Search Results (4,657)

Background: Myocardial perfusion scintigraphy (MPS) is widely used for noninvasive assessment of coronary artery disease, but publicly available datasets suitable for reproducible deep learning segmentation studies remain limited. This paper proposes CardioProto-SegNet, an image-only 3D anatomy-directed segmentation framework for myocardial region delineation using the public Myocardial Perfusion Scintigraphy Image Database v1.0.0 from PhysioNet, which contains 83 patient studies. Methods: The model is implemented as a 3D U-Net-like residual encoder–decoder network enhanced with squeeze-and-excitation channel recalibration and compact prototype-memory refinement at the bottleneck. Because the public dataset does not provide structured clinical variables, all reported results correspond to image-only myocardium segmentation. Results: Experimental evaluation demonstrated reliable segmentation performance on the available public dataset. CardioProto-SegNet achieved a Dice score of 0.7402 on the holdout test split. In five-fold cross-validation, the model obtained a mean Dice of 0.8239, mean IoU of 0.6870, mean accuracy of 0.9943, mean ROC-AUC of 0.9867, and mean PR-AUC of 0.8561. Since confirmed ischemia or infarction labels were not available, an exploratory image-derived subgroup analysis was additionally performed based on myocardial ROI uptake heterogeneity to examine model behavior in lower- and higher-heterogeneity cases. The ablation study showed that residual connections were important for stable segmentation performance, while the deeper variant achieved the highest tested performance, with a Dice score of 0.8290, IoU of 0.7096, and PR-AUC of 0.8831. Conclusions: Overall, the findings suggest that CardioProto-SegNet provides a reproducible public dataset benchmark for myocardium segmentation in MPS and may serve as a foundation for future downstream quantitative and CAD-oriented analysis when larger datasets with clinical labels become available. Full article

Cardiac computed tomography (CT) has evolved into a pivotal tool in precision cardiology, enabling comprehensive, non-invasive evaluation of coronary anatomy, plaque composition, vascular function, and inflammation. From calcium scoring to advanced physiological imaging, CT now integrates multiple layers of cardiovascular information within a unified diagnostic framework. Coronary artery calcium (CAC) quantification provides a robust, reproducible measure of atherosclerotic burden and refines risk estimation beyond traditional algorithms, particularly in asymptomatic individuals with an intermediate likelihood. Building upon this anatomical foundation, coronary CT angiography (CCTA) extends evaluation to the anatomical and morphological characterization of coronary artery disease (CAD), identifying both obstructive and non-obstructive plaques with high prognostic accuracy. The addition of CT-derived fractional flow reserve (FFR-CT) and stress perfusion CT (CTP) bridges anatomy and physiology, improving identification of flow-limiting stenoses and guiding revascularization decisions while reducing unnecessary invasive procedures. Beyond luminal assessment, CT-derived biomarkers such as the perivascular fat attenuation index (pFAI) have introduced a new dimension of vascular inflammation imaging, revealing residual risk even in patients without significant stenosis and suggesting novel pathways for individualized therapeutic targeting. Driven by advances in artificial intelligence and photon-counting detector technology, cardiac CT is transitioning from a purely diagnostic modality to an integrative platform for cardiovascular phenotyping. Taken as a whole, this integration of structural, functional, and biological data provides a genuinely holistic view of coronary health. In practical terms, it shifts clinical decision-making from population-based risk models toward precision-guided patient-specific strategies. Full article

Rhegmatogenous retinal detachment (RRD) is a vision-threatening retinal disorder in which successful anatomical reattachment does not always result in satisfactory functional visual recovery. Optical coherence tomography (OCT) has become an essential non-invasive imaging modality for evaluating retinal microstructure and identifying biomarkers associated with postoperative visual outcomes. This narrative review summarizes and critically evaluates current evidence regarding the prognostic significance of OCT biomarkers following RRD repair. Available studies indicate that primary outer retinal biomarkers, particularly the integrity of the ellipsoid zone (EZ) and external limiting membrane (ELM), are the most reliable predictors of postoperative best-corrected visual acuity. Preservation of these structures reflects photoreceptor viability and is strongly associated with improved visual recovery. Secondary biomarkers, including hyperreflective foci, intraretinal cystic changes, retinal folds, persistent subretinal fluid, and preretinal hyperreflective dots, provide complementary information regarding inflammatory activity, retinal remodeling, and postoperative complications such as epiretinal membrane formation. Although these biomarkers have considerable clinical value for prognostication, patient counseling, and postoperative monitoring, their interpretation is limited by heterogeneity among studies, differences in imaging protocols, and the lack of standardized assessment criteria. Future large-scale prospective studies are needed to validate OCT biomarkers and establish standardized approaches for their integration into routine clinical practice. Full article

The evaluation of stents in coronary computed tomography angiography (CCTA) is still a major topic in cardiovascular imaging. Using Photon-Counting Detector CT (PCCT) may improve the assessment of coronary stents and make on-site CT-FFR analysis feasible for ruling out in-stent restenosis (ISR). In this study, patients with previous coronary stent implantation who underwent CCTA using PCCT and subsequent invasive catheter angiography (ICA) were included. Stent characteristics such as location and length were reported. CT-FFR measurements were taken 1.8 cm before and after the stent, with a value of ≤0.80 defined as hemodynamically significant under respecting the diagnostic accuracy drop in the gray zone between 0.76 and 0.80. Delta CT-FFR with a cut-off value of ≥0.06, indicating hemodynamic significance, was determined. Any ISR and interventional treatment during the following ICA was recorded. Diagnostic performance metrics, including sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV), were calculated for post-stent CT-FFR and Delta CT-FFR in detecting ISR. Patients were followed up to evaluate the rate of major adverse cardiovascular events (MACE) 6 months after CCTA. A total of 19 patients (5 female, 14 male, median age 69 years) were enrolled in this study. In most cases, coronary stents were located in the proximal LAD with a median stent length of 70.2 mm. Pathological CT-FFR < 0.76 distal to the stent was detected in 6 cases (31.6%), while pathological Delta CT-FFR ≥ 0.06 occurred in 14 cases (73.7%). ICA was performed in three of these patients, with ISR confirmed in two cases. These findings yield sensitivity and NPV of 100% for both post-stent CT-FFR and Delta CT-FFR for excluding ISR with a superior specificity (76.5% vs. 29.4%) and overall diagnostic accuracy (78.9% vs. 36.8%) for post-stent CT-FFR. Two patients reported a myocardial infarction in follow-up; however, neither of them was located in the territory of the stented coronary artery. This study outlines the feasibility of on-site CT-FFR analysis using PCCT in excluding ISR in coronary stents with a high diagnostic accuracy. These findings highlight the need to extend the benefits of CT-FFR analysis for non-invasive assessment of possible ISR regarding personalized risk stratification and therapy planning. Full article

Pig vocalization recognition can support non-invasive monitoring in precision livestock farming, but labelled pig-sound recordings are often limited for specific behaviours or physiological states. Under few-shot conditions, deep models may overfit, whereas traditional acoustic features may not fully describe class-specific time-frequency patterns. This study proposed PSA-AP, a pig-sound adaptation pipeline that uses log-Mel spectrograms and integrates SpecAugment-based domain expansion, ImageNet-pretrained ResNet18 knowledge transfer, and ArcFace-based feature alignment. The method was designed to reduce dependence on limited labelled samples, improve task-adapted representation learning, and enhance inter-class separability in the embedding space. Experiments were conducted on a five-class few-shot pig vocalization classification task, including eat, estrous, farrowing (fap), howl, and oink sounds collected from 10 adult Landrace pigs. Using $K=\{5,10,15,20,25,30\}$ labelled wav files per class and five random seeds, each selected training wav file and each held-out test wav file was converted into one 1.0 s log-Mel spectrogram for model training or evaluation. Final evaluation was based on the last checkpoint of each training run. PSA-AP achieved the best mean Accuracy, Macro-F1, and UAR at every K-shot setting. At $K=30$, PSA-AP reached 90.60% Accuracy, 90.49% Macro-F1, and 90.60% UAR, exceeding Raw by 7.80, 7.82, and 7.80 percentage points, respectively. These results indicate that the proposed integration of domain expansion, knowledge transfer, and feature alignment provides a feasible supervised adaptation strategy for few-shot pig vocalization recognition within the current protocol. Full article

Background/Objectives: Topical imiquimod is an established non-invasive treatment for superficial basal cell carcinoma (sBCC). However, data on treatment-associated changes in tumor microvascularization remain limited. This study investigated vascular changes before and after imiquimod therapy using multimodal non-invasive imaging. Methods: In this single-center, prospective observational study, 31 basal cell carcinomas in 20 patients were examined before and 12–16 weeks after topical imiquimod therapy (5%, five times weekly for six weeks) using dermoscopy, dynamic optical coherence tomography (D-OCT), and line-field confocal optical coherence tomography (LC-OCT). Analyses were performed as paired before-and-after comparisons. While approved for sBCC, a small number of thin nodular and infiltrative BCCs were included exploratorily; subgroup analyses were not powered. Results: Dermoscopy showed a nominally significant shift toward smaller vessel diameter categories after therapy (ATS = 8.183, df = 1, p = 0.004). D-OCT-derived parameters (vessel density, vessel diameter, and depth of the vascular plexus) did not show nominally significant changes. LC-OCT showed nominally lower apparent intratumoral flow scores (ATS = 13.285, df = 1, p < 0.001), reduced occurrence of vessel-wall-associated intraluminal structures showing a rolling-like motion pattern (86.7% before treatment versus 33.3% after treatment; ATS = 13.357; df = 1, p < 0.001), and a reduction in maximum vessel diameter (ATS = 6.110, df = 1, p = 0.013). The primary LC-OCT inferential analyses were performed at the lesion level without adjustment for within-patient clustering and should therefore be interpreted as exploratory. An additional patient-cluster-adjusted paired change-score sensitivity analysis for LC-OCT maximum vessel diameter yielded a directionally consistent estimate (−17.81 µm; 95% CI: −34.40 to −1.23; p = 0.037). The primary exploratory endpoints were LC-OCT–based apparent intratumoral flow and maximum vessel diameter; secondary endpoints included dermoscopic and D-OCT–based vascular parameters. In the exploratory response-stratified analysis, the change in LC-OCT-based maximum vessel diameter did not differ significantly among the assigned response groups (Kruskal–Wallis H = 3.870, df = 2, raw p = 0.144; BH-adjusted p = 0.753). Conclusions: LC-OCT detected several exploratory vascular changes between the pre-treatment examination and follow-up and may provide complementary information for the non-invasive assessment of BCC after imiquimod therapy. Given the exploratory design, limited sample size, and lack of systematic histological confirmation, these findings are hypothesis-generating and require validation in larger prospective studies. Full article

Autism spectrum disorder (ASD) is a heterogeneous neurodevelopmental condition, and objective identification based on neuroimaging remains challenging due to inter-subject variability, multi-site heterogeneity, and the complex topology of brain functional networks. Resting-state functional magnetic resonance imaging (rs-fMRI) provides a non-invasive way to characterize intrinsic brain activity, but existing functional-connectivity-based methods often rely on single-scale static representations and insufficiently capture high-order topology, temporal evolution, and phenotypic heterogeneity. This study aims to develop a mathematical and AI-based brain-network modeling framework for ASD identification from rs-fMRI. The proposed method integrates low-order functional connectivity, high-order functional connectivity, phenotypic information, dynamic graph sequences, Transformer-based temporal attention, and static–dynamic gated fusion. Experiments were conducted on the ABIDE-I dataset, including 1112 subjects from 17 acquisition sites, with 539 ASD subjects and 573 typical controls. The proposed static multi-channel model achieved an accuracy of 75.8%, while the dynamic extension achieved a mean accuracy of 78.5% ± 0.7% and an AUC of 0.84 ± 0.01 over repeated runs. The results suggest that jointly modeling multi-scale static topology and dynamic temporal evolution may improve rs-fMRI-based ASD identification and offer a computationally interpretable framework for AI-assisted neuroimaging analysis. Full article

Background: Neurovascular coupling (NVC) is a fundamental mechanism that dynamically matches retinal blood flow to neuronal metabolic demand. While dynamic vessel analysis (DVA) has been established for assessing NVC through flicker-light stimulation, the potential of optical coherence tomography angiography (OCTA) to detect NVC during physiological stimuli, such as dark-to-light adaptation, remains unexplored. Methods: In this prospective cross-sectional study, OCTA imaging was performed in both eyes of 22 healthy participants under dark-adapted (scotopic) and light-adapted (photopic) conditions. Each condition was measured three times consecutively. Macular and peripapillary vessel density (VD) were quantified. Results: After adjustment for measurement order and scan quality, light adaptation significantly increased peripapillary small VD (Δ = +1.30%, p = 0.046; 95% CI: 0.03–2.56%). Peripapillary all VD demonstrated a similar trend but remained borderline significant (Δ = +1.19%, p = 0.069). In contrast, macular VD showed no significant association with light adaptation (Δ = −0.91%, p = 0.11; 95% CI: −2.02 to 0.21%), but was significantly affected by scan quality (Δ = 1.62%, p < 0.001, 95% CI: 1.23–2.02%). Conclusions: In healthy older adults, OCTA detected an increase in peripapillary VD associated with dark-to-light adaptation, reflecting retinal vascular reactivity consistent with neurovascular coupling. The pronounced influence of scan quality and measurement order underscores their importance as critical confounding factors that must be carefully controlled in functional and longitudinal OCTA studies. Together, these findings highlight OCTA’s promise as a non-invasive tool for assessing retinal neurovascular reactivity, while emphasizing the need for scan quality standardization and order correction to ensure reliable interpretation. Full article

Non-invasive Electrical Impedance Tomography (EIT) and Electrical Impedance Spectroscopy (EIS) are emerging as promising techniques for real-time monitoring and quality assessment in food processing and agri-food applications. This study reviews recent advances in impedance-based sensing for fruit characterization and investigates the experimental implementation of multi-electrode impedance measurements for tomographic imaging. Particular attention is devoted to electrode configurations, electrode polarization effects, and equivalent circuit modeling. Experimental measurements were performed on yellow honeydew melon samples using a four-electrode configuration and a impedance analyzer Keysight E4990 (Keysight Technologies, Santa Rosa, USA) over the frequency range from 20 Hz to 1 MHz. The impedance spectra were validated through Kramers–Kronig consistency tests and interpolated using several fractional-order equivalent circuit models, including single-Cole, double-Cole, and Hayden-based models. The results show that four-electrode measurements are less sensitive to electrode-sample interface artifacts than conventional two-electrode approaches, thereby providing a more reliable estimate of the sample impedance, particularly at low frequencies. Among the tested models, the double-Cole model provided the best interpolation accuracy, while the fractional Hayden models effectively described the temporal evolution of extracellular resistance and membrane-related parameters. Preliminary EIT reconstructions further demonstrate the feasibility of non-destructive tomographic imaging for fruit monitoring. These findings support the potential of EIS and EIT as low-cost, portable, and non-invasive tools for smart food quality assessment and precision agriculture applications. Full article

Background: Transient elastography (TE) is widely used for noninvasive assessment of liver fibrosis. In patients undergoing hemodialysis, however, liver stiffness measurements (LSM) may be affected by rapid intradialytic changes in volume status, venous congestion, and other non-fibrotic determinants. We prospectively evaluated peridialytic variability in liver stiffness and the concordance of serum fibrosis indices with elevated LSM in patients receiving maintenance hemodialysis. Methods: In this prospective paired pilot study, 45 adults on maintenance hemodialysis underwent LSM and controlled attenuation parameter (CAP) assessments immediately before and after a dialysis session; paired data were available for 41 patients. The Fibrosis-4 index (FIB-4) and the aspartate aminotransferase-to-platelet ratio index (APRI) were calculated from routine laboratory values. Paired comparisons, correlation analyses, and receiver operating characteristic curves were used to assess within-patient changes and the ability of serum indices to identify elevated pre-dialysis liver stiffness (LSM ≥ 8 kPa). Because no histologic or imaging reference standard for fibrosis was available, these analyses were interpreted as evidence of concordance with elevated LSM rather than as diagnostic accuracy for liver fibrosis. Results: Median LSM was 7.1 kPa (interquartile range [IQR] 5.2–12.1) pre-dialysis and 7.7 kPa (IQR 5.8–12.2) post-dialysis, with no significant paired change (median ΔLSM −0.2 kPa [IQR −1.1 to 1.2]; p = 0.898). However, the proportion with LSM ≥ 8 kPa increased from 36.6% to 46.3%, with 4 of 41 patients (9.8%) newly exceeding the threshold. CAP values showed no significant paired change (p = 0.511). Intradialytic weight loss was not associated with ΔLSM (rho = −0.13, p = 0.44). FIB-4 and APRI showed poor correlation with LSM and limited concordance with elevated LSM (area under the curve 0.553 and 0.578, respectively, with wide confidence intervals). Conclusions: In this exploratory hemodialysis cohort, cohort-level median LSM did not change significantly after dialysis, but clinically relevant individual-level reclassification occurred in approximately 10% of patients. Measurement timing may alter LSM-based classification, underscoring the need for dialysis-specific validation of LSM thresholds and noninvasive assessment strategies. Full article

Colorectal cancer (CRC) remains the third most common malignancy worldwide and a leading cause of cancer mortality, largely driven by metastatic dissemination. Among metastatic routes, lymphatic spread is crucial to determine the prognosis and establish an adequate therapeutic strategy. Lymph node metastasis (LNM) defines stage III disease in the TNM classification, guiding adjuvant chemotherapy and surgical planning. However, nodal staging based on lymphadenectomy and histopathology is invasive, time-consuming, and may lead to overtreatment. Conventional imaging modalities, including computed tomography, magnetic resonance imaging, and endorectal ultrasound, show limited sensitivity and specificity for small or micro-metastatic nodes. Despite multimodal progress, no non-invasive technique reliably identifies malignant nodes in real time. PET–MRI, contrast-enhanced ultrasound, photoacoustic and fluorescence approaches, ICG mapping, and sentinel node biopsy improve detection but remain limited by specificity, cost, or availability. Extranodal extension (ENE) and tumor deposits (TDs) carry major prognostic value, reflecting aggressive biology and association with distant spread. Meanwhile, phylogenetic studies challenge linear dissemination models, indicating that some metastases arise directly from the primary tumor or TDs rather than LNMs. These data support refinement of staging and surgical strategies according to tumor biology rather than purely anatomical criteria. High-frequency quantitative ultrasound (HF-QUS) enables real-time, operator-independent, three-dimensional nodal assessment with reported sensitivity and specificity exceeding 85%. Combined with artificial intelligence and molecular profiling, it may support biologically informed staging, reduce unnecessary surgery, and foster precision oncology. Lymphatic dissemination in CRC offers a platform to merge tumor biology with technological innovation, where advanced imaging, molecular insight, and artificial intelligence may redefine nodal staging toward precision, non-invasive care. Full article

Miscarriage is a common adverse reproductive outcome, and recurrent pregnancy loss (RPL) remains a major challenge in reproductive medicine. Despite advances in genetics, immunology, endocrinology, and endometrial biology, many RPL cases remain unexplained. Conventional statistical approaches may be limited in capturing high-order nonlinear interactions among clinical, imaging, immunological, and molecular factors associated with pregnancy loss unless these interactions are explicitly modeled. Artificial intelligence (AI), including machine learning (ML) and deep learning (DL), has therefore been investigated as a potential framework for reproductive risk prediction and patient stratification. This narrative review summarizes current evidence on AI-based prediction of miscarriage and RPL, with emphasis on its possible translational relevance to infertility treatment and assisted reproductive technology (ART). Clinical data-driven models have shown potentially useful discriminatory performance, while biomarker-integrated ML approaches suggest that immune-inflammatory signatures may contribute to risk estimation. Imaging-based AI, including radiomics from multimodal ultrasound, may also support noninvasive assessment of endometrial receptivity and inform embryo transfer planning. In parallel, the broader ART literature suggests increasing interest in AI for embryo selection, embryology laboratory workflow, and ovarian stimulation prediction. However, the evidence remains limited by retrospective study designs, small datasets, inconsistent RPL definitions, inadequate external validation, and concerns regarding interpretability, fairness, and regulation. Further progress will require multimodal, explainable, and prospectively validated systems linked to clinically meaningful outcomes. AI may ultimately support more individualized reproductive care, but routine clinical implementation remains premature. Full article

Background: Static (at-rest) and dynamic (expression-linked) wrinkles are complementary hallmarks of facial aging. While static wrinkles are widely studied, the objective quantification of dynamic wrinkles during active facial movement remains a novel and underexplored frontier. Quantifying both phenotypes under real-life product use requires objective, non-invasive endpoints alongside standardized clinical grading. Aim: This study aimed to evaluate the clinical and instrumental efficacy of a multi-active topical cream on static and dynamic wrinkles over 8 weeks of twice-a-day use. Methods: After a 2-week washout, we conducted a monocentric, open-label study on 62 Chinese women (25–55) who used the topical cream twice daily for 8 weeks (per-protocol n = 49; dynamic-wrinkle subset n = 41; dermatologist 0–9 grading at T0/Timm/W4/W8). The instrumental endpoints were PRIMOS-CR wrinkle morphometry (forehead, crow’s feet) and periocular high-frequency ultrasound (UC22). Dynamic wrinkles were assessed via high-speed smile imaging (max P10; mean P1–P10). Statistics comprised Wilcoxon’s tests for dermatologist-graded (ordinal) endpoints and repeated-measures ANOVA with Dunnett’s tests for continuous instrumental endpoints (α = 0.05). Results: Improvements were evident at Timm (periorbital elasticity −17.70%, global-face elasticity −15.23%, firmness −19.47%, smoothness −20.16%, radiance −25.75%; all p < 0.001). By Week 8, dermatologist-graded wrinkles generally decreased: crow’s feet −26.89%, under-eye −33.74%, glabellar −35.30%, forehead −34.69% (all p < 0.001). PRIMOS showed reductions in wrinkle area/length (forehead area −8.69%, length −12.05%; crow’s feet area −8.70%, length −16.03%; all p < 0.001). Ultrasound indicated increased periocular epidermal thickness (+26.57%) and density (+12.69%) (both p = 0.005). Dynamic-wrinkle grades improved during smiling (under-eye: max −12.64%, mean −15.74%; crow’s feet: max −15.97%, mean −16.89%; all p < 0.001), with reductions across P1–P10. Conclusions: In real-life, with twice-daily use, the multi-active cream demonstrated significant within-subject improvements in both static and dynamic (expression-linked) wrinkles, as supported by dermatologist grading, PRIMOS 3D wrinkle morphometry, and periocular high-frequency ultrasound. Full article

Background/Objectives: Autoimmune hepatitis (AIH) is a chronic immune-mediated liver disease that may progress to cirrhosis and liver failure if not diagnosed early. Although liver biopsy remains the reference standard for fibrosis assessment, its invasive nature limits routine use. This study aimed to compare the diagnostic performance of ultrasound-based microvascular imaging (MVI), transient elastography (TE), and serum fibrosis indices (APRI and FIB-4) in patients with biopsy-confirmed AIH. Methods: Fifty-five patients with probable or definite AIH according to IAIHG criteria were included in the study. All patients underwent liver biopsy, and fibrosis stage was assessed using the METAVIR system. TE and MVI examinations were performed, and APRI and FIB-4 scores were calculated. Diagnostic performance was evaluated using AUROC, sensitivity, and specificity. Spearman correlation and logistic regression analyses were additionally performed. Results: The mean age of the patients was 49.2 years, and most patients were women. Cirrhosis was present in 58.2% of the cohort. TE demonstrated high diagnostic accuracy, whereas FIB-4 showed moderate performance and APRI demonstrated limited utility. MVI achieved the highest diagnostic performance, with AUROC values of 0.99 for significant fibrosis and 0.97 for cirrhosis. MVI showed the strongest correlation with histological fibrosis stage (r = 0.916, p < 0.001), followed by TE (r = 0.907, p < 0.001). MVI was strongly associated with histologically confirmed cirrhosis (OR 16.7, 95% CI 2.36–118.2, p = 0.004). Conclusions: MVI demonstrates diagnostic performance comparable to TE and may represent a promising adjunctive non-invasive imaging biomarker for fibrosis assessment in AIH. Larger multicenter studies are required for external validation before routine clinical implementation. Full article

Computed tomography (CT) scan is a widely used, non-invasive, sensor-based imaging technique that provides critical intracranial information for rapid stroke assessment. Accurate segmentation of intracerebral hemorrhage (ICH) in sensor-derived CT images is vital for clinical decision-making. Effective intelligent analysis of CT images is key to achieving reliable computer-aided diagnosis. However, existing deep learning methods struggle with complex ICH lesions characterized by blurred boundaries, irregular shapes, and large-scale variations. To address these challenges, this paper proposes TransAMGNet, a hybrid U-shaped network with Transformer integration for ICH CT image segmentation. The network is built on a residual U-Net backbone and introduces a Transformer encoder to strengthen global context modeling, thereby improving the representation of complex lesion morphology. Specifically, in the encoding stage, we design an Adaptive Dual-branch Channel Attention Module (ADCAM), which jointly models global and local channel information to enhance the model’s sensitivity to important feature responses. In the skip-connection pathway, we introduce a Multi-scale Feature Enhancement Module (MFEM), which preserves high-resolution spatial details while supplementing multi-scale contextual information to improve shallow-deep feature fusion. During decoding, a Gate-enhanced Dynamic Upsampling Module (GDUM) is constructed to improve the recovery of lesion boundaries and fine-grained structures through the synergy of gated recalibration and content-aware upsampling. The proposed method is systematically evaluated through comparative experiments and ablation studies. Experimental results show that TransAMGNet outperforms competing methods across multiple evaluation metrics, achieving Dice, Recall, IoU, Precision, and HD95 values of 90.47 ± 0.58%, 87.83 ± 3.71%, 81.26 ± 0.78%, 91.13 ± 0.95%, and 32.94 ± 1.1, respectively. The ablation studies further verify the effectiveness of each module. These results demonstrate that TransAMGNet can effectively improve segmentation performance for complex ICH lesions. Full article