Determining the structure and following the structural evolution of molecules undergoing chemical reactions is one of the key goals of ultrafast molecular physics and chemistry. Recently, Coulomb explosion imaging has emerged as a promising technique for imaging the evolving structure of individual molecules in the gas phase. However, its practical application to structure determination is hampered by the lack of suitable algorithms for directly retrieving the molecular structure from the measured fragment-ion momentum data. Here, we propose a scheme to solve the underlying inverse problem by employing neural networks to infer the initial atomic positions from the final ion momenta on an event-by-event basis. Using this scheme, we retrieve the structure of several polyhalomethane isomers from simulated Coulomb explosion imaging data with an average per-atom position error of ∼0.1 atomic units, i.e., to within 5% of the typical bond lengths. This development paves the way for an automated structure retrieval from Coulomb explosion data one molecule at a time, making it ideally suitable for analyzing pump-probe experiments where several products are formed that need to be distinguished.
Microviscosity is an underexploited pathological cue that can be harnessed to regulate excited-state energy dissipation in organic near-infrared II (NIR-II, 1000-1700 nm) fluorophores. Here we introduce a twisted intramolecular charge transfer molecular rotor that uses microviscosity to gate radiative versus nonradiative decay, delivering a fluorescence "off" state in low-viscosity environments and a weak-but-sufficient NIR-II fluorescence turn-on in viscosity-elevated pathology while keeping nonradiative dissipation dominant for photothermal heating. A tunable donor-π-acceptor library was synthesized to drive absorption/emission toward the NIR-II window and calibrate a "dim-but-hot" photophysical profile. Molecularly dispersed FMR-1105-PEG enables in vivo NIR-II imaging in nonalcoholic fatty liver disease and acetaminophen-induced liver injury models. For tumor translation, micellization retains a measurable viscosity dependence (Förster-Hoffmann-type) while markedly increasing the apparent molar extinction coefficient, thereby enabling efficient 1064 nm photothermal activation, immunogenic cell death hallmarks in vitro, and imaging-guided tumor ablation even under tissue coverage, with favorable biosafety.
To investigate the clinical, imaging, pathological, and molecular characteristics of polymorphous low-grade neuroepithelial tumor of the young (PLNTY) and provide references for clinical diagnosis and treatment. We retrospectively analyzed the clinical data of one pathologically confirmed PLNTY patient, including medical history, clinical symptoms, imaging findings, pathological morphological features, immunohistochemical results and molecular testing data, and reviewed relevant literature. A 30-year-old male patient presented with a 6-year history of drug-resistant epilepsy as the primary manifestation, which progressed from focal facial myoclonus to mixed seizure phenotypes. Cranial magnetic resonance imaging (MRI) demonstrated an ovoid area of abnormal signal intensity in the left temporal pole, exhibiting isointensity on T1-weighted imaging (T1WI), hypointensity on T2-weighted imaging (T2WI) and fluid-attenuated inversion recovery (FLAIR) sequences, and no enhancement following contrast administration; the initial radiological impression was a suspected cavernous hemangioma. Twenty-four-hour video electroencephalography (vEEG) identified abnormal epileptiform discharges originating from the left anterior temporal lobe. Subsequent to left temporal lobe lesionectomy, pathological examination confirmed the diagnosis of polymorphous low-grade neuroepithelial tumor of the young (PLNTY), CNS WHO Grade 1. Microscopically, the lesion was characterized by thin-walled vascular clusters, focal vascular wall hyalinization, calcification, and adjacent reactive glial proliferation. Immunohistochemical (IHC) staining revealed diffuse CD34 positivity, glial fibrillary acidic protein (GFAP) (+), oligodendrocyte transcription factor 2 (Olig-2) (+), and a Ki-67 proliferation index of 1%, indicating low proliferative activity. Molecular testing confirmed BRAF mutation positivity, weak immunoreactivity for BRAF V600E (detected via VE1 antibody), and absence of malignant-associated mutations. Postoperatively, the patient achieved complete seizure control and remained seizure-free with uneventful recovery at the latest follow-up. PLNTY is typically characterized by long-standing drug-resistant epilepsy and has a predilection for the temporal lobe, with potential atypical imaging manifestations (e.g., hypointensity on T2WI as observed in this case). Definitive diagnosis requires the integration of pathological morphology, immunohistochemistry, and molecular testing. PLNTY harboring the BRAF V600E mutation exhibits indolent biological behavior; gross total resection serves as the core therapeutic modality, conferring a favorable prognosis. Molecular subtyping can guide individualized follow-up strategies, and clinicians should be vigilant against misdiagnosis of cases with atypical imaging features.
Magnetic resonance-guided radiotherapy enables repeated on-treatment imaging during chemoradiotherapy for glioblastoma (GBM), providing an opportunity to study dynamic tumor changes. We evaluated longitudinal volumetric change and spatial migration of T2/FLAIR signal abnormalities during treatment using a 1.5 T MR-Linac and examined their association with early disease progression. GBM patients suitable for chemoradiation on a 1.5 T MR-Linac were prospectively enrolled. Regions of T2/FLAIR signal abnormalities were contoured at baseline (F0), during selected treatment fractions, and at one-month post-radiotherapy (PM1). Tumor dynamics were quantified using percentage change in volume relative to F0 (Vrel) and migration distance (dmigrate), defined as the maximum linear displacement relative to F0. Patients were classified as early progressors (disease progression within 6 months after chemoradiation) or non-early progressors. Linear mixed-effects models evaluated longitudinal differences by early progression status, methylation, and extent of resection. Thirty-three patients were included; 28 had ≥ 6 months follow-up and 12 were classified as early progressors. In adjusted linear mixed-effects models, Vrel was higher in early progressors than in non-early progressors from F7 onward (all p < 0.025), with larger differences later in treatment (F15: 38%, p = 0.004; F30: 104%, p = 0.019). Median Vrel was 2% at F1, 2% at F15, and 18% at F30. Median dmigrate increased from 5.0 mm at F1 to 12.6 mm at F30 and was higher in early progressors from F19 onward (difference 5.9 mm, p = 0.02) and at F30 (8.5 mm, p = 0.01). At end of treatment, Vrel ≥ 10% occurred in 12/28 patients (43%) and dmigrate ≥ 10 mm in 13/28 (46%). Among patients with Vrel ≥ 10%, 10/12 (83%) were early progressors; among those with dmigrate ≥ 10 mm, 9/13 (69%) were early progressors. Serial 1.5 T MR-Linac imaging enables longitudinal assessment of T2/FLAIR abnormality volume (Vrel) and spatial migration (dmigrate) during chemoradiotherapy for glioblastoma. Greater increases in Vrel and dmigrate were associated with early progression, supporting the potential role of on-treatment imaging in informing risk-adapted MR-guided radiotherapy strategies.
Cell free DNA (cfDNA) screening for common aneuploidies is now widely integrated into prenatal care, and recent advances in next-generation sequencing, enhanced by unique molecular indexing methodology, have enabled cfDNA for selected single-gene disorders (cfDNA-SGD). Commercially available cfDNA-SGD panels target predominantly either de novo autosomal dominant or X-linked conditions that have recognizable prenatal or neonatal phenotypes. However, it remains a screening modality that relies on placental cell-free DNA, and encounters limitations. In this review, we summarize a currently available cfDNA-SGD panel that screens for 25 autosomal or X-linked dominant conditions corresponding to 30 unique genes, with several genes contributing to multiple conditions on the panel. We highlight the potentially associated prenatal and postnatal phenotypes and the imaging modalities that may help refine a prenatal genetic evaluation. We also discuss how ultrasound, fetal echocardiography, and fetal MRI across all trimesters can refine the prenatal differential diagnosis and help contextualize cfDNA-SGD results, while emphasizing that imaging findings remain the primary guide for management and that confirmatory diagnostic testing is required before clinical decisions and action.
Nosocomial infections are a major health problem worldwide. The increasing time patients spend in hospitals has led to an increase in mortality. Acinetobacter baumannii is an opportunistic pathogen that is a significant factor in nosocomial infections. The main aim of the current study was to investigate biofilm and pathogenicity-related genes among multidrug-resistant A. baumannii strains (n = 54) obtained from patients with respiratory infections. ERIC-PCR was also used to determine their molecular typing correlation. Disk diffusion and MIC methods were used to test antibiotic susceptibility. Biofilm formation was evaluated using crystal violet staining and SEM imaging. In this study, 96.30% of isolates formed biofilms, and 59.25% were strong biofilm producers. The most prevalent biofilm-related genes were pgaC (98.15%), pgaB (92.6%), and pgaA (79.6%), followed by epsA and ompA (74.1%). The isolates demonstrated high resistance to imipenem (100%), cefotaxime (98.15%), followed by cefepime, ceftazidime, levofloxacin, and piperacillin/tazobactam (96.3%). The presence of antibiotic-resistant genes was as follows: blaOXA-58 (20.4%), blaOXA-23 (5.55%), aacC1 (50.0%), aphA6 (45.45%), sulⅡ (63.25%), and sulⅠ (32.65%). The sulⅢ gene was not detected. A dendrogram based on UPGMA revealed significant genetic diversity among the 54 A. baumannii strains. Twenty-two ERIC types were identified, with 14 unique types and 8 common types. This study illuminates a concerning rise in antibiotic resistance and the widespread presence of resistance genes in A. baumannii strains. Furthermore, the high ability of these isolates to form biofilms likely contributes to their enhanced resistance, further complicating eradication efforts. Molecular typing demonstrated considerable genetic diversity.
The tumor microenvironment (TME) is a critical determinant of cancer initiation, progression, and therapeutic response. It comprises not only tumor cells but also immune cells, stromal components, and vasculature that interact through complex molecular signaling networks. In cancers driven by oncoviruses, viral infection represents a unique biological factor that profoundly influences TME formation and remodeling, thereby shaping distinct features such as immune evasion, chronic inflammation, and stromal and vascular reprogramming. Conventional approaches provide valuable insights into the static composition of the TME, but are inherently limited in capturing dynamic cellular behaviors, including immune cell migration, spatiotemporal immune remodeling, and CD8+ T cell interactions with virally infected tumor cells over time, which are critical for understanding immune evasion and therapeutic responses in oncovirus-associated cancers. In this review, we briefly summarize virus-associated cancers and the mechanisms by which oncoviruses actively shape and regulate the TME. We further highlight the importance of intravital imaging as a powerful approach for directly visualizing the spatial and temporal dynamics of the TME in live animals and discuss how this technology can advance our understanding of oncovirus-driven TME remodeling and the development of effective anticancer therapeutic strategies.
While modern imaging technologies offer unprecedented opportunities to observe life across scales, distilling an understanding of the underlying biological processes from these complex, high-dimensional data remains challenging. Computational analysis methods have been lagging behind our ability to produce data, as their development often requires expertise across multiple domains, including life and computer sciences. Annotated image datasets play a key role in fostering the development and improvement of microscopy image analysis methods, as they offer a realistic basis to build upon and invaluable ground truth to evaluate and optimize performance. Drawing inspiration from adjacent fields to microscopy imaging, we discuss in this perspective how sharing annotated datasets has driven progress in computational analysis. We emphasize the critical role that open data standards and infrastructure play in realizing the full scientific potential of annotated image datasets and close by highlighting opportunities for members across the scientific community to cultivate a dynamic ecosystem of data, infrastructure, and analysis methods to elevate research quality and accelerate innovation.
Prostate cancer is a common malignancy in men. With the advent of multiparametric MRI and the Prostate Imaging-Reporting and Data System (PI-RADS) framework for grading lesions, there have been multiple advancements in the management and treatment of the disease. There are several new advancements in prostate MRI technology that suggest further exciting developments in the field, ranging from novel imaging sequences to the application of machine learning methods and artificial intelligence (AI) across the imaging pipeline. In this review, we aim to provide context on the current advancements in prostate MRI and will discuss the benefits and drawbacks of several new imaging techniques including hybrid multidimensional MRI, low field strength MRI, restriction spectrum imaging, Vascular, Extracellular, and Restricted Diffusion for Cytometry in Tumour (VERDICT) MRI, luminal water imaging, and hyperpolarized C-13 MRI. Additionally, we will introduce several novel AI methods that have been proposed to improve MRI image quality, prostate lesion detection and characterization, as well as touch on the ethical implications of AI in medical imaging of the prostate.
Chromatin organization underlies essential genome functions, but its nanoscale organization remains challenging to capture and quantify with precision. Atomic force microscopy (AFM) offers direct structural readouts of DNA and chromatin, yet translating these rich images into reproducible biological metrics has been limited by the lack of standardized, scalable analysis tools. Here we present DNAsight, an automated analysis framework that integrates machine learning-based segmentation with modular, base-pair-calibrated quantification of DNA spatial organization, looping, nucleosome spacing, and protein clustering. Applied across diverse chromatin-associated proteins, DNAsight reveals protein-specific organizational signatures, including topology-dependent compaction by integration host factor, condition-dependent changes in loop-like DNA structures in cohesin-CTCF-precocious dissociation of sisters 5A reactions, and promoter-driven multimerization of GAGA factor clusters. The framework further enables direct extraction of nucleosome spacing distributions from raw AFM images, providing a label-free route to investigate chromatin fiber architecture. Together, these advances establish DNAsight as a generalizable and scalable approach for converting AFM measurements into quantitative insights into the physical principles of chromatin organization.
Psychological traits reflecting neuroticism, depressive symptoms, loneliness, and purpose in life are risk factors of AD dementia; however, the underlying biological mechanisms remain largely unknown. Using multi-omic data from the dorsolateral prefrontal cortex of 822 decedents in the Religious Orders Study and Rush Memory and Aging Project, we utilized a previously derived multi-omic brain molecular pseudotime representing molecular distance from no cognitive impairment (NCI) to AD dementia, and three distinct multi-omic brain molecular subtypes of AD dementia. We first confirmed generalizability of pseudotime and subtypes in two independent samples. We then annotated the subtypes, and explored whether they differed by neuropathologic burden, brain morphology or genetic risk, and found that while these indices differentiated all subtypes from NCI they did not differentiate amongst them. Finally, we tested for differential associations between the psychological traits and the subtypes, adjusting first for age, sex, education, and time to death, and then additionally for 9 common AD and Related Dementias pathologies. We found that in fully adjusted models, neuroticism, loneliness and purpose in life remained differentially associated with some AD subtypes relative to NCI. Our results are consistent with a two-stage model in which (i) upstream genetic risk influences overall disease liability, while (ii) intermediary psychological predispositions align more directly with subtype differentiation capturing AD-related heterogeneity not explained by neuropathology or brain atrophy. These results indicate that psychological risk factors may be associated with AD dementia via multi-omic molecular pathways, predominantly informed by metabolomic dysregulation, capturing heterogeneity not explained by neuropathology.
Pediatric low-grade gliomas (PLGGs) are generally slow-growing tumors associated with favorable long-term outcomes. However, their occurrence in early infancy is rare, particularly when arising in the posterior fossa with extensive dissemination and hydrocephalus. Advances in molecular profiling have identified specific genetic subtypes, including FGFR1-mutated gliomas, which may demonstrate more aggressive clinical behavior than suggested by histology alone. We report a four-month-old female who presented with signs of increased intracranial pressure and sunsetting eyes. Imaging revealed a heterogeneously enhancing exophytic medullary mass with a large cystic component causing tetraventricular hydrocephalus. Following ventriculoperitoneal shunt placement and surgical decompression, histology confirmed a low-grade glioneuronal tumor with low proliferative activity. Molecular analysis identified an FGFR1 mutation and 18q13 deletion, and methylation profiling classified the tumor within the MYB(L1)-family subtype B. Despite benign histologic features, the tumor progressed with cervical cord extension and diffuse spinal leptomeningeal metastases. Targeted therapy with trametinib achieved partial radiologic response before further progression. The patient remains clinically stable under ongoing therapy and multidisciplinary care. This case underscores the critical role of molecular diagnostics in risk stratification and treatment selection, particularly in infants with atypically aggressive PLGG.
Malignant pericardial effusion is a life-threatening manifestation of advanced non-small-cell lung cancer (NSCLC). Cardiac tamponade as an initial presentation is uncommon, and coexistent pathologies can complicate diagnosis and management. A 72-year-old man with mild chronic obstructive pulmonary disease (COPD) and a remote 2-pack-year smoking history was admitted with COVID-19 infection and possible superimposed pneumonia. Initial imaging revealed pericardial inflammation and a small effusion, raising concern for pericarditis. Subsequent echocardiography demonstrated a large pericardial effusion with tamponade physiology requiring pericardiocentesis, while cardiac magnetic resonance and computed tomography suggested possible malignancy. Cytology from pericardial and pleural fluid confirmed adenocarcinoma and bronchoscopy established stage IV epidermal growth factor-mutant NSCLC with cerebellar metastases. The patient experienced recurrent malignant effusions, managed successfully with balloon pericardiotomy and was treated with Osimertinib, resulting in sustained remission of effusions and clinical recovery. Careful review of multimodality imaging also revealed partial anomalous pulmonary venous return with significant left-to-right shunting, which may have exerted a protective hemodynamic effect on tamponade physiology. This may suggest a potential, though speculative, contribution of the anomalous pulmonary venous return to the observed tamponade physiology. At 2-year follow-up, the patient remains clinically and hemodynamically stable. This case illustrates the interesting duality of how multiple concurrent pathologies-including COVID-19 pneumonia, pericarditis, malignant pleuro-pericardial effusions, metastatic lung adenocarcinoma, and PAPVR-can coexist and probably shape the clinical course. The case underscores the indispensable role of multimodality imaging, molecular profiling, and minimally invasive pericardial interventions in guiding diagnosis, therapy, and prognosis in complex cardio-oncology presentations.
Emerging imaging strategies, such as tissue clearing and advanced light-sheet modalities, are transforming the visualization of cell differentiation, tissue reorganization, and morphogenesis in both model and non-model organisms, thereby driving future discoveries of conserved developmental programs and species-specific innovations. In this study, we present a versatile Bessel light sheet microscopy (BLX) system compatible with both immersion and capped air objectives, optimized for whole-mount three-dimensional imaging of cleared tissues. We demonstrate that autofluorescence-based light-sheet imaging enables reliable visualization of CUBIC-R-cleared mouse embryos, revealing microanatomical features such as nephrons during developmental stages. To support solvent-based clearing protocols, we designed a protective lens cap that shields objectives from solvent erosion and economically converts standard air objectives into high-performance immersion objectives. Using this capped BLX configuration, we achieved cellular-resolution imaging of whole PEGASOS-cleared mouse brains. The system's simplified design and open-source availability promote broader adoption for diverse applications in developmental biology.
Diabetic nephropathy (DN) is a major cause of end-stage renal disease, yet the molecular mechanisms driving tubular injury and fibrosis remain poorly defined. Here, we integrated single-cell multiplexed protein imaging, spatial transcriptomics, single-nucleus and single-cell RNA sequencing and chromatin accessibility profiling to comprehensively characterize human DN pathology. Our multi-modal analysis precisely maps kidney cell types and their spatial distributions, immune-fibrotic interactions, and key transcriptional regulators. We identified eight distinct cellular neighborhoods defining the immune-fibrotic microenvironment and uncovered molecular networks driving tubular injury and fibrosis. JUN (encoding c-Jun) emerged as a central regulator of transcriptional reprogramming during tubular injury and fibrogenic remodeling. In a diabetic mouse model, c-Jun is activated in injured proximal tubules. Using an inducible c-Jun mouse model, we demonstrated that tubular-specific c-Jun activation alone is sufficient to induce tubular injury, chronic inflammation, progressive fibrosis, and systemic metabolic alterations, including impaired glucose homeostasis. We also observed reduced expression of SLC4A4, a bicarbonate transporter essential for proximal tubular function, in injured tubules. Together, our findings establish a spatially resolved framework for understanding DN pathogenesis and identify c-Jun as a key mediator of tubular injury and fibrosis in diabetic kidney disease.
CRISPR-based diagnostics offer high sensitivity and specificity for nucleic acid detection, but their translation to point-of-care use remains limited by dependence on benchtop instrumentation, manual reagent handling, and costly optical components. To address these limitations, this study developed a fully integrated, low-cost digital microfluidic (DMF) system capable of automating the complete CRISPR/Cas12 workflow using on-board electronics and smartphone-based imaging. The system incorporates a programmable electrode array capable of precise droplet actuation for sample preparation and reagent mixing, a closed-loop heating module that maintains a stable reaction temperature of 39 °C, and a compact 3D-printed fluorescence imaging unit for end-point signal acquisition. To facilitate rapid and objective interpretation, we implemented a YOLOv11 deep learning model to classify fluorescence outputs into positive or negative results, achieving a mean average precision at 50% of 0.889. We applied the platform to the detection of Mycobacterium tuberculosis (MTB) DNA. The on-chip CRISPR assays reliably detected MTB across a dynamic range from 1 ng/μL down to 10-8 ng/μL, with no signal observed in no-template controls. Overall, the device delivers analytical performance comparable to conventional tube-based CRISPR assays while offering portability, reduced user intervention, and minimized risk of handling errors. These results highlight the potential of the integrated DMF-CRISPR system as a practical and accessible solution for point-of-care molecular diagnostics.
Human serum albumin (HSA), the most abundant plasma protein synthesized by hepatocytes, is vital for maintaining plasma colloid osmotic pressure, regulating redox balance, and transporting nutrients and metabolic wastes. Abnormal HSA levels correlate strongly with multiple diseases, especially liver disorders including cirrhosis, hepatocellular carcinoma and drug-induced liver injury (DILI), making HSA an essential biomarker for clinical diagnosis and prognosis. Although small-molecule fluorescent probes have been developed for HSA detection, major drawbacks persist: most work in the visible spectrum with poor tissue penetration, show low selectivity in complex biological environments, cannot realize subcellular localization or real-time in vivo monitoring. Herein, we fabricated a new near-infrared fluorescent probe TH from rhodamine derivatives for HSA detection. Relying on the twisted intramolecular charge transfer (TICT) mechanism, TH is weakly fluorescent in free form. When bound to the hydrophobic cavity of HSA, its molecular torsion is inhibited, producing prominent near-infrared fluorescence at 720 nm. TH possesses outstanding selectivity toward HSA, with a low detection limit of 2.3 μg/mL and favorable linear response. It also targets mitochondria, allowing real-time subcellular imaging of HSA in live cells. Further verified in a DILI mouse model, TH enables noninvasive real-time monitoring of hepatic HSA dynamics. This work offers a novel near-infrared probe for HSA analysis, holding great promise for auxiliary diagnosis and therapeutic assessment of DILI and other HSA-associated liver diseases.
Fluorescence-guided surgery (FGS) utilizes molecular contrast agents to highlight critical structures or pathological tissues in real time. The premise of FGS is to enable precise surgical decision-making through accurate visualization and quantitative assessment of fluorophore distribution. However, strong effects of diffusion and absorption of fluorescent light in tissue confound fluorescence images, preventing accurate quantitative assessment of the concentration and distribution of fluorescent markers. These optical artifacts may lead to misinterpretation of tissue boundaries and compromised surgical precision, thereby diminishing the capabilities of FGS. Resolving topological depth maps of fluorophore distribution at the millimeter scale is an important first step in performing quantitative sub-surface fluorescence imaging. In this study, we present a spatiotemporal deep learning architecture that utilizes picosecond single-photon avalanche diode (SPAD) sensor images to rapidly recover the depth topology of a fluorophore distribution embedded in diffuse media. The network is designed to work with wide-field, epi-illumination geometry and millimeter spatial resolution. A ConvLSTM-UNet deep learning network was developed for picosecond time-resolved image analysis. This network was trained on 5000 spatiotemporal maps simulated by the optical Monte Carlo method and convolved with the instrument response function (IRF) of the imaging system. The experimental setup utilized a SwissSPAD2 sensor synchronized with a 635 nm picosecond laser diode. Using only 10 selected temporal gates as input, the network could recover depth maps. Reconstruction accuracy was evaluated using mean error metrics across various depths and background concentrations of a fluorophore with a simulated decay time of 100 ps. A total of 75 different test fluorescence video data were evaluated. This set encompassed 15 unique inclusion shapes at five different depths. The network successfully reconstructed fluorescence topography up to 15 mm with a mean absolute error of less than 0.6 mm and mean depth variances below 0.5 mm. The inference time was ∼ 30    ms . Integrating temporal and spatial deep learning networks enabled depth mapping from time-resolved fluorescence data. Utilizing real IRF proved the applicability of SPAD sensors for sub-surface fluorescence mapping.
Different bioconjugation strategies are available for the cysteine (Cys)-specific functionalization of proteins with different payloads, including imaging probes and (pro-)-drugs. Most commonly applied linkers include maleimides (mal); however, because of the sometimes-observed instability of the formed thiosuccinimidyl linkage, its suitability for in vivo applications has been challenged. Consequently, several alternatives have been developed and compared to mal as a benchmark, yet examples of a direct comparison among new methodologies are scarce. We herein report a comparison of the use of mal, phenyloxadiazole methyl sulfone (PODS), and vinylketone (VK) as functional groups for the thiol-specific functionalization of human serum albumin (HSA) via its available free Cys34. Bifunctional chelating agents (BFCA) based on DFO*, identical in all regards but the functional group for bioconjugation, were prepared and conjugated to HSA and the obtained DFO*-HSA conjugates were radiolabeled with Zirconium-89 (89Zr) for positron emission tomography (PET). The efficiency of the conjugation of DFO*-X (X = mal, PODS, VK) to HSA differed significantly, with mal > PODS > VK. Stability studies of the 89Zr-labeled HSA-conjugates indicated good stability for [89Zr]-Zr-DFO*malHSA 11 and [89Zr]-Zr-DFO*-POD-HSA 12 in blood serum but only the latter was found stable in cell culture medium. [89Zr]-Zr-DFO*VK-HSA 13 was excluded from biological experiments due to its surprisingly low stability in all media tested. [89Zr]-Zr-DFO*-POD-HSA 12 was further investigated in CT26-tumor-bearing mice by PET/CT imaging and biodistribution studies. Specific uptake of radioactivity in tumors was high (up to 17% ID/g) and the tumors could be clearly visualized by PET at all time points with excellent tumor-to-background signal (tumor-to-blood ratio 3.2 ± 1.0 after 48 h p.i.). Unexpectedly, the uptake of radioactivity in bones was observed for [89Zr]-Zr-DFO*-POD-HSA 12. Overall, the in vivo performance of [89Zr]-Zr-DFO*-labeled HSA obtained by mal chemistry is the most promising candidate as a companion diagnostic PET imaging probe for the stratification of patients for therapies based on HSA-binding (pro-)-drugs.
Pituitary adenomas represent one of the most common intracranial tumors, and cavernous sinus invasion (CSI remains a major challenge for surgical management. Although the Knosp grading system provides a widely used radiological framework, its subjective nature and inter-observer variability limit diagnostic reliability. In recent years, advanced computational methods have been investigated to improve the preoperative prediction of invasion. This review synthesizes current evidence on the use of radiomics, machine learning (ML), and deep learning (DL) approaches in the detection and assessment of CSI in pituitary adenomas, with particular emphasis on their comparative performance against traditional imaging methods. Studies employing MRI-based radiomic feature extraction, ML classifiers, and convolutional neural networks were analyzed. Reported models commonly incorporated intensity, texture, and shape descriptors, or applied end-to-end DL architectures for automated prediction. Performance metrics such as accuracy, sensitivity, specificity, AUC, and Dice similarity coefficients were compared across studies, with Knosp grade serving as a frequent benchmark. Evidence suggests that ML and DL models consistently outperform conventional MRI interpretation in predicting CSI. Radiomics pipelines integrating quantitative imaging features with clinical variables achieved high diagnostic accuracy, while CNN-based models trained on contrast-enhanced MRI often exceeded AUC values of 0.85. Furthermore, automated segmentation frameworks demonstrated reliable delineation of tumor boundaries, facilitating improved assessment of invasive behavior. Despite promising outcomes, limitations such as small sample sizes, single-center designs, and lack of external validation restrict broad clinical adoption. Radiomics and AI-driven approaches show substantial potential for enhancing preoperative evaluation of pituitary adenomas with CSI. Standardized imaging protocols, multicenter collaborations, and transparent model validation are essential for future integration into neurosurgical decision-making.