Bulent Aslan

Purpose: Clinical assessment of vertebral lesion quality (osteolytic, osteoblastic, mixed) remains subjective, with limited interobserver reliability. This study evaluated a novel application of 3D convolutional neural networks (3D-CNNs) for classifying lesion quality from CT volumes in metastatic cancer patients. Materials and Methods: This retrospective study used CT data from 151 cancer patients planned for radiotherapy for metastatic spine disease (September 2020-July 2024). Leveraging vertebra-level expert annotations, we introduced an unconventional U-Net-based strategy converting coarse voxel-wise predictions into vertebra-level lesion classifications. The final dataset comprised 2,125 vertebrae across four classes (no lesion, osteolytic, osteoblastic, mixed), split into a 3-fold cross-validation set and an independent holdout test set. Model performance was benchmarked against a DenseNet121 baseline and a musculoskeletal radiologist, with Cohen's kappa assessing inter-rater agreement. Results: The 3D model achieved an ensemble accuracy of 84.7%, outperforming DenseNet121 (72.1%), with substantial gains in F1 score, precision, and balanced accuracy. It showed high concordance with the radiologist (Cohen's kappa = 0.76) and comparable sensitivity and specificity across all lesion subtypes. We found both models and the radiologist to struggle with osteolytic lesions, reflecting the difficulty of distinguishing this class from age-related changes in vertebral bone density and architecture caused by benign bone lesions, age-related systemic skeletal disorders and cancer treatments. Conclusions: 3D-CNNs trained with vertebra-level labels can accurately and reliably classify vertebral metastatic lesion quality from CT scans, offering a scalable path toward automated characterization of metastatic spine disease to support clinical decision-making and large-scale radiomics research.

Objective: To assess the impact of split-bolus (SB) single scan CT on the conspicuity of clear cell renal cell carcinoma (ccRCC) metastases compared with single-bolus injection. Methods: This retrospective cohort study included consecutive patients with histologically proven metastatic ccRCC who underwent both SB and single-bolus portal venous abdominal CT within 6 months between 2017 and 2022 in a single tertiary center. SB CT utilized BMI-adjusted contrast dose and kVp (80 to 120) with concurrent arterial and portal venous phases. Single bolus CT utilized BMI-adjusted contrast dose at 120 kVp at the portal venous phase. Wilcoxon rank test compared the conspicuity of metastases between the protocols. Results: Of the 47 patients, 80.9% were male, with a mean age of 67±10.4 years and a BMI of 27.1±5.7. There were 48 paired CTs performed with a median interval of 93 days. Contrast dose was 143±27 ml for SB and 108±26 ml for single-bolus ( P P <0.001. Conclusions: SB scan provides dramatically higher conspicuity of ccRCC metastases as compared with single-bolus portal venous CT.

Purpose: Pontine diffuse midline gliomas (PDMGs) are among the most lethal pediatric brain tumors with a median survival of approximately one year. Reliable prognostic markers are needed to guide treatment strategies and inform families. Our study aims to investigate the prognostic utility of delta-radiomic features in PDMGs. Materials and methods: We retrospectively analyzed 35 pediatric patients with PDMG diagnosed between 2012 and 2020, all treated with radiotherapy plus concomitant and adjuvant temozolomide. Pre- and post-treatment MRI (T1-weighted, T2-weighted, and ADC maps) were subjected to manual segmentation and texture feature extraction using MaZda software. Delta radiomics features were calculated as ratios between post- and pre-treatment values. Diagnostic performance for predicting survival below or above 12 months was assessed using ROC analysis, while overall survival was further evaluated with Kaplan–Meier and Cox regression models. Results: Median overall survival was 13 months (range: 6–69). Among baseline features, only sum average from non-contrast T1 was significant, with lower values (indicating higher heterogeneity) associated with shorter OS. In the post-radiotherapy setting, sum entropy from T1 images emerged as an independent prognostic predictor, with higher values correlating with longer OS. Delta radiomics parameters, particularly sum entropy derived from T1 and T2 sequences, yielded higher AUC values than single post-treatment features, suggesting superior prognostic accuracy. Conclusion: Our preliminary results indicate that delta-radiomics outperforms static texture analysis in predicting overall survival in pediatric PDMGs, and to the best of our knowledge, this is the first study to investigate the prognostic utility of delta-radiomics in this population. Evidence is provided that delta radiomics can serve as a non-invasive marker for early prognostic stratification in pediatric PDMGs. Validation in larger, multi-center cohorts is required to confirm its clinical utility.

Purpose To evaluate the diagnostic contribution of MRI-based morphological and signal features—particularly T2-weighted signal intensity—for differentiating the etiology of pituitary stalk thickening (PST), and to develop imaging-based models for predicting inflammatory and neoplastic pathologies. Methods This retrospective study included 41 adult (51.2%) and pediatric (48.8%) patients with confirmed PST who underwent contrast-enhanced pituitary MRI between 2012 and 2021. Etiologies were classified as congenital/idiopathic, inflammatory/infectious, or neoplastic based on clinical, radiological, or histopathological criteria. Imaging findings including enhancement pattern and T2 signal intensity were assessed in consensus by two neuroradiologists blinded to clinical data. Statistical analyses included univariate and multivariate logistic regression and receiver operating characteristic (ROC) curve analysis. Results Neoplastic lesions were associated with significantly greater stalk thickness (median: 5.9 mm) compared to non-neoplastic lesions (median: 3.83 mm; p = 0.012). T2-weighted hyperintensity was present in 70% of neoplastic lesions, while hypointensity was more frequent in inflammatory/infectious lesions (p = 0.017). A multivariable model incorporating stalk thickness, non-T2 hypointensity and V-shaped enhancement patterns yielded excellent diagnostic performance for neoplastic pathologies (AUC: 0.848; 95% CI: 0.713–0.982). A second model using stalk thickness and T2 hypointensity predicted inflammatory lesions with an AUC of 0.836 (95% CI: 0.715–0.957). Conclusion To our knowledge, this is the first study to propose MRI-based models using stalk morphology and signal features to predict PST etiology. These non-invasive models, developed without clinical input, demonstrate promising diagnostic accuracy and may aid in differential diagnosis. Further validation in larger cohorts is needed.

Background: Medulloblastoma is the most common pediatric malignant brain tumor in the population, with molecular subtypes that differ in prognosis and therapeutic response. Identifying these subtypes before surgery is important for tailoring management. Objective: This study aimed to distinguish medulloblastoma molecular subtypes using a combination of conventional MRI features and MRI-based texture analysis. Materials and methods: We retrospectively analyzed 58 patients with preoperative MRI and histopathologic confirmation of medulloblastoma. Cases were classified into SHH pathway–activated or Group 3/4 subtypes. Morphologic MRI characteristics, apparent diffusion coefficient (ADC) ratios, and texture analysis parameters were compared between the groups. Results: Of the 58 patients, 55.2% had SHH pathway–activated tumors. Morphological features, including location out of the midline or in the cerebellar hemisphere (p<0.001), peri-tumoral edema (p=0.041), macrocysts (p=0.001), nodular involvement/lobulation (p=0.002), and heterogeneous contrast enhancement (p=0.002) were more common in SHH tumors. ADC measurements showed that the solid tumor-to-thalamus ratio was significantly lower in SHH tumors (p<0.001), with a threshold of 0.855 providing 82.1% sensitivity and 92.3% specificity. As for texture analysis parameters, kurtosis (p=0.023), SumOfSqs (p=0.022) and 01-10-50-90% percentile (p=0.011; p=0.001; p=0.006; and p=0.013 respectively) values obtained from ADC images and kurtosis (p=0.041), SumOfSqs (p=0.005), SumVarnc (p=0.014), SumEntrp (p=0.032) values obtained from T1W images were statistically significant in differentiating SHH and group 3/ group 4 medulloblastoma. Conclusion: Integrating MRI morphological features, ADC-based measurements, and texture analysis provides complementary information for non-invasive differentiation of medulloblastoma molecular subtypes.

Objective: To assess the impact of split-bolus (SB) single scan CT on the conspicuity of clear cell renal cell carcinoma (ccRCC) metastases compared to single-bolus injection. Methods: This retrospective cohort study included consecutive patients with histologically proven metastatic ccRCC who underwent both SB and single-bolus portal venous abdominal CT within 6 months between 2017 and 2022 in a single tertiary center. SB CT utilized BMI-adjusted contrast dose and kVp (80-120) with concurrent arterial and portal venous phases. Single bolus CT utilized BMI-adjusted contrast dose at 120kVp at the portal venous phase. Wilcoxon rank test compared the conspicuity of metastases between the protocols. Results: Of the 47 patients, 80.9% were male, with a mean age of 67 ± 10.4 years and a BMI of 27.1 ± 5.7. There were 48 paired CTs performed with a median interval of 93 days. Contrast dose was 143 ± 27 cc for SB and 108 ± 26 cc for single-bolus ( p < 0.001). 66 metastases were analyzed, with an average size of 2 cm: 48.5% in the pancreas, 28.8% in skeletal muscle, and 22.7% in the liver. The median contrast-to-noise ratio (CNR) was higher with SB compared to single-bolus for all metastases (3.0 vs. 1.4), pancreatic metastases (2.7 vs. 1.4), muscle metastases (5.2 vs. 2.0), and liver metastases (2.8 vs. 0.9), all p < 0.001. Conclusions: SB scan provides dramatically higher conspicuity of ccRCC metastases as compared to single-bolus portal venous CT.

Advancements in understanding the clinical, biochemical, and genetic aspects of primary mitochondrial disorders, along with the identification of a broad range of phenotypes frequently involving the central nervous system, have opened a new and crucial area in neuroimaging. This expanding knowledge presents significant challenges for radiologists in clinical settings, as the neuroimaging features and their associated metabolic abnormalities become more complex. This review offers a comprehensive overview of the key neuroimaging features associated with the common primary mitochondrial disorders. It highlights both the classical imaging findings and the emerging diagnostic insights related to several previously identified causative genes for these diseases. The review also provides an in-depth description of the clinicoradiologic presentations and potential underlying mitochondrial defects, aiming to enhance diagnostic abilities of radiologists in identifying primary mitochondrial diseases in their clinical practice. Graphical abstract

Purpose Given the high prevalence of vertebral fractures post-radiotherapy in patients with metastatic spine disease, accurate and rapid muscle segmentation could support efforts to quantify muscular changes due to disease or treatment and enable biomechanical modeling for assessments of vertebral loading to improve personalized evaluation of vertebral fracture risk. This study presents a deep-learning approach for segmenting the complete volume of the trunk muscles from clinical CT images trained using sparsely annotated data. Materials and Methods we extracted 2,009 axial CT images at the midpoint of each vertebral level (T4 to L4) from clinical CT of 148 cancer patients. The key extensor and flexor muscles (up to 8 muscles per side) were manually contoured and labeled per image in the thoracic and lumbar regions. We first trained a 2D nnU-Net deep-learning model on these labels to segment key extensor and flexor muscles. Using these sparse annotations per spine, we trained the model to segment each muscle’s entire 3D volume Results The proposed method achieved comparable performance to manual segmentations, as assessed by expert radiologists, with a mean Dice score above 0.769. Significantly, the model drastically reduced segmentation time, from 4.3-6.5 hours for manual segmentation of 14 single axial CT images to approximately 1 minute for segmenting the complete thoracic-abdominal 3D volume. Conclusion The approach demonstrates high potential for automating 3D muscle segmentation, significantly reducing the manual intervention required for generating musculoskeletal models, and could be instrumental in enhancing clinical decision-making and patient care in radiation oncology. Summary A deep learning 2D nnU-Net model, trained on a sparse set of 2D muscle annotations, successfully segmented the entire volume of 20 thoracolumbar muscles from cancer patients’ clinical CT data. The model showed a remarkable increase in segmentation efficacy and generalizability, achieving comparable performance to manual segmentations in delineating each muscle anatomy. Key Points ▪ A deep learning model (2D nnU-Net), developed using a sparse set of single axial CT-slice at each mid-per vertebral level, containing manual image annotation of 20 thoracic and lumbar muscles, achieved comparable performance to manual segmentations, as assessed by expert radiologists, with a mean Dice score above 0.769. ▪ The model drastically reduced segmentation time, from 4.3-6.5 hours for manual segmentation of 14 single axial CT images to approximately 1 minute for segmenting the complete thoracic-abdominal 3D volume. ▪ Radiologist assessment based on a Likert scale (0-5) for clinical acceptability of the muscle anatomical segmentation showed strong model performance for a representative sample of clinical CT data (a (mean(SD) of 4.66 (0.73)) and external data (4.66 (0.73).

Background Diagnosis and treatment of pseudotumor cerebri syndrome in children is still a challenge for clinicians. The aim of this study is to reveal the influence of presentation age and CSF opening pressure on long-term prognosis of pseudotumor cerebri and share our clinical data of the very young age (≤ 5-year) group. Method This retrospective study includes the patients followed by the Marmara University Pediatric Neurology Clinic between years 2012 and 2020 diagnosed with definite, probable, or suggestive pseudotumor cerebri syndrome according to modified Friedman criteria. Patients were classified into three groups according to presentation age: group 1: ≤ 5 years old; group 2: 6–10 years; and group 3 > 10 years old. CSF opening pressure was also categorized into three groups as CSF < 20 cmH20; CSF 20–30 cmH20; and CSF > 30 cmH20. Results One hundred three patients, 62.1% female (n = 64), were enrolled in the study. Group 1 consisted of 16 patients (60% male), group 2 consisted of 30 patients (63.3% female), and group 3 consisted of 57 patients (66.7% female). The mean CSF opening pressure did not differ between the three age groups in our study (p > 0.05). Treatment response was not correlated with CSF opening pressure. Papilledema presence and level of CSF opening pressure were independent of age (p > 0.05). Conclusions Age at presentation and CSF opening pressure at diagnosis are not any predictive factors that influence long-term prognosis of pseudotumor cerebri syndrome in children. Evaluation and follow-up of children should be done in personalized approach.