BMC Medical Imaging最新文献

{"title":"Open-radiomics: a collection of standardized datasets and a technical protocol for reproducible radiomics machine learning pipelines.","authors":"Khashayar Namdar, Matthias W Wagner, Birgit B Ertl-Wagner, Farzad Khalvati","doi":"10.1186/s12880-025-01855-2","DOIUrl":"10.1186/s12880-025-01855-2","url":null,"abstract":"<p><strong>Background: </strong>As an important branch of machine learning pipelines in medical imaging, radiomics faces two major challenges namely reproducibility and accessibility. In this work, we introduce open-radiomics, a set of radiomics datasets along with a comprehensive radiomics pipeline based on our proposed technical protocol to investigate the effects of radiomics feature extraction on the reproducibility of the results.</p><p><strong>Methods: </strong>We curated large-scale radiomics datasets based on three open-source datasets; BraTS 2020 for high-grade glioma (HGG) versus low-grade glioma (LGG) classification and survival analysis, BraTS 2023 for O6-methylguanine-DNA methyltransferase (MGMT) classification, and non-small cell lung cancer (NSCLC) survival analysis from the Cancer Imaging Archive (TCIA). We used the BraTS 2020 open-source Magnetic Resonance Imaging (MRI) dataset to demonstrate how our proposed technical protocol could be utilized in radiomics-based studies. The cohort includes 369 adult patients with brain tumors (76 LGG, and 293 HGG). Using PyRadiomics library for LGG vs. HGG classification, we created 288 radiomics datasets; the combinations of 4 MRI sequences, 3 binWidths, 6 image normalization methods, and 4 tumor subregions. We used Random Forest classifiers, and for each radiomics dataset, we repeated the training-validation-test (60%/20%/20%) experiment with different data splits and model random states 100 times (28,800 test results) and calculated the Area Under the Receiver Operating Characteristic Curve (AUROC).</p><p><strong>Results: </strong>Unlike binWidth and image normalization, the tumor subregion and imaging sequence significantly affected performance of the models. T1 contrast-enhanced sequence and the union of Necrotic and the non-enhancing tumor core subregions resulted in the highest AUROCs (average test AUROC 0.951, 95% confidence interval of (0.949, 0.952)). Although several settings and data splits (28 out of 28800) yielded test AUROC of 1, they were irreproducible.</p><p><strong>Conclusions: </strong>Our experiments demonstrate the sources of variability in radiomics pipelines (e.g., tumor subregion) can have a significant impact on the results, which may lead to superficial perfect performances that are irreproducible.</p><p><strong>Clinical trial number: </strong>Not applicable.</p>","PeriodicalId":9020,"journal":{"name":"BMC Medical Imaging","volume":"25 1","pages":"312"},"PeriodicalIF":3.2,"publicationDate":"2025-08-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12323200/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144783441","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Photon counting detector CT-derived virtual non-contrast images of the liver: comparison of conventional and liver-specific algorithms across arterial and portal venous phase scans.","authors":"Anna-Katharina Gerstner, Franka Risch, Luca Canalini, Gerlig Widmann, Elke R Gizewski, Stefanie Bette, Simon Hellbrueck, Thomas Kroencke, Josua A Decker","doi":"10.1186/s12880-025-01849-0","DOIUrl":"10.1186/s12880-025-01849-0","url":null,"abstract":"","PeriodicalId":9020,"journal":{"name":"BMC Medical Imaging","volume":"25 1","pages":"311"},"PeriodicalIF":3.2,"publicationDate":"2025-08-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12323134/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144783442","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Combined nomogram for differentiating adrenal pheochromocytoma from large-diameter lipid-poor adenoma using multiphase CT radiomics and clinico-radiological features.","authors":"Zujuan Shan, Xinzhang Zhang, Yiwen Zhang, Shuailong Wang, Junfeng Wang, Xin Shi, Lin Li, Zhenhui Li, Liuyang Yang, Hao Liu, Wenliang Li, Junfeng Yang, Liansheng Yang","doi":"10.1186/s12880-025-01835-6","DOIUrl":"10.1186/s12880-025-01835-6","url":null,"abstract":"<p><strong>Background and objective: </strong>Adrenal incidentalomas (AIs) are predominantly adrenal adenomas (80%), with a smaller proportion (7%) being pheochromocytomas(PHEO). Adenomas are typically non-functional tumors managed through observation or medication, with some cases requiring surgical removal, which is generally safe. In contrast, PHEO secrete catecholamines, causing severe blood pressure fluctuations, making surgical resection the only treatment option. Without adequate preoperative preparation, perioperative mortality risk is significantly high.A specialized adrenal CT scanning protocol is recommended to differentiate between these tumor types. However, distinguishing patients with similar washout characteristics remains challenging, and concerns about efficiency, cost, and risk limit its feasibility. Recently, radiomics has demonstrated efficacy in identifying molecular-level differences in tumor cells, including adrenal tumors. This study develops a combined nomogram model, integrating key clinical-radiological and radiomic features from multiphase CT, to enhance accuracy in distinguishing pheochromocytoma from large-diameter lipid-poor adrenal adenoma (LP-AA).</p><p><strong>Methods: </strong>A retrospective analysis was conducted on 202 patients with pathologically confirmed adrenal PHEO and large-diameter LP-AA from three tertiary care centers. Key clinico-radiological and radiomics features were selected to construct models: a clinico-radiological model, a radiomics model, and a combined nomogram model for predicting these two tumor types. Model performance and robustness were evaluated using external validation, calibration curve analysis, machine learning techniques, and Delong's test. Additionally, the Hosmer-Lemeshow test, decision curve analysis, and five-fold cross-validation were employed to assess the clinical translational potential of the combined nomogram model.</p><p><strong>Results: </strong>All models demonstrated high diagnostic performance, with AUC values exceeding 0.8 across all cohorts, confirming their reliability. The combined nomogram model exhibited the highest diagnostic accuracy, with AUC values of 0.994, 0.979, and 0.945 for the training, validation, and external test cohorts, respectively. Notably, the unenhanced combined nomogram model was not significantly inferior to the three-phase combined nomogram model (p > 0.05 in the validation and test cohorts; p = 0.049 in the training cohort).</p><p><strong>Conclusions: </strong>The combined nomogram model reliably distinguishes between PHEO and LP-AA, shows strong clinical translational potential, and may reduce the need for contrast-enhanced CT scans.</p><p><strong>Clinical trial number: </strong>Not applicable.</p>","PeriodicalId":9020,"journal":{"name":"BMC Medical Imaging","volume":"25 1","pages":"313"},"PeriodicalIF":3.2,"publicationDate":"2025-08-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12323233/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144783439","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A dual self-attentive transformer U-Net model for precise pancreatic segmentation and fat fraction estimation.","authors":"Ashok Shanmugam, Prianka Ramachandran Radhabai, Kavitha Kvn, Agbotiname Lucky Imoize","doi":"10.1186/s12880-025-01852-5","DOIUrl":"10.1186/s12880-025-01852-5","url":null,"abstract":"<p><p>Accurately segmenting the pancreas from abdominal computed tomography (CT) images is crucial for detecting and managing pancreatic diseases, such as diabetes and tumors. Type 2 diabetes and metabolic syndrome are associated with pancreatic fat accumulation. Calculating the fat fraction aids in the investigation of β-cell malfunction and insulin resistance. The most widely used pancreas segmentation technique is a U-shaped network based on deep convolutional neural networks (DCNNs). They struggle to capture long-range biases in an image because they rely on local receptive fields. This research proposes a novel dual Self-attentive Transformer Unet (DSTUnet) model for accurate pancreatic segmentation, addressing this problem. This model incorporates dual self-attention Swin transformers on both the encoder and decoder sides to facilitate global context extraction and refine candidate regions. After segmenting the pancreas using a DSTUnet, a histogram analysis is used to estimate the fat fraction. The suggested method demonstrated excellent performance on the standard dataset, achieving a DSC of 93.7% and an HD of 2.7 mm. The average volume of the pancreas was 92.42, and its fat volume fraction (FVF) was 13.37%.</p>","PeriodicalId":9020,"journal":{"name":"BMC Medical Imaging","volume":"25 1","pages":"315"},"PeriodicalIF":3.2,"publicationDate":"2025-08-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12323108/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144783438","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Exploring the relationships between CT and pathological characteristics and gene mutations in neoplastic ground glass nodules.","authors":"Zi-Ya Zhao, Si-Zhu Liu, Xue-Ping Chen, Yang-Li Zhang, Bin-Jie Fu, Wang-Jia Li, Fa-Jin Lv, Zhi-Gang Chu","doi":"10.1186/s12880-025-01851-6","DOIUrl":"10.1186/s12880-025-01851-6","url":null,"abstract":"<p><strong>Background: </strong>Neoplastic ground glass nodules (GGNs) are relatively indolent tumors, with slow progression in invasiveness and computed tomography (CT) features. This study aimed to explore the correlation between pathological and CT characteristics and gene mutations in neoplastic GGNs.</p><p><strong>Methods: </strong>We retrospectively analyzed 1,348 neoplastic GGNs from January 2019 to November 2022, including 290 adenocarcinomas in situ (AIS), 448 microinvasive adenocarcinomas (MIA), and 610 invasive adenocarcinomas (IAC). The correlations between patients' characteristics, pathological subtypes and grades, CT features, changes in follow-up, and gene mutations were analyzed.</p><p><strong>Results: </strong>Solid component (odds ratio [OR] = 1.493; P = 0.014), larger size (OR = 1.049; P = 0.006), ill-defined boundary (OR = 1.368; P = 0.027), and lobulation sign (OR = 1.824; P = 0.001) were revealed as independent CT predictors of gene mutation. From AIS to IAC, the epidermal growth factor receptor (EGFR) mutation rate significantly increased (P < 0.01), while the kirsten rat sarcoma viral oncogene (KRAS) and the anaplastic lymphoma kinase (ALK) mutation rates significantly decreased (P < 0.05). Among IACs, mutation rate was the highest in the intermediate-grade ones (P < 0.05). Gene mutations were more frequently detected in nodules showing changes during follow-up (76.3% vs. 61.1%, P = 0.02), especially in those with more than 2-year follow-up (77.1% vs. 43.7%, P = 0.023). However, the specific changes were not associated with gene mutations (P = 0.273).</p><p><strong>Conclusions: </strong>Gene mutations in neoplastic GGNs were associated with CT features, pathological subtypes and grades, and changes observed during long-term follow-up.</p>","PeriodicalId":9020,"journal":{"name":"BMC Medical Imaging","volume":"25 1","pages":"314"},"PeriodicalIF":3.2,"publicationDate":"2025-08-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12323122/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144783440","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Differentiating between adrenocortical carcinoma and pheochromocytoma by a CT-based radiomics model: a multicenter retrospective study.","authors":"Yinyao Chao, Hongzhang Zhu, Wenyi Yang, Haohua Yao, Nan Ma, Xianda Chen, Jing Zhao, Huali Ma, Zhenhua Liu, Hui Han, Zhuowei Liu, Kai Yao, Yiyao Li, Peng Wu, Jingtong Zhang, Bin Li, Shengjie Guo","doi":"10.1186/s12880-025-01842-7","DOIUrl":"10.1186/s12880-025-01842-7","url":null,"abstract":"<p><strong>Background: </strong>Adrenocortical carcinoma (ACC), a rare and highly malignant adrenal gland tumor, exhibits computed tomography (CT) characteristics that resemble those of the less malignant pheochromocytoma (PHEO). While biochemical evaluation is widely accepted for differentiating between ACC and PHEO, non-functioning tumors remain a diagnostic challenge. The similarity in CT imaging and atypical hormone levels can lead to suboptimal accuracy in diagnosis, leading to inappropriate clinical interventions. This study aims to differentiate between large (≥ 4 cm) ACC and PHEO with radiomics features based on contrast-enhanced CT.</p><p><strong>Methods: </strong>In this retrospective study, 158 patients who received pathological diagnoses of ACC or PHEO between January 2011 and September 2023 were enrolled from three institutions. Radiomics features were extracted from different phases of contrast-enhanced CT and then selected by a two-step procedure. The radiomics model was developed in a cohort of 109 patients from Institution 1, then the model performance was evaluated in the external test cohort of 49 patients from Institutions 2 and 3. The area under the receiver operating characteristic curve (AUC) of the radiomics model was compared with two radiologists using the DeLong test. Hormone testing results were collected to determine the presence of excess cortisol or catecholamines. SHapley Additive exPlanations (SHAP) was used to improve the interpretability of the radiomics model.</p><p><strong>Results: </strong>We developed and evaluated a radiomics model consisting of ten selected CT-based radiomics features. In the external test cohort, the proposed radiomics model achieved high accuracy (86%), specificity (88%), and sensitivity (81%) in differentiating between ACC and PHEO and outperformed 2 radiologists (AUC 0.920 vs. 0.786, 0.629). This radiomics model showed strong capabilities in differentiating biochemically negative ACC and PHEO (with an accuracy of 80%). Moreover, its performance remained consistent even when cortisol and catecholamine levels were simultaneously elevated. Furthermore, SHAP provided quantitative explanations for the radiomics model and visualized the diagnostic process.</p><p><strong>Conclusions: </strong>The interpretable CT-based radiomics model outperforms radiologists in differentiating between ACC and PHEO, especially when hormone testing results are atypical.</p>","PeriodicalId":9020,"journal":{"name":"BMC Medical Imaging","volume":"25 1","pages":"310"},"PeriodicalIF":3.2,"publicationDate":"2025-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12317479/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144764436","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A brain tumor segmentation enhancement in MRI images using U-Net and transfer learning.","authors":"Amin Pourmahboubi, Nazanin Arsalani Saeed, Hamed Tabrizchi","doi":"10.1186/s12880-025-01837-4","DOIUrl":"10.1186/s12880-025-01837-4","url":null,"abstract":"<p><p>This paper presents a novel transfer learning approach for segmenting brain tumors in Magnetic Resonance Imaging (MRI) images. Using Fluid-Attenuated Inversion Recovery (FLAIR) abnormality segmentation masks and MRI scans from The Cancer Genome Atlas's (TCGA's) lower-grade glioma collection, our proposed approach uses a VGG19-based U-Net architecture with fixed pretrained weights. The experimental findings, which show an Area Under the Curve (AUC) of 0.9957, F1-Score of 0.9679, Dice Coefficient of 0.9679, Precision of 0.9541, Recall of 0.9821, and Intersection-over-Union (IoU) of 0.9378, show how effective the proposed framework is. According to these metrics, the VGG19-powered U-Net outperforms not only the conventional U-Net model but also other variants that were compared and used different pre-trained backbones in the U-Net encoder.Clinical trial registrationNot applicable as this study utilized existing publicly available dataset and did not involve a clinical trial.</p>","PeriodicalId":9020,"journal":{"name":"BMC Medical Imaging","volume":"25 1","pages":"307"},"PeriodicalIF":3.2,"publicationDate":"2025-07-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12312496/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144759091","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"The use of diffusion-weighted magnetic resonance imaging and parametric response mapping for disease outcome prediction in nasopharyngeal carcinoma.","authors":"Akarapong Teeraakaravipas, Napat Ritlumlert, Yothin Rakvongthai, Tunchanok Paprad, Chawalit Lertbutsayanukul, Nutchawan Jittapiromsak","doi":"10.1186/s12880-025-01847-2","DOIUrl":"10.1186/s12880-025-01847-2","url":null,"abstract":"<p><strong>Background: </strong>Nasopharyngeal carcinoma (NPC) shows variable treatment responses due to tumor heterogeneity and individual radiosensitivity, complicating the early identification of patients at risk for recurrence. Developing reliable imaging biomarkers could help predict treatment outcomes, enabling timely treatment adjustments and improved prognosis. Therefore, we aimed to evaluate the use of the apparent diffusion coefficient (ADC), based on diffusion-weighted imaging, and parametric response mapping (PRM), a voxel-wise imaging analysis method, in predicting treatment outcomes of patients with NPC.</p><p><strong>Methods: </strong>This retrospective and prospective cohort study included 70 patients with NPC, treated with radiotherapy or concurrent chemoradiation therapy with or without induction chemotherapy. Imaging examinations were performed before (pre-treatment) and 5 weeks after initiating treatment (intra-treatment). Tumor volume at pre- and intra-treatment, percentage change in tumor volume (%∆Vol), pre- and intra-treatment ADC, percentage change in ADC (%∆ADC), and voxels with increased ADC values within the tumor (PRM+) were used to predict correlation with treatment outcomes. Poor outcomes were defined as developing locoregional recurrence, distant metastases, or death. The primary endpoint was progression-free survival, defined as the time to these events. Kaplan-Meier survival analysis, Cox regression, and multivariate models were used to determine predictive factors.</p><p><strong>Results: </strong>Overall, 17 and 53 patients had poor and good outcomes, respectively. The PRM+ was lower in patients with poor outcomes than in those with good outcomes (22.4% vs. 64.1%; p < 0.001). In the multivariate analyses, cut-off values of PRM+ < 35% and initial T-stage 3-4 were identified as two risk factors associated with poor outcomes, with adjusted hazard ratios (95% confidence intervals) of 22.53 (5.09-99.8; p < 0.001), and 3.45 (1.10-10.77; p = 0.033), respectively.</p><p><strong>Conclusions: </strong>Low PRM+ and high initial T-stage were associated with poor treatment outcomes. Therefore, PRM+ can be a predictive tool for NPC treatment outcomes. Integrating PRM into clinical practice could enhance individualized treatment planning, leading to better patient outcomes and reduced treatment-related side effects.</p>","PeriodicalId":9020,"journal":{"name":"BMC Medical Imaging","volume":"25 1","pages":"308"},"PeriodicalIF":3.2,"publicationDate":"2025-07-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12315438/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144759094","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Impact of large language models and vision deep learning models in predicting neoadjuvant rectal score for rectal cancer treated with neoadjuvant chemoradiation.","authors":"Hyun Bin Kim, Hong Qi Tan, Wen Long Nei, Ying Cong Ryan Shea Tan, Yiyu Cai, Fuqiang Wang","doi":"10.1186/s12880-025-01844-5","DOIUrl":"10.1186/s12880-025-01844-5","url":null,"abstract":"<p><p>This study aims to explore Deep Learning methods, namely Large Language Models (LLMs) and Computer Vision models to accurately predict neoadjuvant rectal (NAR) score for locally advanced rectal cancer (LARC) treated with neoadjuvant chemoradiation (NACRT). The NAR score is a validated surrogate endpoint for LARC. 160 CT scans of patients were used in this study, along with 4 different types of radiology reports, 2 generated from CT scans and other 2 from MRI scans, both before and after NACRT. For CT scans, two different approaches with convolutional neural network were utilized to tackle the 3D scan entirely or tackle it slice by slice. For radiology reports, an encoder architecture LLM was used. The performance of the approaches was quantified by the Area under the Receiver Operating Characteristic curve (AUC). The two different approaches for CT scans yielded [Formula: see text] and [Formula: see text] while the LLM trained on post NACRT MRI reports showed the most predictive potential at [Formula: see text] and a statistical improvement, p = 0.03, over the baseline clinical approach (from [Formula: see text] to [Formula: see text])). This study showcases the potential of Large Language Models and the inadequacies of CT scans in predicting NAR values. Clinical trial number Not applicable.</p>","PeriodicalId":9020,"journal":{"name":"BMC Medical Imaging","volume":"25 1","pages":"306"},"PeriodicalIF":3.2,"publicationDate":"2025-07-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12312340/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144759092","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"TA-SSM net: tri-directional attention and structured state-space model for enhanced MRI-Based diagnosis of Alzheimer's disease and mild cognitive impairment.","authors":"Sichen Bao, Fengbo Zheng, Lifen Jiang, Qiuyuan Wang, Yong Lyu","doi":"10.1186/s12880-025-01836-5","DOIUrl":"10.1186/s12880-025-01836-5","url":null,"abstract":"<p><p>Early diagnosis of Alzheimer's disease (AD) and its precursor, mild cognitive impairment (MCI), is critical for effective prevention and treatment. Computer-aided diagnosis using magnetic resonance imaging (MRI) provides a cost-effective and objective approach. However, existing methods often segment 3D MRI images into 2D slices, leading to spatial information loss and reduced diagnostic accuracy. To overcome this limitation, we propose TA-SSM Net, a deep learning model that leverages tri-directional attention and structured state-space model (SSM) for improved MRI-based diagnosis of AD and MCI. The tri-directional attention mechanism captures spatial and contextual information from forward, backward, and vertical directions in 3D MRI images, enabling effective feature fusion. Additionally, gradient checkpointing is applied within the SSM to enhance processing efficiency, allowing the model to handle whole-brain scans while preserving spatial correlations. To evaluate our method, we construct a dataset from the Alzheimer's Disease Neuroimaging Initiative (ADNI), consisting of 300 AD patients, 400 MCI patients, and 400 normal controls. TA-SSM Net achieved an accuracy of 90.24% for MCI detection and 95.83% for AD detection. The results demonstrate that our approach not only improves classification accuracy but also enhances processing efficiency and maintains spatial correlations, offering a promising solution for the diagnosis of Alzheimer's disease.</p>","PeriodicalId":9020,"journal":{"name":"BMC Medical Imaging","volume":"25 1","pages":"309"},"PeriodicalIF":3.2,"publicationDate":"2025-07-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12315388/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144759093","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}