Objective To investigate the value of a combined model integrating ultrasound radiomics and clinical characteristics in evaluating clinical pregnancy after frozen embryo transfer. Methods A total of 137 women who underwent frozen embryo transfer at the First Affiliated Hospital of Anhui Medical University from October 2023 to March 2024 were enrolled in this study. Baseline clinical data were collected, and ultrasound images and data were acquired 1 day before embryo transfer. Independent clinical predictors were selected using logistic regression analysis to construct a clinical model. Radiomics features of the endometrium and junctional zone were extracted from ultrasound images to build a radiomics model and generate a radiomics score (rad-score). A fusion model was subsequently developed by combining the rad-score with the independent clinical predictors. All models were trained and validated using five-fold cross-validation. ROC curves were constructed to compare the predictive performance of the three models for pregnancy outcomes. The Shapley additive explanations (SHAP) method was applied to interpret the contribution of each feature to the predictive results. Results Endometrial blood flow resistivity index (P=0.024) and pulsatility index (P=0.035) were identified as independent clinical predictors. ROC curve analysis demonstrated that the fusion model achieved the optimal predictive performance. The mean area under the curve (AUC) values of the training and validation cohorts were 0.822 and 0.821, respectively. Conclusion The combined model established based on clinical data and ultrasound images can effectively predict pregnancy outcomes. With the SHAP method, clinicians can better understand and interpret the predictive outcomes.

Objective A three-classification model based on APT imaging radiomics was constructed to accurately distinguish normal controls (NM), early-stage and middle-to-late-stage Parkinson's disease (PD). Through explainability analysis, the diagnostic value of key brain regions and features was clarified. Methods A total of 99 subjects from the Second Affiliated Hospital of Xinjiang Medical University from January 2022 to January 2024 were retrospectively enrolled, including 37 healthy controls, 36 patients with early-stage PD, and 26 patients with advanced-stage PD. All subjects underwent brain APT sequence scanning.Six brain nuclei, caudate nucleus (CN), putamen (PUT), globus pallidus (GP), red nucleus (RN), substantia nigra (SN), and nucleus accumbens (NAc), were manually segmented to extract 107 radiomic features. Key features were selected to construct diagnostic models using six machine learning algorithms. Model performance was evaluated using area under the receiver operating characteristic curve (AUC) and accuracy. Shapley additive explanations (SHAP) analysis was employed to decipher model decision logic and quantify feature contributions across disease stages. Results After three-step feature screening, 15 key radiomics features were identified. The combined LR model demonstrated optimal performance: training set macro-AUC=0.889 (95% CI: 0.827-0.943), micro-AUC=0.895 (95% CI: 0.837-0.946); The test set macro-AUC was 0.859 (95% CI: 0.707-0.975) and micro-AUC was 0.854 (95% CI: 0.704-0.967), significantly outperforming other models. SHAP analysis revealed key feature contribution patterns across PD stages: SN and RN features were critical for early-stage and normal classification, while GLCM autocorrelation coefficients of the PUT nucleus and RN features were core contributors for mid-to-late stage classification. Conclusion The LR combined model based on APT radiomics effectively achieves PD three-category diagnosis and staging. SN, RN, and PUT nuclei serve as core imaging biomarkers for PD pathological progression. SHAP analysis clearly elucidates the model's decision-making mechanism, providing an imaging tool for PD precision diagnosis and treatment that combines performance with interpretability.

Objective To develop and validate a CT-based radiogenomics model driven by metabolic reprogramming for the noninvasive prediction of immunotherapy efficacy in patients with non-small cell lung cancer (NSCLC). Methods This study integrated transcriptomic, clinical, and CT image data. Differentially expressed genes (DEGs) associated with metabolic reprogramming in NSCLC were identified using The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) databases. A metabolic reprogramming risk score (MRRS) model was established via Cox regression analysis, and its correlation with the tumor immune microenvironment was evaluated. Radiomic features were extracted from CT images of NSCLC patients in The Cancer Imaging Archive (TCIA) utilizing PyRadiomics. Following feature selection via LASSO regression, a radiogenomics model with MRRS as the target variable was constructed and evaluated. Furthermore, an independent validation cohort comprising 206 patients with advanced NSCLC who received immunotherapy was enrolled from Shanxi Provincial People's Hospital. ROC curves were employed to quantify the predictive performance of the model for immunotherapy efficacy. Results A total of 156 metabolic reprogramming-related DEGs were identified. From these candidates, nine key genes were selected to construct the MRRS model. In the TCGA-NSCLC training set, the AUCs for predicting 1-, 3- and 5-year overall survival were 0.638, 0.685 and 0.648, respectively, which were higher than those based on conventional clinical parameters. Patients in the high-risk group exhibited decreased immune cell infiltration (P<0.01) and poorer overall survival (P<0.001). Subsequently, six optimal radiomic features were selected from CT images to formulate the radiogenomics model, which achieved AUCs of 0.742 and 0.726 in the training and test sets for predicting MRRS, respectively. In the independent immunotherapy cohort, the radiogenomics model yielded an AUC of 0.704 for predicting treatment efficacy, effectively distinguishing responders from non-responders. Conclusion The proposed CT-based radiogenomics model enables the noninvasive assessment of metabolic reprogramming status in NSCLC. Furthermore, it demonstrates promising clinical potential for predicting the efficacy of immunotherapy.

Objective To develop a nomogram model for predicting vessels encapsulating tumor clusters (VETC)-positive hepatocellular carcinoma (HCC) based on preoperative contrast-enhanced CT imaging features combined with clinical characteristics. Methods Clinical and radiological data of 174 patients diagnosed with HCC by pathology admitted to Nanfang Hospital, Southern Medical University from October 2019 to April 2022, were retrospectively collected. The patients were randomly divided into a training cohort (n=121) and a validation cohort (n=53) in a 7:3 ratio. Based on CD34 immunostaining results, the percentage of tumor area occupied by the VETC pattern (VETC index) was calculated, and patients were classified into the VETC-positive group (≥55%) and the VETC-negative group (<55%) using this threshold. Independent predictive factors significantly associated with VETC-positive HCC were identified using univariate and multivariate logistic regression analyses, and a nomogram prediction model was constructed. The diagnostic performance of the model was evaluated using the ROC curve, and its calibration was assessed by calibration curves. Results Non-rim arterial phase hyperenhancement, enhancing capsule, intratumoral artery, and arterial peritumoral enhancement were identified as independent predictive factors for VETC-positive HCC (P<0.05). The nomogram model established by these factors demonstrated good performance, with the AUC of 0.815 in the training cohort and 0.760 in the validation cohort. The predicted probabilities from the nomogram model demonstrated high consistency with the actual incidence rates. Conclusion The nomogram model based on preoperative contrast-enhanced CT imaging features exhibits favorable predictive performance for VETC-positive HCC, which may facilitate the identification of high-risk individuals with HCC and inform clinical diagnosis and treatment decisions.

Objective To develop a multivariate model based on ultrasound features for accurate qualitative diagnosis of superficial lymph nodes involved by lymphoproliferative disorders (LPD). Methods A retrospective analysis was conducted on data from 209 patients who underwent lymph node biopsy with histological, immunohistochemical, and genetic testing at West China Hospital of Sichuan University from January 2022 to December 2024, and their clinical information and ultrasound images were collected. Two experienced sonographers independently interpreted and recorded the imaging features. Using pathology as the gold standard, the included patients were divided into benign group and malignant group, univariate analysis identified significant features, which were then used to build a multivariate logistic regression model to evaluate diagnostic performance. Subgroup analyses were performed in benign and malignant groups to assess associations between ultrasound findings and pathological diagnoses. Results A total of 209 patients were included (74 malignant, 135 benign). Univariate and multivariate analyses identified four predictive variables: age, short axis, cortical heterogeneity and vascular distribution. The multivariate model achieved an AUC of 0.826, and demonstrated significantly superior diagnostic performance compared to conventional ultrasound (P<0.001), particularly for cervical lymph nodes (P<0.05). Features significantly differentiating reactive hyperplasia from benign LPD included lymph node size and vascular distribution (P<0.05). Features showing significant differences between Hodgkin and Non-Hodgkin lymphoma included age and size (P<0.05). Conclusion The multivariate model based on ultrasonic features improves the qualitative diagnostic efficacy of superficial lymph nodes involved by LPD, and sonographic characteristics demonstrate potential for differentially diagnosing benign and malignant subtypes of superficial lymph node diseases.

Objective To investigate the association of maximum stiffness (E_max) measured by shear wave elastography (SWE) and the resistive index (RI) obtained by pulsed Doppler ultrasonography (PDU) with molecular subtypes of breast cancer. Methods A total of 137 patients with breast cancer (137 lesions) who underwent preoperative breast ultrasound and subsequent surgical resection from January 2024 to May 2025 were included. SWE parameters (E_max, E_min, and E_mean), PDU-derived RI, and pathological data, including histological type, grade, hormone receptor status, HER-2 status, and Ki-67 proliferation index, were analyzed. Univariate analysis and multiple linear regression were performed to evaluate the relationship between imaging parameters and pathological findings. Results Among the 137 lesions, invasive ductal carcinoma was the most common histological type, and Luminal A type was the predominant molecular subtype. Triple-negative type (TNBC) lesions exhibited significantly higher E_max values compared to Luminal A type, Luminal B type, and HER-2-positive type. The highest RI values were observed in TNBC, followed by HER-2-positive, Luminal B, and Luminal A subtypes. Multiple linear regression analysis indicated that tumor size, histological grade, and molecular subtype were independent predictors of both E_max and RI (P<0.05). Conclusion Increased tumor stiffness and higher resistive index may be associated with aggressive histopathological features in breast cancer. Softer tumors tend to correlate with the Luminal A subtype, whereas stiffer tumors are more likely to be TNBC.

Objective To elucidate the correlation between ultrasonographic features and mitochondrial oxidative stress-related gene expression in invasive ductal breast carcinoma. Methods A total of 38 patients diagnosed with invasive ductal carcinoma of the breast from January 2022 to August 2023 at the First Hospital of Shanxi Medical University were enrolled in this study. All participants had complete clinical records, multimodal ultrasound imaging data, and whole transcriptome sequencing profiles. To identify mitochondrial oxidative stress-related genes in breast cancer, we integrated differentially expressed genes from TCGA database with mitochondrial gene sets from MitoCarta 3.0 and oxidative stress-related genes from GeneCards. Key gene subgroups were then identified through Kaplan-Meier survival analysis, STRING protein-protein interaction network analysis, and Cytoscape module analysis. The expression levels of these key genes were subsequently extracted from the 38 breast cancer patients, and their correlations with clinical parameters and ultrasound features were systematically analyzed. Results Nine key genes were identified and classified into two subgroups: subgroup 1 (MRPL12, MRPL13, MRPS12, MRPL14, NME3, HSPE1) and subgroup 2 (DNA2, POLQ, RECQL4). Regarding ultrasound features, MRPS12 and MRPL14 expression was significantly upregulated in tumors with non-spiculated margins. In lesions exhibiting UMA-CPP>10%, MRPL12, MRPL13, MRPS12, and MRPL14 were significantly overexpressed, whereas DNA2 expression was downregulated. MRPL12 expression was notably elevated in cases with Alder grade III and perivascular distribution (P<0.05). Correlation analysis revealed that PI was positively correlated with MRPL13 and MRPS12, while AUC showed positive correlations with MRPL13 and HSPE1. AT was negatively correlated with MRPS12 and MRPL14, and TTP was negatively correlated with MRPS12. Regarding elasticity parameters, E_mean was positively associated with MRPL12 and MRPS12; E_max with MRPL12 and NME3; E_min with MRPL12; and E_sd with MRPL12, MRPL13, and MRPS12. Conversely, E_sd demonstrated a negative correlation with DNA2 (P<0.05). Notably, the majority of ultrasound features significantly associated with key genes were predominantly concentrated among four mitochondrial ribosomal protein genes MRPL12, MRPL13, MRPS12, and MRPL14 and exhibited consistent directional trends. Conclusion Mitochondrial oxidative stress-related gene expression correlates with specific ultrasound features of invasive ductal breast carcinoma, providing a non-invasive imaging basis for assessing mitochondrial oxidative stress within the tumor microenvironment.

Objective To investigate the imaging characteristics of ¹⁸F-FDG PET/CT in patients with adult-onset Still's disease (AOSD) and evaluate their association with disease activity. Methods We retrospectively analyzed ¹⁸F-FDG PET/CT imaging features, clinical manifestations, and laboratory results of 23 patients with AOSD who were treated at Northern Theater General Hospital from May 2016 to December 2024. We calculated the maximum standardized uptake value ratios of the spleen, bone marrow, and lymph nodes to the liver (SLR_max, BLR_max, and LyLR_max, respectively). Spearman's rank correlation analysis was performed to evaluate the relationships between these metabolic parameters and laboratory indicators as well as clinical activity scores. Multiple comparisons were adjusted using false discovery rate (FDR) correction, with a corrected q-value <0.05 considered significant. Inter-rater agreement was evaluated using the intraclass correlation coefficient). Results All 23 patients demonstrated abnormal FDG uptake, primarily involving lymph nodes (100%), most commonly in cervical, axillary, supraclavicular, and inguinal regions. Increased splenic uptake was observed in all patients (100%), while bone marrow involvement was identified in 21 patients (91.3%). Additional uptake sites included the liver (n=1, 4.3%), shoulder joints (n=4, 17.4%), submandibular glands (n=4, 17.4%), tonsils (n=3, 13.0%), and cutaneous nodules (n=2, 8.7%). The mean liver SUV_max was 3.6±0.5, while the spleen and bone marrow showed SUV_max values of 4.1±0.7 and 4.4±0.9, respectively. The median lymph node SUV_max was 4.8(4.1, 12.5), with corresponding SLR_max, BLR_max, and LyLR_max values of 1.14±0.09, 1.25±0.31, and 1.42(1.09,3.42). Clinical activity system scores ranged from 5-10[7(6-8)]. SLR_max and LyLR_max showed positive correlations with CRP (ρ=0.471 and 0.560, respectively, uncorrected P<0.05). BLR_max and LyLR_max were also positively correlated with the clinical activity system scores (ρ=0.432 and 0.416; uncorrected P<0.05). None of these correlations remained statistically significant after FDR correction. Additionally, lymph node SUV_max was positively correlated with lymph node length (ρ=0.518, P<0.05). Conclusion ¹⁸F-FDG PET/CT can comprehensively depict the extent of systemic involvement in AOSD. Although the correlations between metabolic parameters and clinical indicators did not reach statistical significance after FDR correction, the observed trends still suggest a potential association between them.

Objective To investigate the correlation between quantitative parameters of amide proton transfer-weighted (APTw) magnetic resonance imaging and p53 protein expression in breast cancer, as well as its predictive value for TP53 gene mutation status. Methods Data from 72 patients with pathologically confirmed breast cancer at Guangdong Second Provincial General Hospital between March 2024 and September 2025 were retrospectively collected. All patients underwent 3T MRI including conventional sequences and APTw scanning. Based on immunohistochemical results of p53 protein, patients were divided into the TP53 mutant group (n=42) and wild-type group (n=30). The TP53 mutant group was further subdivided into missense mutation subgroup (n=32) and nonsense mutation subgroup (n=10). APTw images were processed using the CEST-Matlab program. Regions of interest (ROI) of lesions were manually delineated in ImageJ software, and APT_mean and APT_max were measured and compared between groups. Spearman correlation analysis was performed to evaluate the correlation between APT parameters and p53 protein expression, and receiver operating characteristic (ROC) curves were plotted to analyze diagnostic efficacy. Results APT_mean and APT_max in the TP53 mutant group were significantly higher than those in the wild-type group (P<0.001). APT_mean showed a moderate positive correlation with p53 protein expression (r=0.48, P<0.001). No significant differences in APT parameters were found between the TP53 missense and nonsense mutation subgroups (P>0.05). The area under the ROC curve(AUC) of APT_mean in diagnosing TP53 mutation was 0.838, with a sensitivity of 92.9% and specificity of 66.7% at the cutoff value of 2.135%. Conclusion Quantitative APTw parameter (APT_mean) exhibits a moderate positive correlation with p53 protein expression in breast cancer and has favorable predictive efficacy for TP53 mutation, which is expected to provide an imaging basis for individualized treatment and prognostic evaluation of breast cancer.

Objective To evaluate the predictive value of CT imaging features in the diagnosis of secondary pulmonary tuberculosis (PTB) and to develop a diagnostic model based on multidimensional imaging features for assessing its performance. Methods A retrospective analysis was performed on clinical and CT imaging data from 246 patients with pulmonary shadows admitted to Guangzhou Chest Hospital from January 2023 to December 2024. The cohort comprised 160 patients with secondary PTB (PTB group) and 86 with non-tuberculous pulmonary conditions (non-PTB group). Binary logistic regression analysis was used to identify risk factors and construct a combined diagnostic model, and receiver operating characteristic (ROC) curve analysis was performed to evaluate model performance. Results Univariate analysis revealed that halo sign (OR=4.196, P=0.008), centrilobular nodules (OR=88.290, P=0.001), consolidation (OR=3.260, P=0.013), calcification (OR=2.547, P=0.031), and pleural thickening (OR=2.762, P=0.017) as risk factors for secondary PTB, while lesion distribution in the right middle lobe (OR=0.352, P=0.044) and left lingular segment (OR=0.190, P=0.003) were protective factors.Multivariate analysis identified seven independent predictors (halo sign, centrilobular nodules, consolidation, calcification, pleural thickening, right middle lobe involvement, left lingular segment involvement) . The combined model achieved an area under the curve (AUC) of 0.881 (95% CI: 0.834-0.929), with a sensitivity of 87.5% and specificity of 79.1%, demonstrating superior diagnostic performance compared to any single imaging feature (DeLong test Z=3.89-10.50, all P<0.001). At the optimal threshold of 0.601, the positive and negative predictive values were 73.2% and 90.3%, respectively. Conclusion The combined diagnostic model based on multidimensional CT imaging features enhances the identification of secondary PTB and holds promise as a valuable tool for clinical auxiliary diagnosis.

Objective To compare the effect of different reconstruction algorithms on CT angiography (CTA) image quality, and to explore the advantages and potential of deep learning image reconstruction(DLIR)in optimizing CTA. Methods Apex CT was used to scan the QSP Phantom which containing 9 test tubes. The target test tubes were 1 pure water and 5 iodine solution (one 3.75, 7.5, 15 mgI/mL each, and two iodine solution 30 mg/mL). The original scanning data were reconstructed using the adaptive statistical iterative reconstruction V (ASiR-V) 0%-100% (interval 10%) and three kinds of weight Deep Learning Image Reconstruction (High/Middle/Lower-strength DLIR, DLIR-L/M/H) ,14 algorithms reconstructed images with layer thickness of 1.25 mm for comparative analysis. 9 fixed layers in the target test tube were selected to place region of interest (ROI), standard deviation (SD) value of pure water was used as background, and the SD of CT value of pure water and iodine solution in the target test tube was measured. The effects of different reconstruction algorithms on CT value, noise and contrast noise ratio (CNR) of pure water and iodine solution images were calculated and compared. The mean CT value, noise and CNR of 30 mgI/ml iodine solution were compared between the center and the edge of the phantom. Results There were no statistical differences in CT values between pure water and iodine concentration solutions of different reconstruction algorithms (P>0.05). Compared with 30 mgI/ml iodine solution at the center and the edge of the phantom, CT values and CNR decreased, and the difference was statistically significant (P<0.05). There was no statistical difference in the noise of ASIR-V 90% and ASIR-V 100% reconstructed images (P>0.05). The noise of various iodine concentration solutions decreased with the increase of the weight of ASIR-V and DLIR, and the noise of DLIR-H was the lowest. There was no significant difference in the noise of the reconstructed image between ASIR-V 60% and DLIR-L (P>0.05), and there was no significant difference in the noise of the reconstructed image between ASIR-V 80% and ASIR-V 90% and DLIR-M (P>0.05). Except that there was no statistical difference in CNR of reconstructed images between 0%-60% and 10%ASIR-V weights, CNR of various iodine concentration solutions increased with the increase of ASIR-V and DLIR weights, and the CNR of DLIR-H was the largest. There were no significant differences in the CNR of ASIR-V 60%, ASIR-V 70% and DLIR-L reconstructed images (P>0.05), and no significant differences in the CNR of ASIR-V 90% and DLIR-M reconstructed images (P>0.05). Conclusion The DLIR reconstruction algorithm can improve the image quality of CTA without changing the CT value, DLIR-H with the lowest noise of and the highest CNR.

Objective To develop a closed-loop healthcare management pathway for prediabetes based on wearable devices and to evaluate its effectiveness in improving glycemic control. Methods A prospective single-arm study was conducted. A total of 125 patients with pre-diabetes diagnosed at the Health Management Center of Nanfang Hospital, Southern Medical University from April to December 2024, and at the Physical Examination Center of Linzhi People's Hospital from January to April 2025, were selected as the study subjects. Continuous glucose monitoring was performed using smart wearable devices. Individualized dietary and exercise prescriptions were delivered, and remote monitoring and follow-up management were conducted via a smart healthcare cloud platform, forming a closed-loop management pathway. The subjects were divided into three groups based on age: 25-50 years (n=31), 51-55 years (n=43), and ≥56 years (n=51). Additionall, patients were divided into a well-controlled group (<6.1 mmol/L, n=59) and a poorly controlled group (≥6.1 mmol/L, n=66) based on baseline fasting plasma glucose levels. The clinical characteristics of patients with different glycemic control statuses were compared. A linear mixed-effects model was used to analyze dynamic changes in blood glucose and their influencing factors. Results After the 12-week intervention, patients' fasting blood glucose decreased from 6.10 mmol/L at baseline to 5.00 mmol/L (P<0.001), and the glycemic control rate increased from 47.2% to 83.2% (P<0.001). The mixed-effects model showed a significant decrease in fasting blood glucose over the intervention time (β=-0.07, P<0.001). Compared with the 25-50 years group, patients aged 51-55 years (β=0.36, P=0.037) and ≥56 years (β=0.34, P=0.040) had higher fasting blood glucose levels; patients with a medical history or medication use had significantly higher fasting blood glucose than those without (β=0.46, P=0.003). Conclusion The closed-loop healthcare management pathway based on wearable devices can effectively reduce blood glucose levels in prediabetic patients, achieving personalized and precise management, and provides a scalable new strategy for the early prevention and control of diabetes.

Postoperative analgesia after traditional open and laparoscopic surgeries primarily relies on opioid-based patient-controlled analgesia, which can lead to a series of adverse effects, such as postoperative nausea and vomiting, and postoperative delirium, seriously compromising patient health. With the successful application and widespread promotion of the enhanced recovery after surgery (ERAS) concept, as well as the development of visualization techniques, multimodal analgesia has made significant progress. It offers clear advantages in promoting accelerated patient recovery and ensuring perioperative safety. This article elaborates on the recent research progress of visualized nerve blocks in postoperative analgesia for gastric surgery, summarizing the mechanisms of postoperative pain after gastric surgery, standards for evaluating postoperative analgesia, commonly used nerve block drugs and their combinations, as well as various nerve block methods for gastric surgery, aiming to provide patients with a more rational postoperative pain management plan.

Glioblastoma (GBM) is the most aggressive primary malignant tumor of the central nervous system, necessitating multimodal imaging for accurate diagnosis and treatment. While MRI provides superior soft-tissue contrast, it has significant limitations in delineating true tumor boundaries, differentiating true progression from pseudoprogression, and assessing metabolic activity. PET/CT offers complementary functional and molecular imaging by characterizing tumor metabolism, though its limited spatial resolution constrains anatomical precision. The integrated imaging strategy combining tyrosine metabolism PET/CT and multiparametric MRI enables comprehensive assessment of GBM biology through the fusion of morphological, functional-metabolic, and molecular information. This review systematically explores integrated imaging strategies combining tyrosine metabolism PET/CT, primarily using 1?F-FET and 11C-MET, with multiparametric MRI. It focuses on the application value of this multimodal methodology in critical aspects of GBM management, including non-invasive preoperative grading, accurate delineation of tumor biological margins, optimization of radiotherapy planning, and early assessment of therapeutic efficacy and recurrence. By synthesizing these advances, this article offers a perspective for GBM management grounded in multimodal molecular imaging.

Peripheral nerve sheath tumors (PNSTs) represent a significant subtype of soft tissue neoplasms. Different clinical management and prognosis are associated with distinct subtypes, making accurate differentiation crucial. MRI, known for its high soft-tissue resolution, has significantly enhanced its efficacy in the differential diagnosis of PNSTs through the integration of multimodal MRI and radiomics.Therefore,This review summarizes the advancements in structural and functional MRI, molecular and metabolic MRI, and radiomics for the differential diagnosis of PNSTs, providing references and directions for clinical management and future research.

In recent years, artificial intelligence, particularly radiomics and deep learning technologies, has achieved remarkable progress in the precise diagnosis and treatment of breast cancer. However, most existing studies and reviews have primarily focused on improving model performance while neglecting the issue of limited interpretability, which gives rise to the "black-box" problem and hinders the clinical adoption of artificial intelligence systems. To address this limitation, the present review adopts the perspective of explainable artificial intelligence (XAI) and provides a systematic overview of recent advances in XAI-based approaches for breast cancer imaging diagnosis and treatment. Special emphasis is placed on the latest developments in explainable radiomics and interpretable deep learning. This review aims to offer new insights for advancing precision medicine in breast cancer.

Minimal hepatic encephalopathy (MHE) is a common central nervous system complication in patients with cirrhosis, whose core feature is impaired cognitive function. In recent years, brain network science has provided a new perspective for revealing the neuropathological mechanisms of MHE. Damage to the default mode network (DMN), the most active brain network during rest, is considered a core mechanism underlying the cognitive impairment in MHE. Resting-state functional magnetic resonance imaging (rs-fMRI) allows for the non-invasive and precise assessment of both structural damage and functional dysfunction of the DMN in MHE patients. This review focuses on recent research progress regarding DMN impairment in MHE patients. It systematically summarizes the abnormal changes in the DMN and their association with cognitive deficits from multiple dimensions, including intra-DMN brain activity, within-network functional connectivity, and aberrant interactions with other brain networks (e.g., the salience network and central executive network). The objective of this review is to provide an objective neuroimaging basis for the early identification, understanding of pathological mechanisms, and treatment efficacy evaluation of MHE.

Ovarian cancer is one of the most common cancers of the female reproductive system and the main cause of cancer death in women. Therefore, early diagnosis and effective treatment monitoring are critical to improve the survival rate of patients. At present, spectral CT, as a new technology, can obtain iodine concentration, effective atomic number, spectral curve slope and other parameters through post-processing, which is helpful for accurate qualitative and quantitative analysis of ovarian lesions.This article reviews the basic principles of spectral CT and its application progress in quantitative parameters in differential diagnosis, efficacy evaluation and metastasis monitoring of ovarian cancer, and looks forward to future research directions, aiming to provide reference for clinical practice and related research.

Neonatal white matter damage (NWMD) is one of the most common forms of injury leading to adverse neurodevelopmental outcomes in neonates, ranging from good prognosis to severe neurological sequelae and even death, which can cause serious economic burdens to families and society. With the rapid development of science and technology, artificial intelligence models based on MRI are expected to become an important tool to assist clinical disease screening, diagnosis and prognosis assessment. However, current research still faces various challenges, mainly focusing on the small amount of data, the lack of standards, cross-center commonality, and ethical privacy. In future breakthroughs, we can focus on the innovation of multimodal fusion technology and the application of self-supervised learning. At the same time, it is also necessary to promote cross-institutional collaboration to build standard datasets, develop ethical data tools, design explainable AI systems, and promote the transformation from algorithm innovation to clinical implementation.