In this work, a pipeline for automating the segmentation of xenografts in mouse models was developed. T₂-weighted (T2-wt) MRI images from mice implanted with six different prostate cancer patient-derived xenografts (PDX) in the kidneys, liver, and tibia were used. The segmentation pipeline included a slice classifier to identify the slices that had tumors and subsequent training and validation using several U-Net-based segmentation architectures. Multiple combinations of the algorithm and training images for different sites were evaluated for inference quality. Results and Conclusions: The slice classifier network achieved 90% accuracy in identifying slices containing tumors. Among the various segmentation architectures tested, the dense residual recurrent U-Net achieved the highest performance in kidney tumors. When evaluated across the kidneys, tibia, and liver, this architecture performed the best when trained on all data as compared to training on only data from a single site (and inferring on a multi-site tumor images), achieving a Dice score of 0.924 across the test set. Full article

Background/Objectives: Radiomic features exhibit a correlation with tumor size on pretreatment images. However, on post-treatment images, this association is influenced by treatment efficacy and varies between responders and non-responders. This study introduces a novel model, called baseline-referenced Delta radiomics, which integrates the association between radiomic features and tumor size into Delta radiomics to predict chemotherapy response in liver metastases from breast cancer (BC) and colorectal cancer (CRC). Materials and Methods: A retrospective study analyzed contrast-enhanced computed tomography (CT) scans of 83 BC patients and 84 CRC patients. Among these, 57 BC patients with 106 liver lesions and 37 CRC patients with 109 lesions underwent post-treatment imaging after systemic chemotherapy. Radiomic features were extracted from up to three lesions per patient following manual segmentation. Tumor response was assessed by measuring the longest diameter and classified according to RECIST 1.1 criteria as progressive disease (PD), partial response (PR), or stable disease (SD). Classification models were developed to predict chemotherapy response using pretreatment data only, Delta radiomics, and baseline-referenced Delta radiomics. Model performance was evaluated using confusion matrix metrics. Results: Baseline-referenced Delta radiomics performed comparably or better than established radiomics models in predicting tumor response in chemotherapy-treated patients with liver metastases. The sensitivity, specificity, and balanced accuracy in predicting response ranged from 0.66 to 0.97, 0.81 to 0.97, and 80% to 90%, respectively. Conclusions: By integrating the relationship between radiomic features and tumor size into Delta radiomics, baseline-referenced Delta radiomics offers a promising approach for predicting chemotherapy response in liver metastases from breast and colorectal cancer. Full article

Background: Zygomatico-maxillary complex (ZMC) fractures are prevalent facial injuries with significant functional and aesthetic implications. Computer-assisted surgery (CAS) offers precise surgical planning and outcome evaluation. The study aimed to evaluate the application of CAS in the analysis of ZMC fracture outcomes and to propose a reproducible workflow for surgical outcome assessment using cephalometric landmarks. Methods: A retrospective cohort study was conducted on 16 patients treated for unilateral ZMC fractures at the Maxillofacial Surgery Unit of Siena University Hospital (2017–2024). Inclusion criteria included ZMC fractures classified as Zingg B or C, treated via open reduction and internal fixation (ORIF). Pre- and post-operative CT scans were processed for two- and three-dimensional analyses. Discrepancies between CAS-optimized reduction and achieved surgical outcomes were quantified using cephalometric landmarks and volumetric assessments. Results: Out of the 16 patients (69% male, mean age 48.1 years), fractures were predominantly on the right side (81%). CAS comparison between the post-operative and the contralateral side revealed significant asymmetries along the X and Y axes, particularly in the fronto-zygomatic suture (FZS), zygo-maxillary point (MP), and zygo-temporal point (ZT). Computer-assisted comparison between the post-operative and the CAS-simulated reductions showed statistical differences along all three orthonormal axes, highlighting the challenges in achieving ideal symmetry despite advanced surgical techniques. CAS-optimized reductions demonstrated measurable improvements compared to traditional methods, underscoring their utility in outcome evaluation. Conclusions: CAS technology enhances the precision of ZMC fracture outcome evaluation, allowing for detailed comparison between surgical outcomes and virtual simulations. Its application underscores the potential for improved surgical planning and execution, especially in complex cases. Future studies should focus on expanding sample size, refining workflows, and integrating artificial intelligence to automate processes for broader clinical applicability. Full article

Objectives: To assess the elasticity values in breast tissues using magnetic resonance elastography (MRE) and examine the association between elasticity values of invasive breast cancer with prognostic factors and the pathologic response to neoadjuvant systemic therapy (NST). Methods: A total of 57 patients (mean age, 54.1 years) with invasive breast cancers larger than 2 cm in diameter on ultrasound were prospectively enrolled. The elasticity values (mean, minimum, and maximum) of invasive breast cancers, normal fibroglandular tissues, and normal fat tissues were measured via MRE using a commercially available acoustic driver and compared. Elasticity values of breast cancers were compared according to prognostic factors and pathologic responses in patients who received NST before surgery. Receiver operating curve analysis was performed to evaluate the predictive efficacy of elasticity values in terms of pathological response. Results: Among the 57 patients, the mean elasticity value of invasive breast cancers was significantly higher than that of normal fibroglandular tissue and normal fat tissue (7.90 ± 5.80 kPa vs. 2.54 ± 0.80 kPa vs. 1.32 ± 0.33 kPa, all ps < 0.001). Invasive breast cancers with a large diameter (>4 cm) exhibited significantly higher mean elasticity values relative to tumors with a small diameter (≤4 cm) (11.65 ± 7.22 kPa vs. 5.87 ± 3.58 kPa, p = 0.002). Among 24 patients who received NST, mean, minimum, and maximum elasticity values significantly differed between the pathologic complete response (pCR) and non-pCR groups (all ps < 0.05). For the mean elasticity value, the area under the curve value for distinguishing pCR and non-pCR groups was 0.880 (95% confidence interval, 0.682, 0.976; p < 0.001). Conclusions: The elasticity values of invasive breast cancers measured via breast MRE showed a positive correlation with tumor size and showed potential in predicting the therapeutic response in patients receiving NST. Full article

Background: The aim of this study was to analyse patients who underwent bronchoscopy for suspected foreign body aspiration (FBA) and to evaluate the properties of computed tomography (CT) in preventing unnecessary bronchoscopy, which carries the risk of serious complications. Methods: All patients younger than 18 years of age who were evaluated for FBA at a tertiary children’s hospital between June 2014 and February 2023. Results: A total of 165 children were included in this study. In Group I (only bronchoscopy), the detection rate of FBA was 77.9%, whereas in Group II (CT ± bronchoscopy), the detection rate of FBA was 93%. Additionally, the negative diagnosis rate in Group II (7%) was significantly higher compared to Group I (22.1%). Conclusions: Low-dose chest CT is a highly effective and reliable imaging method for the diagnosis of FBA due to its rapid performance, minimal radiation exposure, and high sensitivity and specificity; therefore, it can prevent unnecessary bronchoscopies in suspicious cases and increase the positive bronchoscopy rate. Full article

Background/Objectives: Diabetes-related pancreatic changes on MRI remain unclear. Thus, we evaluated the pancreatic changes on MRI in patients with both type 1 diabetes (T1D) and type 2 diabetes (T2D) using multiparametric MRI. Methods: This prospective study involved patients with T1D or T2D who underwent upper abdominal 3-T MRI. Additionally, patients without impaired glucose metabolism were retrospectively included as a control. The imaging data included pancreatic anteroposterior (AP) diameter, pancreas-to-muscle signal intensity ratio (SIR) on fat-suppressed T1-weighted image (FS-T1WI), apparent diffusion coefficient (ADC) value, T1 value on T1 map, proton density fat fraction (PDFF), and mean secretion grade of pancreatic juice flow on cine-dynamic magnetic resonance cholangiopancreatography (MRCP). The MR measurements were compared using one-way analysis of variance and the Kruskal–Wallis test. Results: Sixty-one patients with T1D (n = 7) or T2D (n = 54) and 21 control patients were evaluated. The pancreatic AP diameters were significantly smaller in patients with T1D than in patients with T2D (p < 0.05). The average SIR on FS-T1WI was significantly lower in patients with T1D than in controls (p < 0.001). The average ADC and T1 values of the pancreas were significantly higher in patients with T1D than in patients with T2D (p < 0.01) and controls (p < 0.05). The mean secretion grade of pancreatic juice flow was significantly lower in patients with T1D than in controls (p = 0.019). The average PDFF of the pancreas was significantly higher in patients with T2D than in controls (p = 0.029). Conclusions: Patients with T1D had reduced pancreas size, increased pancreatic T1 and ADC values, and decreased pancreatic juice flow on cine-dynamic MRCP, whereas patients with T2D had increased pancreatic fat content. Full article

Background/Objectives: This study explores the application of vision transformers to predict early responses to stereotactic radiosurgery in patients with brain metastases using minimally pre-processed magnetic resonance imaging scans. The objective is to assess the potential of vision transformers as a predictive tool for clinical decision-making, particularly in the context of imbalanced datasets. Methods: We analyzed magnetic resonance imaging scans from 19 brain metastases patients, focusing on axial fluid-attenuated inversion recovery and high-resolution contrast-enhanced T1-weighted sequences. Patients were categorized into responders (complete or partial response) and non-responders (stable or progressive disease). Results: Despite the imbalanced nature of the dataset, our results demonstrate that vision transformers can predict early treatment responses with an overall accuracy of 99%. The model exhibited high precision (99% for progression and 100% for regression) and recall (99% for progression and 100% for regression). The use of the attention mechanism in the vision transformers allowed the model to focus on relevant features in the magnetic resonance imaging images, ensuring an unbiased performance even with the imbalanced data. Confusion matrix analysis further confirmed the model’s reliability, with minimal misclassifications. Additionally, the model achieved a perfect area under the receiver operator characteristic curve (AUC = 1.00), effectively distinguishing between responders and non-responders. Conclusions: These findings highlight the potential of vision transformers, aided by the attention mechanism, as a non-invasive, predictive tool for early response assessment in clinical oncology. The vision transformer (ViT) model employed in this study processes MRIs as sequences of patches, enabling the capture of localized tumor features critical for early response prediction. By leveraging patch-based feature learning, this approach enhances robustness, interpretability, and clinical applicability, addressing key challenges in tumor progression prediction following stereotactic radiosurgery (SRS). The model’s robust performance, despite the dataset imbalance, underscores its ability to provide unbiased predictions. This approach could significantly enhance clinical decision-making and support personalized treatment strategies for brain metastases. Future research should validate these findings in larger, more diverse cohorts and explore the integration of additional data types to further optimize the model’s clinical utility. Full article

Purpose: Sex classification is a major benchmark of previous work in learning on the structural connectome, a naturally occurring brain graph that has proven useful for studying cognitive function and impairment. While graph neural networks (GNNs), specifically graph convolutional networks (GCNs), have gained popularity lately for their effectiveness in learning on graph data, achieving strong performance in adult sex classification tasks, their application to pediatric populations remains unexplored. We seek to characterize the capacity for GNN models to learn connectomic patterns on pediatric data through an exploration of training techniques and architectural design choices. Methods: Two datasets comprising an adult BRIGHT dataset (N = 147 Hodgkin’s lymphoma survivors and N = 162 age similar controls) and a pediatric Human Connectome Project in Development (HCP-D) dataset (N = 135 healthy subjects) were utilized. Two GNN models (GCN simple and GCN residual), a deep neural network (multi-layer perceptron), and two standard machine learning models (random forest and support vector machine) were trained. Architecture exploration experiments were conducted to evaluate the impact of network depth, pooling techniques, and skip connections on the ability of GNN models to capture connectomic patterns. Models were assessed across a range of metrics including accuracy, AUC score, and adversarial robustness. Results: GNNs outperformed other models across both populations. Notably, adult GNN models achieved 85.1% accuracy in sex classification on unseen adult participants, consistent with prior studies. The extension of the adult models to the pediatric dataset and training on the smaller pediatric dataset were sub-optimal in their performance. Using adult data to augment pediatric models, the best GNN achieved comparable accuracy across unseen pediatric (83.0%) and adult (81.3%) participants. Adversarial sensitivity experiments showed that the simple GCN remained the most robust to perturbations, followed by the multi-layer perceptron and the residual GCN. Conclusions: These findings underscore the potential of GNNs in advancing our understanding of sex-specific neurological development and disorders and highlight the importance of data augmentation in overcoming challenges associated with small pediatric datasets. Further, they highlight relevant tradeoffs in the design landscape of connectomic GNNs. For example, while the simpler GNN model tested exhibits marginally worse accuracy and AUC scores in comparison to the more complex residual GNN, it demonstrates a higher degree of adversarial robustness. Full article

Background/Objectives: Correct pulmonary nodule volumetry and categorization is paramount for accurate diagnosis in lung cancer screening programs. CT scanners with slice thicknesses of multiple millimetres are still common worldwide, and slice thickness has an adverse effect on the accuracy of the pulmonary nodule volumetry. Methods: We propose a deep learning based super-resolution technique to generate thin-slice CT images from thick-slice CT images. Analysis of the lung nodule volumetry and categorization accuracy was performed using commercially available AI-based lung cancer screening software. Results: The accuracy of pulmonary nodule categorization increased from 72.7 percent to 94.5 percent when thick-slice CT images were converted to generated-thin-slice CT images. Conclusions: Applying the super-resolution-based slice generation on thick-slice CT images prior to automatic nodule evaluation significantly increases the accuracy of pulmonary nodule volumetry and corresponding pulmonary nodule category. Full article