Cristian Constantin Volovăț

Background: Metastatic-site frequencies reflect both how extensively a cancer spreads and how commonly an organ is involved across cancers. We assessed whether cancer–organ affinity beyond these overall effects is reproducible between autopsy and clinical cohorts. Methods: We analyzed 796,211 organ-involvement events from 332,195 Japanese malignant autopsy cases, covering 76 primary cancers and 48 target sites. We quantified observed-to-expected affinity with empirical-Bayes shrinkage and screened associations with multiple-testing correction. Robust associations underwent anatomical adjudication and fixed-margin testing. External validation compared 143 harmonized cancer–organ pairs (13 cancers, 11 targets) with an independently analyzed MSK-MET-derived clinical matrix. Results: We identified 176 robust enriched associations; 10 remained in a conservative analysis of patterns compatible with distant-organ spread. All 176 remained significant under exact fixed-margin tests. Among 5000 fixed-margin randomized matrices, none contained more than three robust associations, versus 176 observed (Monte Carlo p = 2.00 × 10⁻⁴). Conventional metastatic-site fractions were highly correlated across cohorts (Spearman ρ = 0.861). Residual affinities also showed concordance (Spearman ρ = 0.669; 95% bootstrap confidence interval, 0.416–0.839; Pearson r = 0.747), with matching enrichment or depletion directions in 76.9% of pairs. However, only three of the ten selected distant-organ associations could be compared individually across cohorts, and two showed enrichment in both datasets. Conclusions: Residual metastatic affinity captures partly reproducible population-level patterns beyond metastatic frequency. Validation was restricted to harmonizable categories; these aggregate findings neither estimate individual metastatic risk nor confirm lesion-level distant spread.

Background/Objectives: Early prediction of pathologic complete response (pCR) during neoadjuvant therapy could improve diagnostic assessment of treatment response, adaptive treatment monitoring, and early prognostic stratification in breast cancer, but longitudinal public MRI archives are not immediately ready for reproducible paired-image modeling. The aim of this study was to develop a reproducible longitudinal breast MRI workflow from the public ACRIN 6698/BMMR2 archive and to compare handcrafted radiomics with frozen deep image embeddings for early treatment-response modeling and imaging-based pCR prediction. Methods: The raw archive was reduced to a validated 183-patient cohort with matched T0 and T1 cropped DCE image–mask pairs, preserved predefined train/test assignment, and binary pCR labels. Handcrafted radiomic features were extracted at T0 and T1 using a MIRP-based workflow, longitudinal delta descriptors were constructed, and train-only cleaning, filtering, and feature selection were applied before predictive modeling. In parallel, lesion-centered paired 2.5D tensors were generated for image-based deep analysis. Exploratory end-to-end paired deep-learning models were evaluated but showed substantial overfitting. Therefore, a frozen pretrained ResNet18 feature-extraction strategy was used to derive T0, T1, DELTA, and AVG image embeddings, which were then modeled using train-only selection and classical classifiers. Results: In the handcrafted radiomics branch, the strongest refined model was the T1-only random forest model with 10 selected features, achieving an AUROC of 0.670 and an AUPRC of 0.405 on the independent test subset. In the frozen deep-embedding branch, the best configuration by AUROC was the DELTA-only random forest model with 30 selected features (AUROC 0.672, AUPRC 0.408), whereas the strongest configuration by AUPRC was the AVG-only logistic regression model with 100 selected features (AUROC 0.666, AUPRC 0.472). End-to-end paired deep-learning models were technically feasible but were not retained because of marked overfitting. Conclusions: A complex public longitudinal breast MRI archive can be converted into a reproducible modeling cohort for early response assessment. Within this framework, handcrafted radiomics and frozen deep image embeddings yielded comparable exploratory discrimination but emphasized different representation spaces: the strongest handcrafted signal was concentrated in the early-treatment T1 phenotype, whereas the strongest deep signal was concentrated in latent longitudinal change and integrated paired-image state. These findings indicate that longitudinal breast MRI-derived biomarkers warrant further investigation for early treatment-response assessment; however, the present models remain exploratory and require independent external validation before any clinical application.

Background/Objectives: Autism Spectrum Disorder (ASD) is associated with structural brain alterations, particularly involving white matter and connectivity. Artificial intelligence (AI) enhances the detection of subtle neuroanatomical changes. This study aimed to characterize structural abnormalities and volumetric patterns in children with ASD using AI-assisted MRI. Methods: This retrospective study included 90 children diagnosed with ASD. Brain MRI scans were analyzed using the CE-certified AI platform mdbrain. Structural findings were classified into corpus callosum anomalies, white matter signal abnormalities (WMSA), ventriculomegaly, other abnormalities, or no detectable changes. Group differences were assessed using ANOVA and Kruskal–Wallis tests with Tukey post hoc analysis. Logistic regression, principal component analysis (PCA), and linear discriminant analysis (LDA) were applied. Results: WMSA were identified in 23.3% of patients, followed by other anomalies (27.8%), corpus callosum anomalies (8.9%), ventriculomegaly (8.9%), and no abnormalities (31.1%). Total white matter volume was significantly reduced in pathological groups and was the only independent predictor. PCA identified three principal components reflecting shared temporo-parietal covariance, hemispheric asymmetry, and a white matter-related axis. Exploratory LDA demonstrated partial separation among anomaly categories. Conclusions: Children with ASD in this cohort showed heterogeneous but partially structured MRI alterations involving both focal and global volumetric changes. Reduced total white matter volume was the most consistent multivariable association with structural abnormalities. AI-assisted morphometric analysis may support structural phenotyping in ASD. These findings are exploratory and require confirmation in larger, prospectively validated cohorts before biomarker applications can be considered.

Cancer immunotherapy increasingly relies on nucleic acid-based vaccines, yet achieving efficient and safe delivery remains a critical limitation. Polyplexes—electrostatic complexes of cationic polymers and nucleic acids—have emerged as versatile carriers offering greater chemical tunability and multivalent targeting capacity compared to lipid nanoparticles, with lower immunogenicity than viral vectors. This review summarizes key design principles governing polyplex performance, including polymer chemistry, architecture, and assembly route—emphasizing microfluidic fabrication for improved size control and reproducibility. Mechanistically, effective systems support stepwise delivery: tumor targeting, cellular uptake, endosomal escape (via proton-sponge, membrane fusion, or photochemical disruption), and compartment-specific cargo release. We discuss therapeutic applications spanning plasmid DNA, siRNA, miRNA, mRNA, and CRISPR-based editing, highlighting preclinical data across multiple tumor types and early clinical evidence of on-target knockdown in human cancers. Particular attention is given to physiological barriers and engineering strategies—including size-switching systems, charge-reversal polymers, and tumor-penetrating peptides—that improve intratumoral distribution. However, significant challenges persist, including cationic toxicity, protein corona formation, manufacturing variability, and limited clinical responses to date. Current evidence supports polyplexes as a modular platform complementary to lipid nanoparticles in selected oncology indications, though realizing this potential requires continued optimization alongside rigorous translational development.

Background/Objectives: Glioblastoma (GBM) exhibits heterogeneous, nonlinear invasion patterns that challenge conventional modeling and radiomic prediction. Most deep learning approaches describe the morphology but rarely capture the dynamical stability of tumor evolution. We propose an AI framework that approximates a latent attractor landscape of GBM morphodynamics—stable basins in a continuous manifold that are consistent with reproducible morphologic regimes. Methods: Multimodal MRI scans from BraTS 2020 (n = 494) were standardized and embedded with a 3D autoencoder to obtain 128-D latent representations. Unsupervised clustering identified latent basins (“attractors”). A neural ordinary differential equation (neural-ODE) approximated latent dynamics. All dynamics were inferred from cross-sectional population variability rather than longitudinal follow-up, serving as a proof-of-concept approximation of morphologic continuity. Voxel-level perturbation quantified local morphodynamic sensitivity, and proof-of-concept control was explored by adding small inputs to the neural-ODE using both a deterministic controller and a reinforcement learning agent based on soft actor–critic (SAC). Survival analyses (Kaplan–Meier, log-rank, ridge-regularized Cox) assessed associations with outcomes. Results: The learned latent manifold was smooth and clinically organized. Three dominant attractor basins were identified with significant survival stratification (χ² = 31.8, p = 1.3 × 10⁻⁷) in the static model. Dynamic attractor basins derived from neural-ODE endpoints showed modest and non-significant survival differences, confirming that these dynamic labels primarily encode the morphodynamic structure rather than fixed prognostic strata. Dynamic basins inferred from neural-ODE flows were not independently prognostic, indicating that the inferred morphodynamic field captures geometric organization rather than additional clinical risk information. The latent stability index showed a weak but borderline significant negative association with survival (ρ = −0.13 [−0.26, −0.01]; p = 0.0499). In multivariable Cox models, age remained the dominant covariate (HR = 1.30 [1.16–1.45]; p = 5 × 10⁻⁶), with overall C-indices of 0.61–0.64. Voxel-level sensitivity maps highlighted enhancing rims and peri-necrotic interfaces as influential regions. In simulation, deterministic control redirected trajectories toward lower-risk basins (≈57% success; ≈96% terminal distance reduction), while a soft actor–critic (SAC) agent produced smoother trajectories and modest additional reductions in terminal distance, albeit without matching the deterministic controller’s success rate. The learned attractor classes were internally consistent and clinically distinct. Conclusions: Learning a latent attractor landscape links generative AI, dynamical systems theory, and clinical outcomes in GBM. Although limited by the cross-sectional nature of BraTS and modest prognostic gains beyond age, these results provide a mechanistic, controllable framework for tumor morphology in which inferred dynamic attractor-like flows describe latent organization rather than a clinically predictive temporal model, motivating prospective radiogenomic validation and adaptive therapy studies.

Colorectal cancer (CRC), a prevalent malignancy, is a significant global health concern. The intricate interplay of genetic mutations, inflammatory processes, and environmental factors underscores the complexity of CRC’s etiology. The human gut harbors a diverse microbial community that plays a key role in maintaining homeostasis and influencing various aspects of host physiology. Perturbations in the gut microbiome (GM) composition and function have been implicated in CRC carcinogenesis. This bidirectional relationship involves microbial contributions to inflammation, DNA damage, and immune modulation, shaping the tumor microenvironment (TME). Bacteriophages, viruses that infect bacteria, contribute to the microbiome’s diversity and function by influencing bacterial abundance and composition. These phages can impact host–microbiome interactions, potentially influencing CRC risk. Furthermore, they can be manipulated to transport targeted medication, without being metabolized. Antibiotics exert selective pressures on the gut microbiome, leading to shifts in bacterial populations and potential dysbiosis. Probiotics can modulate the composition and activity of the GM and could be considered adjunctive therapy in the treatment of CRC. Understanding the intricate balance between bacteriophages, antibiotics–probiotics, and the GM is essential for comprehending CRC etiology and progression.

Background: Survival prediction in patients with brain metastases remains a major clinical challenge, where timely and individualized prognostic estimates are critical for guiding treatment strategies and patient counseling. Methods: We propose a novel hybrid deep learning framework that integrates volumetric MRI-derived imaging biomarkers with structured clinical and demographic data to predict overall survival time. Our dataset includes 148 patients from three institutions, featuring expert-annotated segmentations of enhancing tumors, necrosis, and peritumoral edema. Two convolutional neural network backbones—ResNet-50 and EfficientNet-B0—were fused with fully connected layers processing tabular data. Models were trained using mean squared error loss and evaluated through stratified cross-validation and an independent held-out test set. Results: The hybrid model based on EfficientNet-B0 achieved state-of-the-art performance, attaining an R² score of 0.970 and a mean absolute error of 3.05 days on the test set. Permutation feature importance highlighted edema-to-tumor ratio and enhancing tumor volume as the most informative predictors. Grad-CAM visualizations confirmed the model’s attention to anatomically and clinically relevant regions. Performance consistency across validation folds confirmed the framework’s robustness and generalizability. Conclusions: This study demonstrates that multimodal deep learning can deliver accurate, explainable, and clinically actionable survival predictions in brain metastases. The proposed framework offers a promising foundation for integration into real-world oncology workflows to support personalized prognosis and informed therapeutic decision-making.

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.

In the current context, breast cancer survivors are characterized by an increased life expectancy through the introduction of screening methods but also of the multidisciplinary therapeutic approach. Thus, the appearance and persistence of adverse effects associated with cancer and its treatment, including cognitive dysfunction, can influence the quality of life in these patients. It is imperative to define the ideas that are incorporated in the vast concept of "cognitive dysfunction associated with cancer" or "oncobrain," alongside to describe the processes that lead to it. Therefore, certain physiological, but also pathological, values of the patient's profile influence the appearance and progression of chemobrain. Among the most frequently cited factors in the literature are age, disorders such as anxiety and depression, and educational level. Furthermore, based on the etiology of neurodegenerative diseases, researchers have developed multiple theories about the causes of chemobrain, many of which rely on data from clinical practice. Given the wide range of changes involved, it's crucial to identify the tests that can both subjectively and objectively demonstrate the cognitive dysfunction associated with the various targeted cognitive impairments. These observations will pave the way for progress in the therapy of these adverse effects.

Background/Objectives : This study investigates the utility of vision transformers (ViTs) for predicting early responses to SRS using a minimal pre-processing approach on MRI images. Methods: We analyzed MRI scans from 19 patients with BM, focusing on axial fluid-attenuated inversion recovery (FLAIR) and high-resolution contrast-enhanced T1-weighted (CE T1w) sequences. Patients were classified as responders (complete or partial response) or non-responders (stable or progressive disease). Results: Our findings demonstrate that ViTs can effectively predict treatment responses, achieving an overall accuracy of 97%. The model exhibited high precision (96% for progression and 97% for regression) and strong recall rates (92% for progression and 99% for regression), distinguishing between treatment outcomes reliably. The confusion matrix analysis further supports its reliability, with minimal misclassifications. The model also achieved an almost perfect area under the ROC curve (AUC = 0.99), indicating accurate differentiation between responders and non-responders across various thresholds. Conclusions : These findings highlight the potential of the Vision Transformer model as a non-invasive predictive tool in clinical settings, significantly influencing clinical decision-making processes and enhancing patient management for brain metastases. By improving early response predictions, this model contributes to the development of personalized treatment strategies in oncology. Future research should focus on validating these results in larger, diverse cohorts, integrating additional data types, and refining the model to further enhance its utility in clinical practice.