Objective To investigate the clinical utility of deep learning-based multiparametric 1?F-FDG PET/CT radiomics for preoperative non-invasive prediction of histological subtypes in non-small cell lung cancer (NSCLC). Methods This retrospective study included 279 patients with pathologically confirmed NSCLC (187 adenocarcinomas and 92 squamous cell carcinomas) who underwent 1?F-FDG PET/CT at the Affiliated Cancer Hospital of Xinjiang Medical University from July 2018 to December 2024. The cohort was randomly assigned to training (n=195) and validation (n=84) sets at a 7:3 ratio. Tumor segmentation was performed using 3DSlicer, and segmentation consistency was evaluated using the intraclass correlation coefficient. Handcrafted radiomics features were extracted from PET and CT images using PyRadiomics, whereas deep learning features were extracted using a ResNet-50 architecture. Clinical, radiomics, and deep learning models were constructed using logistic regression, support vector machine, K-nearest neighbors, and ExtraTrees classifiers. The optimal model was subsequently integrated with clinical risk factors to develop a combined prediction model. Discrimination, calibration, and clinical utility were assessed using ROC curves, calibration curves, and decision curve analysis. Results Multivariate logistic regression analysis revealed that age, smoking history, lesion location, maximum standardized uptake value, and CYFRA21-1 were independent predictors for distinguishing lung adenocarcinoma from squamous cell carcinoma. In the training cohort, the clinical model, CT radiomics model, PET radiomics model, CT deep learning model, PET deep learning model, and combined model achieved areas under the curve (AUCs) of 0.941, 0.894, 0.865, 0.882, 0.879, and 0.958, respectively; the corresponding AUCs in the validation cohort were 0.922, 0.792, 0.868, 0.848, 0.873, and 0.948. Decision curve analysis indicated that the combined model provided a superior clinical net benefit. Conclusion Integrating clinical factors with 1?F-FDG PET/CT deep learning radiomics enables accurate preoperative non-invasive prediction of NSCLC histological subtypes and holds considerable promise for clinical application.

Objective To evaluate the clinical value of emerging ultrasound viscoelasticity imaging (VI) and its combination with shear wave elastography (SWE) in the staging of chronic kidney disease (CKD). Methods A total of 127 CKD patients who visited the First Affiliated Hospital of Bengbu Medical University between July 2025 and December 2025 were enrolled. Based on their glomerular filtration rate and the staging criteria of the Kidney Disease: Improving Global Outcomes guidelines, the patients were categorized into five stages. Using CKD stage 3 as the cutoff, they were divided into an early-to-mid stage group (CKD stages 1-3, n=71) and a late-stage group (CKD stages 4-5, n=56). On the day following admission, ultrasonography was performed to acquire baseline parameters. Under conventional two-dimensional ultrasound, the left renal cortical thickness as well as the length, thickness, and width of the left kidney were measured (kidney volume was calculated using an empirical formula). Superb microvascular imaging was then employed to measure the peak systolic velocity (PSV), end-diastolic velocity (EDV), and resistive index (RI) of the left interlobar artery (IRA). Subsequently, SWE and VI were used to obtain the maximum and mean values of the elasticity (E), viscosity (V), and dispersion (D) coefficients in the left renal cortex, denoted as E_max, E_mean, V_max, V_mean, D_max, and D_mean. Each parameter was measured five times, and the average value was taken as the final result.The differences in the above parameters between the early-to-mid stage group and the late-stage group were compared, and the SWE-derived and VI-derived parameters that demonstrated the highest diagnostic performance for the late-stage group were evaluated. Results In the early-to-mid stage group, RI, E_max, E_mean, V_max, V_mean, and D_max (P0.001), as well as D_mean (P=0.02), were significantly lower than those in the late-stage group, whereas renal volume, renal cortical thickness, PSV, and EDV were significantly higher (P0.001). RI, E_max, E_mean, V_max, V_mean, D_max, and D_mean were positively correlated with both CKD stage and serum creatinine level. Among them, V_max exhibited the strongest correlation (with CKD stage: r=0.658, P0.001; with serum creatinine: r=0.723, P0.001). In contrast, renal volume, renal cortical thickness, PSV, and EDV were negatively correlated with CKD stage and serum creatinine level, with renal cortical thickness showing the strongest negative correlation (r=-0.628, P0.001; r=-0.632, P0.001). For the identification of late-stage CKD, the AUCs of the SWE-derived and VI-derived parameters were as follows: E_max=0.806, E_mean=0.761, V_max=0.883, V_mean=0.808, D_max=0.741, D_mean=0.624. Statistically significant SWE and VI parameters from the univariate analysis were entered into a binary logistic regression model with forward stepwise selection. The results identified V_max and E_max as independent factors for late-stage CKD. The combined diagnostic model constructed using these two parameters yielded an AUC of 0.888, which was significantly superior to that of E_max alone (P=0.002), but did not differ significantly from that of V_max alone, the parameter with the highest individual performance (P=0.605). Conclusion VI holds diagnostic value for CKD staging, with V_max demonstrating the highest diagnostic performance. Combining the optimal VI parameter with the SWE parameter further enhances the diagnostic accuracy for CKD.

Objective To investigate changes in brain functional indicators in patients with minimal hepatic encephalopathy (MHE) using analysis methods such as amplitude of low-frequency fluctuations (ALFF), fractional amplitude of low-frequency fluctuations (fALFF), regional homogeneity (ReHo), and degree of centrality (DC). Methods A prospective study enrolled 25 patients diagnosed with MHE at the First Department of Hepatology of the Affiliated Hospital of Shaanxi University of Traditional Chinese Medicine from August to October 2025, constituting the MHE group. Additionally, 20 healthy individuals from the same period were selected as the healthy control (HC) group. ALFF, fALFF, ReHo, and DC analyses were employed to identify brain regions exhibiting statistically significant differences in brain functional indicators between the two groups. Results Compared with the HC group, the MHE group exhibited elevated ALFF values in the cerebellum, left parahippocampal gyrus, and right hippocampus, and reduced ALFF values in the right precuneus, left medial frontal cortex, and medial prefrontal cortex (P0.005 at the voxel level, cluster level P0.05, GRF-corrected); fALFF values were elevated in the right insula and reduced in the left infratemporal gyrus and right precuneus; ReHo values were elevated in the right orbitofrontal cortex and hippocampus, and reduced in the left angular gyrus, infratemporal gyrus, right supramarginal gyrus, and posterior central gyrus; DC values were elevated in the left superior frontal gyrus; they were reduced in the left lingual gyrus, with all differences being statistically significant (voxel-wise P0.005, cluster-level P0.05, GRF-corrected). Conclusion Patients with MHE exhibit widespread and complex patterns of cerebral functional disturbances, primarily involving the default mode network, but also affecting the limbic system, visual pathways, and sensorimotor cortex. Furthermore, these patients demonstrate regional functional compensation or an imbalance between inhibitory and excitatory mechanisms, providing a new perspective for understanding the complexity of neuropathological changes in MHE.

Objective To investigate the predictive value of a nomogram model integrating inflammatory markers and chest CT imaging features for extensive-stage small cell lung cancer (SCLC). Methods Clinical and imaging data of 122 patients with pathologically confirmed SCLC at the Second Affiliated Hospital of Anhui Medical University from March 2020 to November 2025 were retrospectively analyzed. According to the two-stage classification, patients were divided into limited-stage SCLC (LS-SCLC, n=38) and extensive-stage SCLC (ES-SCLC, n=84). The two groups were compared in terms of neutrophil to lymphocyte ratio (NLR), platelet to lymphocyte ratio, monocyte to lymphocyte ratio, systemic immune-inflammation index, and CT imaging features. Multivariate logistic regression analysis was performed to identify independent predictors of ES-SCLC, which were subsequently used to construct a nomogram. The diagnostic performance and clinical utility of the model were evaluated using ROC curves, calibration curves, and decision curve analysis. Results Multivariate logistic regression analysis showed that NLR (OR=4.022, 95% CI: 1.352-11.968, P=0.012), largest diameter (OR=2.029, 95% CI: 1.236-3.332, P=0.005), bronchial obstruction (OR=9.714, 95% CI: 2.448-38.547, P=0.001), and mediastinal lymphadenopathy (OR=8.901, 95% CI: 1.972-40.173, P=0.004) were independent risk factors for ES-SCLC. The nomogram incorporating these four predictors demonstrated excellent diagnostic performance with an AUC of 0.924 (95% CI: 0.879-0.969), a sensitivity of 83.3%, and a specificity of 89.5%. Bootstrap resampling validation confirmed good calibration, and decision curve analysis indicated significant clinical net benefit. Conclusion The nomogram model based on an inflammatory marker (NLR) and CT imaging features (largest diameter, bronchial obstruction, and mediastinal lymphadenopathy) can effectively predict ES-SCLC.

Objective To develop a preoperative prediction model based on two-dimensional ultrasound and The Bethesda System for Reporting Thyroid Cytopathology (BSRTC) categories, and to evaluate its predictive value for complete nodule absorption within 12 months after radiofrequency ablation (RFA). Methods Preoperative data from 112 patients with 131 nodules who underwent RFA in the First Affiliated Hospital of Bengbu Medical College from January 2022 to June 2024 were retrospectively analyzed. The variables included gender, age, initial nodule volume, internal composition, and the Chinese Thyroid Imaging Reporting and Data System (C-TIRADS) category. The volume reduction ratios at 1, 3, 6, 12 months after RFA were also assessed and recorded as VRR1, VRR3, VRR6, and VRR12, respectively. According to whether 100% VRR was achieved within 12 months after RFA, the nodules were divided into a complete absorption group (n=33) and an incomplete absorption group (n=98). Logistic regression analysis was performed to identify independent factors associated with absorption, and a prediction model was constructed accordingly. The model was visualized using a nomogram, and its performance was evaluated using the Hosmer-Lemeshow test, calibration curves with bootstrap validation, ROC curve analysis, and decision curve analysis. Results Univariate analysis showed that C-TIRADS category, BSRTC category, and initial nodule volume were significantly associated with absorption time (P0.05). Logistic regression analysis confirmed that C-TIRADS categories 4b and 4c, BSRTC categories V and VI, and initial nodule volume were independent risk factors for incomplete absorption (P0.05). The combined prediction model showed good model fit (Hosmer-Lemeshow test P=0.933). The calibration curve demonstrated good predictive accuracy, and ROC curve analysis showed strong discriminatory ability, with an AUC of 0.854. Decision curve analysis indicated that the model had good clinical utility. Conclusion The prediction model incorporating C-TIRADS category, initial nodule volume, and BSRTC category demonstrated good predictive performance. This model may help accurately predict nodule absorption after RFA and optimize clinical management strategies.

Objective To explore the alterations of whole-brain degree centrality (DC) in patients with primary dysmenorrhea (PDM) and the changes in DC before and after acupuncture treatment by resting-state functional magnetic resonance imaging (rs-fMRI). Methods A total of 25 patients with PDM (PDM group) and 27 healthy controls (HC group) were recruited from Shaanxi University of Traditional Chinese Medicine between October 2022 and October 2023. All participants underwent rs-fMRI scanning on days 1 to 3 of menstruation, and their clinical symptoms were assessed using the Visual Analogue Scale (VAS), COX Menstrual Symptom Scale (CMSS), Self-rating Anxiety Scale (SAS) and Self-rating Depression Scale (SDS). Patients in the PDM group received consecutive acupuncture at Guanyuan (CV4) and bilateral Sanyinjiao (SP6) one week prior to the next menstruation until menstrual onset. SPM and DPABI software based on the Matlab platform were adopted for data preprocessing. The dcm2nii tool was used to convert original DICOM data into NIFTI format, followed by slice-timing correction, head motion correction, spatial normalization, nuisance regression, linear drift removal, and band-pass filtering at 0.01-0.1 Hz. After preprocessing, the DPABI 3.0 toolkit was applied to calculate DC values of brain images. Brain regions with significant intergroup DC differences and DC changes in the PDM group before and after acupuncture were analyzed. Results There were no statistically significant differences in age, height and body weight between the PDM group and HC group (P0.05), whereas significant differences were observed in disease duration, VAS, CMSS, SAS and SDS scores (P0.001). The scores of VAS, CMSS, SAS and SDS during menstruation differed significantly in the PDM group before and after acupuncture treatment (P0.05). Compared with the HC group, the PDM group exhibited elevated DC values in the left middle temporal gyrus, left dorsolateral superior frontal gyrus and left posterior cingulate cortex. After acupuncture, the PDM group showed increased DC values in the left postcentral gyrus and left paracentral lobule, alongside decreased DC values in the left lenticular nucleus/caudate nucleus and right lenticular nucleus (Gaussian Random Field correction, voxel-level P0.001, cluster-level P0.05, two-tailed). Conclusion rs-fMRI combined with DC analysis reveals abnormal degree centrality in multiple brain regions of PDM patients. Acupuncture exerts neural effects mainly on pain conduction pathways. It can relieve clinical symptoms of PDM and modulate brain regional DC values, which provides imaging evidence for further elucidating the central pain and analgesic mechanisms underlying primary dysmenorrhea.

Objective To explore the value of a predictive model based on radiomics signatures from the venous phase of enhanced CT, combined with preoperative clinical independent risk factors, in non-invasively predicting lymphovascular invasion (LVI) in patients with advanced gastric cancer (AGC). Methods A retrospective analysis was conducted on 210 patients with pathologically confirmed AGC at the First Affiliated Hospital of Bengbu Medical University from January 2023 to July 2025, including 134 with LVI and 76 without LVI. Preoperative CT features of gastric cancer patients were analyzed to construct a radiomics model. A clinical prediction model was established through logistic regression analysis. The radiomics model and clinical risk factors were combined to construct a clinical-radiomics model. The discriminatory efficacy of the models was verified using the ROC curve and the area under the curve (AUC). The diagnostic efficacy was evaluated using calibration curves and decision curve analysis. Results Multivariate analysis identified lymph node metastasis (OR=5.01) and nerve invasion (OR=3.32) as independent risk factors for LVI (P=0.032). LASSO regression selected 9 key radiomics features to construct the Rad-score. In the validation set, the AUC values of the clinical model, radiomics model, and combined model were 0.751, 0.665, 0.797, respectively. The combined model was numerically superior to the clinical model, with good calibration (P=0.412), and decision curve analysis showed a higher clinical net benefit within a broader threshold range. SHAP analysis visualization indicated that lymph node metastasis and nerve invasion were the most significant features contributing to the model's prediction. Conclusion The venous phase CT radiomics combined with clinical risk factors model constructed in this study can effectively integrate quantitative imaging information with key clinical indicators, providing a new method with good efficacy and clinical translational potential for non-invasive, individualized prediction of LVI in AGC patients before surgery.

Objective To predict the risk of clinically significant prostate cancer (csPCa) by establishing a nomogram model based on shear wave elastography (SWE) and multiparameter magnetic resonance imaging (mpMRI). Methods A total of 145 patient from the Department of Urology, Affiliated Hospital of Yan 'an University, who were suspected of prostate cancer and underwent ultrasound-guided prostate biopsy from September 2020 to December 2022 were selected. They were divided into csPCa group (n=40) and non-csPCa group (n=105) based on the pathological results of biopsy. The difference in clinical indicator and imaging indicator between the two groups were compared. Multivariate logistic stepwise regression analysis was used to screen independent predictors of csPCa, and a nomogram model was constructed based on the independent predictor. Calibration curve was used to evaluate the goodness of fit of the model, ROC curve was used to evaluate the predictive power of the model, and decision curve analysis was used to evaluate the clinical application value of the model. Results Multivariate stepwise Logistic regression analysis showed that E_mean difference, PV and PI-RADS were independent predictors of csPCa risk. The nomogram model shows good calibration and discrimination ability, and the area under the ROC curve is 0.957 (95% CI:0.924-0.990). Decision curve analysis shows that the nomogram model has a good clinical net return for predicting csPCa risk when the threshold probability is in the range of 0 to 0.90. Conclusion The nomogram model based on SWE and mpMRI has good clinical prediction efficiency, and can provide a more objective and practical basis for making a reasonable diagnosis and treatment plan for patients with suspected prostate cancer.

Objective To investigate the predictive value of radiomics models developed from pre-neoadjuvant chemotherapy intratumoral and peritumoral features of primary breast lesions on dynamic contrast-enhanced magnetic resonance imaging for neoadjuvant chemotherapy response in breast cancer. Methods A total of 388 breast cancer patients with complete pretreatment dynamic contrast-enhanced magnetic resonance imaging who received neoadjuvant chemotherapy at the Shandong Medical and Pharmaceutical University Hospital from January 2018 to July 2025 were retrospectively enrolled. After lesion segmentation and radiomic feature extraction from intratumoral and peritumoral ROIs, feature dimension reduction and selection were performed sequentially using the minimum redundancy maximum relevance algorithm and least absolute shrinkage and selection operator regression. Three radiomics models (intratumoral, peritumoral, and combined intratumoral-peritumoral) were established to predict neoadjuvant chemotherapy response. Model discrimination was assessed via the area under receiver operating characteristic curve and decision curve analysis, and the calibration of the optimal model was verified by calibration curves. Results In total, 1361 intratumoral and 1286 peritumoral radiomic features were extracted from dynamic contrast-enhanced magnetic resonance imaging ROIs. Five predictive features were screened for the intratumoral model and another five for the peritumoral model; the combined model was built incorporating all 10 selected features. The combined model achieved AUCs of 0.913 in the training cohort and 0.846 in the validation cohort, superior to the intratumoral model (0.845, 0.771) and peritumoral model (0.788, 0.747). Decision curve analysis demonstrated superior net clinical benefit of the combined model, and its calibration curve revealed favorable consistency between predicted and observed response probabilities. Conclusion The intratumoral and peritumoral imaging-based radiomics model before chemotherapy can effectively predict the efficacy of neoadjuvant chemotherapy in breast cancer patients,which is expected to assist clinical decision-making.

Objective To explore the impact of virtual unenhanced (VUE) of spectral CT based on deep learning image reconstruction (DLIR) on image quality and lesion detection rates in contrast-enhanced abdominal CT scanning. Methods A total of 50 patients who underwent unenhanced and dual-phase (arterial and portal venous phases) contrast-enhanced abdominal CT scans using the spectral GSI mode from May to July 2025 were collected in Affiliated Hospital of Shaanxi University of Chinese Medicine. True unenhanced (TUE) images were reconstructed with ASiR-V 40% (120 kVp-like). Four groups of VUE images were reconstructed using medium and high level DLIR (DLIR-M and DLIR-H) based on the dual-phase contrast-enhanced data. The CT values, noise (SD), signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR) and lesion detection rates of the five groups were measured and compared, and the 5-point Likert method was adopted to evaluate the overall image. Inter-observer agreement was assessed using the Kappa test. One-way ANOVA analysis and Kruskal-Wallis H test were used to compare the subjective and objective evaluation indexes of the five groups.The lesion detection rates of the five groups were compared using the chi-square test. Results There was no statistically significant difference in CT values between TUE and the four VUE groups (P0.05). The four VUE groups had lower SD, higher SNR and CNR than TUE (all P0.05). Inter-observer agreement for image quality scores was good (all Kappa0.8), and the image quality score of the VUE_V-H group was the highest, showing a statistically significant difference compared with TUE (P0.001). There was no statistically significant difference in lesion detection rates among the five groups(P0.05). The effective dose (ED) of VUE combined with dual-phase scanning was reduced by approximately 33.35%, and the ED was (10.71±3.27) mSv. Conclusion The abdominal VUE image based on DLIR can significantly improve image quality without reducing the lesion detection rates. It is suggested that the VUE_V images reconstructed using DLIR-H have the potential value of replacing TUE scans to reduce radiation doses.

Objective To construct a fusion model based on multiparametric magnetic resonance imaging (mpMRI) radiomic features combined with clinical complete blood count for predicting lymphovascular space invasion (LVSI) in endometrial carcinoma (EC). Methods Clinical and imaging data of 108 EC patients from Northwest Women's and Children's Hospital were collected retrospectively. On T2WI fat-saturated sequences and diffusion-weighted imaging (DWI) sequences, the regions of interest (ROI) of tumors and the entire uterus were manually delineated on each transverse section. Tumor volume, uterine volume, and tumor/uterine volume ratio (TVR) were measured and calculated. Clinical indicators (including laboratory complete blood count indicators) and tumor morphological indicators were screened by univariate analysis, then fused with radiomic models to construct EC LVSI prediction models using random forest (RF) algorithm. Results Univariate analysis showed that clinical indicators such as age, carbohydrate antigen 199 (CA199), and lymphocyte ratio, as well as tumor morphological indicators such as TVR, DWI lesion volume, and axial tumor short diameter were significantly correlated with LVSI (P0.05). The area under the curve (AUC) of the radiomic model constructed based on DWI sequence for predicting LVSI was 0.80, and the AUC of the fusion model of radiomics and tumor morphology was 0.94. The RF model fused with screened clinical indicators, tumor morphological indicators, and radiomic features showed the optimal predictive performance, with AUC of 0.97, accuracy of 0.91, and specificity of 0.82. Its efficacy was significantly superior to the single radiomic model and the fusion model of radiomics and tumor morphology, and the calibration curve showed good consistency between predicted probabilities and actual conditions (P0.05). Conclusion The fusion model based on mpMRI radiomic features (including morphological features) combined with clinical indicators has reliable predictive value for LVSI in EC patients. It can effectively assist clinical preoperative evaluation of tumor invasion risk, thereby guiding the formulation of individualized surgical plans and helping optimize clinical treatment strategies.

Hepatocellular carcinoma is characterized by an insidious onset and remains challenging to diagnose at an early stage. Conventional serological biomarkers and morphological imaging modalities still exhibit limited sensitivity for the detection of early-stage tumors and small lesions. Glypican-3 (GPC3), a glycosylphosphatidylinositol-anchored heparan sulfate proteoglycan, is highly expressed in most HCC tissues while being scarcely detectable in normal adult liver tissues, making it an attractive target for molecular imaging and targeted therapy of HCC. Recent advances in radiolabeling strategies and ligand engineering have facilitated the development of various GPC3-targeted antibody-based and peptide-based radiotracers. These radiotracers have demonstrated favorable tumor-targeting capability and imaging performance in preclinical studies, and some candidates have already entered early-stage clinical evaluation with expanding applications in radionuclide theranostics. This review summarizes recent advances in GPC3-targeted radiotracers for the diagnosis and treatment of HCC, with particular emphasis on probe design, imaging characteristics, early clinical translation, and theranostic potential. The review aims to provide insights into the rational development and future clinical translation of GPC3-targeted radiopharmaceuticals.

Neuroendocrine prostate cancer (NEPC) is a highly aggressive and treatment-resistant variant that emerges following therapy for castration-resistant prostate cancer. Its pathogenesis is driven not only by tumor cell-intrinsic mutations but also critically by the tumor microenvironment (TME). This review summarizes how key cellular and non-cellular components within the TME of NEPC interact to collectively promote neuroendocrine differentiation and therapy resistance. Given the limited efficacy of current treatment options, this review further explores the promising prospects of immunotherapy, including immune checkpoint inhibitors and novel targeted therapies against antigens such as DLL3. A deeper understanding of the dynamic interactions between NEPC and its TME is crucial for developing effective combination immunotherapies to overcome treatment resistance and improve patient outcomes.

Knee osteoarthritis (KOA) is a highly prevalent degenerative joint disorder among the middle-aged and elderly population. Given the irreversible pathological alterations at its middle to advanced stages, early detection and prevention of disease progression have long been the cornerstone of clinical KOA management. MRI enables clear visualization of structural changes in the soft tissues of the knee joint; in particular, quantitative MRI allows for quantitative assessment of the biochemical properties of articular cartilage, providing a solid imaging foundation for the precise management of KOA. With advances in AI technology, the integration of deep learning with MRI has opened up new avenues for KOA diagnosis and treatment. Herein, we review the research progress of MRI combined with AI technology in the early diagnosis, progression prediction, total knee arthroplasty risk assessment, and perioperative adjuvant management of KOA. We also discuss the core bottlenecks of current technologies and outline prospects for future development, to provide a reference for the clinical translation and application of this integrated technology.

Gallbladder cancer (GBC), the most common malignancy of the biliary system, is often diagnosed at an advanced stage with a poor prognosis because of its atypical early symptoms and highly aggressive biological behavior. Although conventional two-dimensional ultrasound is the preferred imaging modality for GBC screening, it remains limited in differential diagnosis and refined assessment such as tumor staging. Multimodal ultrasound integrates high-resolution grayscale imaging, color Doppler hemodynamic assessment, contrast-enhanced ultrasound evaluation of microvascular perfusion, and shear-wave elastography assessment of tissue stiffness, thereby providing a multidimensional diagnostic framework and improving diagnostic accuracy. This review summarizes the current applications, limitations, and future directions of multimodal ultrasound in GBC diagnosis, with the aim of providing ultrasound diagnostic reference for early identification, risk stratification, and clinical decision-making.

Early identification of tumor lesions and implementation of individualized precision therapy have long been pivotal challenges and core objectives in medical research. Targeted nano ultrasound contrast agents achieve precise recognition of specific molecular markers within tumor cells or the tumor microenvironment by modifying the contrast agent surface with specific ligands, thereby enabling specific imaging of tumor sites. Targeted nanobubbles are emerging as significant contrast agents and drug delivery carriers to targeted regions. This paper provides a systematic review of recent advancements in the research and development of tumor-targeted nano ultrasound contrast agents, covering aspects such as construction methods, targeted modification strategies, and the synergistic application of imaging and therapy. Furthermore, it explores potential directions for future development in light of the key challenges encountered in clinical translation, aiming to offer theoretical insights and practical guidance for the design and optimization of novel multifunctional ultrasound diagnostic and therapeutic probes.

The pathological process of Alzheimer's disease (AD) begins 10-20 years before clinical diagnosis, making the development of non-invasive biomarkers for ultra-early identification and monitoring crucial. In recent years, the retina, serving as an in vivo, non-invasive, and high-resolution imaging extension of the central nervous system, has become a forefront focus in AD research. Studies suggest that retinal pathological changes can mirror core events in the AD brain. However, significant controversies remain regarding their specificity as early biomarkers, their temporal sequence, and their clinical translation pathways, necessitating systematic review and integration. This article reviews the theoretical basis of the retina as a "window to the brain" and its multimodal imaging findings in AD. It focuses on analyzing the latest evidence for multi-level biomarkers, ranging from retinal structural thickness and microvascular networks to molecular pathology and functional electrophysiology. The review elaborates on how high-resolution technologies like optical coherence tomography reveal early changes, and evaluates the potential value of retinal markers in the early and differential diagnosis of AD (e.g., distinguishing it from vascular dementia and glaucoma) and in predicting disease progression. This article aims to systematically delineate the current research landscape and controversies, providing a theoretical foundation and a clear pathway for the development and clinical translation of retinal imaging-based early biomarkers for AD.

Ultrasound elastography (UE) enables quantitative or semi-quantitative assessment of tissue biomechanical properties, providing important functional information for the noninvasive diagnosis and risk stratification of superficial organ diseases. Given the superficial anatomical location of the breast and thyroid, and the fact that their associated malignancies typically exhibit marked stiffness heterogeneity, UE has gradually evolved into an important adjunct to conventional ultrasound examination. This review systematically summarizes the clinical value of strain elastography (SE) and shear wave elastography (SWE) in breast and thyroid diseases, with particular emphasis on the significance of the stiff rim sign in differentiating benign and malignant breast lesions, the role of UE in the risk re-stratification of intermediate-risk thyroid nodules, as well as common technical pitfalls and confounding factors encountered in clinical practice. In breast imaging, UE can assist in optimizing Breast Imaging Reporting and Data System (BI-RADS) categorization and reducing unnecessary biopsies of benign lesions. In thyroid imaging, the Chinese Thyroid Imaging Reporting and Data System (C-TIRADS) has demonstrated important clinical value in the evaluation of category 4 nodules and nodules with indeterminate fine-needle aspiration cytology. In addition, recent advances in artificial intelligence (AI) for automated region-of-interest selection, real-time quality control, and multimodal diagnosis are summarized. In the context of domestic multicenter studies and guideline development, the future direction of UE toward greater standardization, reproducibility, and precision is further discussed.

Tumor microenvironment hypoxia is a key driver of malignant progression in glioma, which affects tumor invasiveness, chemoradiotherapy tolerance, and patient prognosis. Accurate evaluation of hypoxia is therefore critical for individualized treatment decision-making. In recent years, imaging-based assessment of tumor hypoxia has become a prevailing research trend. Starting from the multimodal imaging technical system, this paper systematically analyzes the research value of the technical principles of diffusion kurtosis imaging, dynamic contrast-enhanced magnetic resonance imaging, blood oxygen level-dependent magnetic resonance imaging, intravoxel incoherent motion imaging, and magnetic resonance spectroscopy in the evaluation of tumor hypoxia. It also discusses the cutting-edge advances of emerging techniques including bimodal imaging, nanotechnology, MRI habitat analysis, and nuclear medicine. Future research directions focus on multiparameter combined imaging, artificial intelligence-assisted habitat analysis, and integrated diagnosis and therapy with nanoprobes. Establishing reliable correlations among imaging, pathology, and molecular signatures is the key to promoting clinical translation.

The pre-stage of chronic obstructive pulmonary disease (COPD) (Pre-COPD) represents a critical window for the onset and progression of COPD, where early accurate identification and intervention are pivotal for slowing or even halting disease advancement. Currently, single-modality data exhibit limitations in Pre-COPD diagnosis and prognostic evaluation, whereas the "multimodal feature fusion" strategy integrating clinical, imaging, and biological data has emerged as a cutting-edge research approach. This article systematically reviews the conceptual framework of Pre-COPD, types of multimodal features, development of precision diagnostic models, stratified prognostic management, existing technical bottlenecks and solutions, and future research directions. It aims to provide systematic theoretical references and practical guidance for early accurate identification, risk stratification, and personalized prognostic management of Pre-COPD, thereby advancing COPD prevention and treatment strategies and facilitating the implementation of precision medicine.