Yujie Feng

Introduction: Lymphovascular invasion (LVI) is associated with poor outcomes in lung adenocarcinoma, yet accurate preoperative prediction in clinical stage IA disease remains difficult. We aimed to develop and validate a multicenter ensemble machine-learning model that integrates intratumoral heterogeneity (ITH) score and clinicoradiological features for preoperative LVI prediction. Methods: In this retrospective multicenter prediction-model development and external validation study, 1,527 patients with surgically confirmed stage IA lung adenocarcinoma from three tertiary centers were included. Patients from two centers formed the training cohort (n = 1,068), and patients from the third center formed the external validation cohort (n = 459). The intratumoral heterogeneity (ITH) score was quantified from preoperative computed tomography (CT) by integrating global and local pixel-distribution patterns. Ten base machine-learning algorithms and 31 stacking configurations were evaluated. Model performance was assessed using area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, specificity, positive predictive value, negative predictive value, and F1 score. SHapley Additive exPlanations (SHAP) were used for model interpretation and feature selection. Results: The XGBoost plus random forest stacking model ranked first overall according to the comprehensive score and achieved the best integrated diagnostic performance. In comparative analyses, the stacking model showed a more balanced diagnostic profile than the radiomics-only, ITH-only, and clinicoradiological-only models, with the highest overall mean performance, although the radiomics-only model showed a slightly higher AUC and sensitivity. SHAP analysis identified the ITH score as the dominant contributor to model output. Conclusion: A stacking model integrating ITH score and clinicoradiological features enables reliable preoperative prediction of LVI in stage IA lung adenocarcinoma and may support individualized risk stratification and surgical planning.

Compressed sensing (CS) reconstruction is a powerful technique for image compression and reconstruction. Most existing deep learning and deep unfolding reconstruction methods: 1) do not sufficiently analyze the proximal subproblem and naturally derive network architectures from mathematical theory; 2) lack the ability to correct errors and recover details via multidomain, multiscale, and multiresolution analysis; and 3) depend only on the previous stage and overlooking valuable information from earlier stages. In this study, we propose a Nesterov-optimized memory and geometric distillation learning (Nest-MGDL) framework, which effectively integrates interpretability and adaptability to achieve robust and generalizable deep learning performance. By reformulating the proximal subproblem as the decomposition and distillation of memory and geometric priors across diverse domains and integrating a local multilevel feature aggregation (LMFA) strategy, Nest-MGDL enables theoretically grounded network design and accurate multiscale correction of errors. Inspired by the third-Nesterov (N-III) acceleration scheme with fast convergence, we incorporate gradient information from measurement errors across multiple historical stages to achieve more consistent and accurate reconstruction. It provides new insights into leveraging historical measurement errors and designing explainable network architectures. All parameters in our Nest-MGDL are learnable end-to-end with adaptive initialization, enhancing model flexibility and ensuring stable convergence. Extensive experiments show that Nest-MGDL outperforms other state-of-the-art methods.

Objectives: The preoperative differentiation of adenocarcinomas in situ, minimally invasive adenocarcinoma, and invasive adenocarcinoma using computed tomography (CT) is crucial for guiding clinical management decisions. However, accurately classifying ground-glass nodules poses a significant challenge. Incorporating quantitative intratumoral heterogeneity scores may improve the accuracy of this ternary classification. Materials and methods: In this multicenter retrospective study, we developed ternary classification models by leveraging insights from both base and stacking ensemble machine learning models, incorporating intratumoral heterogeneity scores along with clinical-radiological features to distinguish adenocarcinomas in situ, minimally invasive adenocarcinoma, and invasive adenocarcinoma. The machine learning models were trained, and final model selection depended on maximizing the macro-average area under the curve (macro-AUC) in both the internal and external validation sets. Results: Data from 802 patients from three centers were divided into a training set (n = 477) and an internal test set (n = 205), in a 7:3 ratio, with an additional external validation set comprising 120 patients. The stacking classifier exhibited superior performance relative to the other models, achieving macro-AUC values of 0.7850 and 0.7717 for the internal and external validation sets, respectively. Moreover, an interpretability analysis utilizing the Shapley Additive Explanation identified four key features of this ternary classification: intratumoral heterogeneity score, nodule size, nodule type, and age. Conclusion: The stacking classifier, recognized as the optimal algorithm for integrating the intratumoral heterogeneity score and clinical-radiological features, effectively served as a ternary classification model for assessing the invasiveness of lung adenocarcinoma in chest CT images. Critical relevance statement: Our stacking classifier integrated intratumoral heterogeneity scores and clinical-radiological features to improve the ternary classification of lung adenocarcinoma invasiveness (adenocarcinomas in situ/minimally invasive adenocarcinoma/invasive adenocarcinoma), aiding in precise diagnosis and clinical decision-making for ground-glass nodules. Key Points: The intratumoral heterogeneity score effectively assessed the invasiveness of lung adenocarcinoma. The stacking classifier outperformed other methods for this ternary classification task. Intratumoral heterogeneity score, nodule size, nodule type, and age predict invasiveness. Graphical

Recently, deep learning methods have gained remarkable achievements in the field of image restoration for remote sensing (RS). However, most existing RS image restoration methods focus mainly on conventional first-order degradation models, which may not effectively capture the imaging mechanisms of remote sensing images. Furthermore, many RS image restoration approaches that use deep learning are often criticized for their lacks of architecture transparency and model interpretability. To address these problems, we propose a novel progressive restoration network for high-order degradation imaging (HDI-PRNet), to progressively restore different image degradation. HDI-PRNet is developed based on the theoretical framework of degradation imaging, also Markov properties of the high-order degradation process and Maximum a posteriori (MAP) estimation, offering the benefit of mathematical interpretability within the unfolding network. The framework is composed of three main components: a module for image denoising that relies on proximal mapping prior learning, a module for image deblurring that integrates Neumann series expansion with dual-domain degradation learning, and a module for super-resolution. Extensive experiments demonstrate that our method achieves superior performance on both synthetic and real remote sensing images.

The accurate prediction of soil organic matter (SOM) content is essential for promoting sustainable soil management and addressing global climate change. Due to multiple factors such as topography and climate, especially in mountainous areas, SOM spatial prediction faces significant challenges. The main novelty of this study lies in proposing a geographic positional encoding mechanism that embeds geographic location information into the feature representation of a Transformer model. The encoder structure is further modified to enhance spatial awareness, resulting in the development of the Geo-Positional Transformer (GPTransformer). Furthermore, this model is integrated with a 1D-CNN to form a dual-branch neural network called the Geo-Positional Transformer-CNN (GPTransCNN). This study collected 1490 topsoil samples (0–20 cm) from cultivated land in Longshan County to develop a predictive model for mapping the spatial distribution of SOM across the entire cultivated area. Different models were comprehensively evaluated through ten-fold cross-validation, ablation experiments, and uncertainty analysis. The results show that GPTransCNN has the best performance, with an R² improvement of approximately 43% over the Transformer, 19% over the GPTransformer, and 15% over the 1D-CNN. This study demonstrates that by incorporating geographic positional information, GPTransCNN effectively combines the global modeling capabilities of the GPTransformer with the local feature extraction strengths of the 1D-CNN, which can improve the accuracy of SOM mapping in mountainous areas. This approach provides data support for sustainable soil management and decision-making in response to global climate change.

Remote sensing images are essential for many applications of the earth’s sciences, but their quality can usually be degraded due to limitations in sensor technology and complex imaging environments. To address this, various remote sensing image deblurring methods have been developed to restore sharp and high-quality images from degraded observational data. However, most traditional model-based deblurring methods usually require predefined hand-crafted prior assumptions, which are difficult to handle in complex applications. On the other hand, deep learning-based deblurring methods are often considered as black boxes, lacking transparency and interpretability. In this work, we propose a new blind deblurring learning framework that utilizes alternating iterations of shrinkage thresholds. This framework involves updating blurring kernels and images, with a theoretical foundation in network design. Additionally, we propose a learnable blur kernel proximal mapping module to improve the accuracy of the blur kernel reconstruction. Furthermore, we propose a deep proximal mapping module in the image domain, which combines a generalized shrinkage threshold with a multi-scale prior feature extraction block. This module also incorporates an attention mechanism to learn adaptively the importance of prior information, improving the flexibility and robustness of prior terms, and avoiding limitations similar to hand-crafted image prior terms. Consequently, we design a novel multi-scale generalized shrinkage threshold network (MGSTNet) that focuses specifically on learning deep geometric prior features to enhance image restoration. Experimental results on real and synthetic remote sensing image datasets demonstrate the superiority of our MGSTNet framework compared to existing deblurring methods.

Proximal gradient-based optimization is one of the most common strategies to solve inverse problem of images, and it is easy to implement. However, these techniques often generate heavy artifacts in image reconstruction. One of the most popular refinement methods is to fine-tune the regularization parameter to alleviate such artifacts, but it may not always be sufficient or applicable due to increased computational costs. In this work, we propose a deep geometric incremental learning framework based on the second Nesterov proximal gradient optimization. The proposed end-to-end network not only has the powerful learning ability for high-/low-frequency image features, but also can theoretically guarantee that geometric texture details will be reconstructed from preliminary linear reconstruction. Furthermore, it can avoid the risk of intermediate reconstruction results falling outside the geometric decomposition domains and achieve fast convergence. Our reconstruction framework is decomposed into four modules including general linear reconstruction, cascade geometric incremental restoration, Nesterov acceleration, and post-processing. In the image restoration step, a cascade geometric incremental learning module is designed to compensate for missing texture information from different geometric spectral decomposition domains. Inspired by the overlap-tile strategy, we also develop a post-processing module to remove the block effect in patch-wise-based natural image reconstruction. All parameters in the proposed model are learnable, an adaptive initialization technique of physical parameters is also employed to make model flexibility and ensure converging smoothly. We compare the reconstruction performance of the proposed method with existing state-of-the-art methods to demonstrate its superiority. Our source codes are available at https://github.com/fanxiaohong/Nest-DGIL.

Proximal gradient-based optimization is one of the most common strategies to solve inverse problem of images, and it is easy to implement. However, these techniques often generate heavy artifacts in image reconstruction. One of the most popular refinement methods is to fine-tune the regularization parameter to alleviate such artifacts, but it may not always be sufficient or applicable due to increased computational costs. In this work, we propose a deep geometric incremental learning framework based on the second Nesterov proximal gradient optimization. The proposed end-to-end network not only has the powerful learning ability for high-/low-frequency image features, but also can theoretically guarantee that geometric texture details will be reconstructed from preliminary linear reconstruction. Furthermore, it can avoid the risk of intermediate reconstruction results falling outside the geometric decomposition domains and achieve fast convergence. Our reconstruction framework is decomposed into four modules including general linear reconstruction, cascade geometric incremental restoration, Nesterov acceleration, and post-processing. In the image restoration step, a cascade geometric incremental learning module is designed to compensate for missing texture information from different geometric spectral decomposition domains. Inspired by the overlap-tile strategy, we also develop a post-processing module to remove the block effect in patch-wise-based natural image reconstruction. All parameters in the proposed model are learnable, an adaptive initialization technique of physical parameters is also employed to make model flexibility and ensure converging smoothly. We compare the reconstruction performance of the proposed method with existing state-of-the-art methods to demonstrate its superiority. Our source codes are available at https://github.com/fanxiaohong/Nest-DGIL .

Introduction: Ferroptosis-related gene (FRG) signature is important for assessing novel therapeutic approaches and prognosis in glioma. We trained a deep learning network for determining FRG signatures using multiparametric magnetic resonance imaging (MRI).Methods: FRGs of patients with glioma were acquired from public databases. FRG-related risk score stratifying prognosis was developed from The Cancer Genome Atlas (TCGA) and validated using the Chinese Glioma Genome Atlas. Multiparametric MRI-derived glioma images and the corresponding genomic information were obtained for 122 cases from TCGA and The Cancer Imaging Archive. The deep learning network was trained using 3D-Resnet, and threefold cross-validation was performed to evaluate the predictive performance.Results: The FRG-related risk score was associated with poor clinicopathological features and had a high predictive value for glioma prognosis. Based on the FRG-related risk score, patients with glioma were successfully classified into two subgroups (28 and 94 in the high- and low-risk groups, respectively). The deep learning networks TC (enhancing tumor and non-enhancing portion of the tumor core) mask achieved an average cross-validation accuracy of 0.842 and an average AUC of 0.781, while the deep learning networks WT (whole tumor and peritumoral edema) mask achieved an average cross-validation accuracy of 0.825 and an average AUC of 0.781.Discussion: Our findings indicate that FRG signature is a prognostic indicator of glioma. In addition, we developed a deep learning network that has high classification accuracy in automatically determining FRG signatures, which may be an important step toward the clinical translation of novel therapeutic approaches and prognosis of glioma.