Automatic Tumor Segmentation Research Articles

Development and Validation of a Deep Learning System to Differentiate HER2-Zero, HER2-Low, and HER2-Positive Breast Cancer Based on Dynamic Contrast-Enhanced MRI.

Previous studies explored MRI-based radiomic features for differentiating between human epidermal growth factor receptor 2 (HER2)-zero, HER2-low, and HER2-positive breast cancer, but deep learning's effectiveness is uncertain. This study aims to develop and validate a deep learning system using dynamic contrast-enhanced MRI (DCE-MRI) for automated tumor segmentation and classification of HER2-zero, HER2-low, and HER2-positive statuses. Retrospective. One thousand two hundred ninety-four breast cancer patients from three centers who underwent DCE-MRI before surgery were included in the study (52 ± 11 years, 811/204/279 for training/internal testing/external testing). 3 T scanners, using T1-weighted 3D fast spoiled gradient-echo sequence, T1-weighted 3D enhanced fast gradient-echo sequence and T1-weighted turbo field echo sequence. An automated model segmented tumors utilizing DCE-MRI data, followed by a deep learning models (ResNetGN) trained to classify HER2 statuses. Three models were developed to distinguish HER2-zero, HER2-low, and HER2-positive from their respective non-HER2 categories. Dice similarity coefficient (DSC) was used to evaluate the segmentation performance of the model. Evaluation of the model performances for HER2 statuses involved receiver operating characteristic (ROC) curve analysis and the area under the curve (AUC), accuracy, sensitivity, and specificity. The P-values <0.05 were considered statistically significant. The automatic segmentation network achieved DSC values of 0.85 to 0.90 compared to the manual segmentation across different sets. The deep learning models using ResNetGN achieved AUCs of 0.782, 0.776, and 0.768 in differentiating HER2-zero from others in the training, internal test, and external test sets, respectively. Similarly, AUCs of 0.820, 0.813, and 0.787 were achieved for HER2-low vs. others, and 0.792, 0.745, and 0.781 for HER2-positive vs. others, respectively. The proposed DCE-MRI-based deep learning system may have the potential to preoperatively distinct HER2 expressions of breast cancers with therapeutic implications. 4 TECHNICAL EFFICACY: Stage 3.

Liver tumor segmentation method combining multi-axis attention and conditional generative adversarial networks.

In modern medical imaging-assisted therapies, manual annotation is commonly employed for liver and tumor segmentation in abdominal CT images. However, this approach suffers from low efficiency and poor accuracy. With the development of deep learning, automatic liver tumor segmentation algorithms based on neural networks have emerged, for the improvement of the work efficiency. However, existing liver tumor segmentation algorithms still have several limitations: (1) they often encounter the common issue of class imbalance in liver tumor segmentation tasks, where the tumor region is significantly smaller than the normal tissue region, causing models to predict more negative samples and neglect the tumor region; (2) they fail to adequately consider feature fusion between global contexts, leading to the loss of crucial information; (3) they exhibit weak perception of local details such as fuzzy boundaries, irregular shapes, and small lesions, thereby failing to capture important features. To address these issues, we propose a Multi-Axis Attention Conditional Generative Adversarial Network, referred to as MA-cGAN. Firstly, we propose the Multi-Axis attention mechanism (MA) that projects three-dimensional CT images along different axes to extract two-dimensional features. The features from different axes are then fused by using learnable factors to capture key information from different directions. Secondly, the MA is incorporated into a U-shaped segmentation network as the generator to enhance its ability to extract detailed features. Thirdly, a conditional generative adversarial network is built by combining a discriminator and a generator to enhance the stability and accuracy of the generator's segmentation results. The MA-cGAN was trained and tested on the LiTS public dataset for the liver and tumor segmentation challenge. Experimental results show that MA-cGAN improves the Dice coefficient, Hausdorff distance, average surface distance, and other metrics compared to the state-of-the-art segmentation models. The segmented liver and tumor models have clear edges, fewer false positive regions, and are closer to the true labels, which plays an active role in medical adjuvant therapy. The source code with our proposed model are available at https://github.com/jhliao0525/MA-cGAN.git.

Deep Learning Technique for automatic Liver and Liver Tumor Segmentation in CT Images

Deep Learning Technique for automatic Liver and Liver Tumor Segmentation in CT Images

Automatic segmentation, classification, and prediction of pelvic bone tumors using deep learning techniques

AbstractMRI and CT images are widely utilized for detecting tumors in internal organs. However, accurately distinguishing tumors using these images only, poses a challenge. This demands human expertise to effectively mark tumors in a given MRI or CT image, which is time-consuming. In this study, MONAI Label and MONAI Dynamic U-Net (DynU-Net) were used to segment tumors in any given 3D CT image of the pelvic bone, aiming to assist radiologists further. This method helps to run the model without needing a GPU which is better than traditional approaches. In addition, a convolutional neural network (CNN) was used to classify the tumors as benign or malignant and to predict three grades of tumors (low, medium, and high). The use of CNN in classification and prediction gives higher results than other studies. A dataset of 178 3D CT picture images was employed to feed the networks with the help of Adam optimizer and Categorical cross-entropy. We employ a set of 3D CT scans because of their greater contrast and spatial resolution which is better used for pelvic bone tumors. The Dice similarity coefficient (DSC) and Hausdorff distance (HD) have been used to perform calculations in segmentation. The results show a DSC of 0.7660280 and an HD of 16.75480. A range of performance metrics, such as sensitivity, specification, and F1-score for classification and prediction methods, are used to evaluate the accuracy of the proposed system. The system has accuracy (99.4%) for classification and (97.8%) for prediction. These findings indicate that MONAI Label is effective for automatic segmentation of tumors in a given CT scan; with high accuracy. Moreover, CNN is useful for classification and prediction systems with high accuracy. However, achieving better results is possible with an abundance of training samples.

FEBE-Net: Feature Exploration Attention and Boundary Enhancement Refinement Transformer Network for Bladder Tumor Segmentation

The automatic and accurate segmentation of bladder tumors is a key step in assisting urologists in diagnosis and analysis. At present, existing Transformer-based methods have limited ability to restore local detail features and insufficient boundary segmentation capabilities. We propose FEBE-Net, which aims to effectively capture global and remote semantic features, preserve more local detail information, and provide clearer and more precise boundaries. Specifically, first, we use PVT v2 backbone to learn multi-scale global feature representations to adapt to changes in bladder tumor size and shape. Secondly, we propose a new feature exploration attention module (FEA) to fully explore the potential local detail information in the shallow features extracted by the PVT v2 backbone, eliminate noise, and supplement the missing fine-grained details for subsequent decoding stages. At the same time, we propose a new boundary enhancement and refinement module (BER), which generates high-quality boundary clues through boundary detection operators to help the decoder more effectively preserve the boundary features of bladder tumors and refine and adjust the final predicted feature map. Then, we propose a new efficient self-attention calibration decoder module (ESCD), which, with the help of boundary clues provided by the BER module, gradually and effectively recovers global contextual information and local detail information from high-level features after calibration enhancement and low-level features after exploration attention. Extensive experiments on the cystoscopy dataset BtAMU and five colonoscopy datasets have shown that FEBE-Net outperforms 11 state-of-the-art (SOTA) networks in segmentation performance, with higher accuracy, stronger robust stability, and generalization ability.

NIMG-54. AUTOMATED IMAGING RESPONSE EVALUATION SYSTEM (AIRES) FOR NEURO-ONCOLOGY: PROTOTYPE SOFTWARE TOOL FOR DETERMINING STANDARDIZED RADIOGRAPHIC RESPONSE ASSESSMENT FOR GLIOBLASTOMA CLINICAL TRIALS

Abstract Standardized radiographic response assessment criteria like Modified Response Assessment in Neuro-Oncology (mRANO) and now the RANO 2.0 are utilized for evaluating progression-free survival and objective response rate during glioblastoma clinical trials. After measuring tumor size with either bidimensional diameters or volumetric segmentations, the crucial and final time-consuming step of categorizing progression, stable disease, and response based on multiple criteria and numerical thresholds (e.g. ≥40% volumetric growth = progressive disease) is currently done manually. We developed an application prototype, Automated Imaging Response Evaluation System (AIRES) for Neuro-Oncology, that uses the mRANO criteria to label each timepoint as treatment response, stable disease, or disease progression based on either bidimensional or volumetric measurements as well as optional inputs of clinical status and corticosteroid dosage. These inputs may be uploaded via CSV files, which can integrate with automated tumor segmentation volumetry workflows, or entered in the AIRES graphical user interface (GUI). The AIRES GUI displays a patient report consisting of a table and graph that contain the lesion size, percent change from baseline, percent change from nadir, and the mRANO response category for each timepoint. To test AIRES in a real-world application, tumor size measurements from 367 scans across 41 patients from NUTMEG, a phase II multicenter clinical trial, were used to perform mRANO reads and calculate progression-free survival and time to response. AIRES-assisted reads performed by reader 1 were significantly quicker than fully manual reads performed by reader 2 (37.1±8.6 vs 107.7±58.6 seconds, P&amp;lt;0.0001). These results show that the AIRES app can make the reads quicker and potentially feasible for non-trained personnel, while reducing the amount of error-prone manual steps. Additionally, the underlying logic of AIRES is easily adaptable to integrate other response criteria including but not limited to RANO 2.0 and other future response assessment updates.

Automatic tumor segmentation and lymph node metastasis prediction in papillary thyroid carcinoma using ultrasound keyframes.

Accurate preoperative prediction of cervical lymph node metastasis (LNM) for papillary thyroid carcinoma (PTC) patients is essential for disease staging and individualized treatment planning, which can improve prognosis and facilitate better management. To establish a fully automated deep learning-enabled model (FADLM) for automated tumor segmentation and cervical LNM prediction in PTC using ultrasound (US) video keyframes. The bicentral study retrospective enrolled 518 PTC patients, who were then randomly divided into the training (Hospital 1, n=340), internal test (Hospital 1, n=83), and external test cohorts (Hospital 2, n=95). The FADLM integrated mask region-based convolutional neural network (Mask R-CNN) for automatic thyroid primary tumor segmentation and ResNet34 with Bayes strategy for cervical LNM diagnosis. A radiomics model (RM) using the same automated segmentation method, a traditional radiomics model (TRM) using manual segmentation, and a clinical-semantic model (CSM) were developed for comparison. The dice similarity coefficient (DSC) was used to evaluate segmentation performance. The prediction performance of the models was validated in terms of discrimination and clinical utility with the area under the receiver operator characteristic curve (AUC), heatmap analysis, and decision curve analysis (DCA). The comparison of the predictive performance among different models was conducted by DeLong test. The performances of two radiologists compared with FADLM and the diagnostic augmentation with FADLM's assistance were analyzed in terms of accuracy, sensitivity and specificity using McNemar's x2 test. The p-value less than 0.05 was defined as a statistically significant difference. The Benjamini-Hochberg procedure was applied for multiple comparisons to deal with Type I error. The FADLM yielded promising segmentation results in training (DSC: 0.88±0.23), internal test (DSC: 0.88±0.23), and external test cohorts (DSC: 0.85±0.24). The AUCs of FADLM for cervical LNM prediction were 0.78 (95% CI: 0.73, 0.83), 0.83 (95% CI: 0.74, 0.92), and 0.83 (95% CI: 0.75, 0.92), respectively. It all significantly outperformed the RM (AUCs: 0.78vs. 0.72; 0.83vs. 0.65; 0.83vs. 0.68, all adjusted p-values<0.05) and CSM (AUCs: 0.78vs. 0.71; 0.83vs. 0.62; 0.83vs. 0.68, all adjusted p-values<0.05) across the three cohorts. The RM offered similar performance to that of TRM (AUCs: 0.61vs. 0.63, adjusted p-value =0.60) while significantly reducing the segmentation time (3.3±3.8 vs. 14.1±4.2 s, p-value <0.001). Under the assistance of FADLM, the accuracies of junior and senior radiologists were improved by 18% and 15% (all adjusted p-values<0.05) and the sensitivities by 25% and 21% (all adjusted p-values<0.05) in the external test cohort. The FADLM with elaborately designed automated strategy using US video keyframes holds good potential to provide an efficient and consistent prediction of cervical LNM in PTC. The FADLM displays superior performance to RM, CSM, and radiologists with promising efficacy.

A convolutional attention model for predicting response to chemo-immunotherapy from ultrasound elastography in mouse tumor models

BackgroundIn the era of personalized cancer treatment, understanding the intrinsic heterogeneity of tumors is crucial. Despite some patients responding favorably to a particular treatment, others may not benefit, leading to the varied efficacy observed in standard therapies. This study focuses on the prediction of tumor response to chemo-immunotherapy, exploring the potential of tumor mechanics and medical imaging as predictive biomarkers. We have extensively studied “desmoplastic” tumors, characterized by a dense and very stiff stroma, which presents a substantial challenge for treatment. The increased stiffness of such tumors can be restored through pharmacological intervention with mechanotherapeutics.MethodsWe developed a deep learning methodology based on shear wave elastography (SWE) images, which involved a convolutional neural network (CNN) model enhanced with attention modules. The model was developed and evaluated as a predictive biomarker in the setting of detecting responsive, stable, and non-responsive tumors to chemotherapy, immunotherapy, or the combination, following mechanotherapeutics administration. A dataset of 1365 SWE images was obtained from 630 tumors from our previous experiments and used to train and successfully evaluate our methodology. SWE in combination with deep learning models, has demonstrated promising results in disease diagnosis and tumor classification but their potential for predicting tumor response prior to therapy is not yet fully realized.ResultsWe present strong evidence that integrating SWE-derived biomarkers with automatic tumor segmentation algorithms enables accurate tumor detection and prediction of therapeutic outcomes.ConclusionsThis approach can enhance personalized cancer treatment by providing non-invasive, reliable predictions of therapeutic outcomes.

Automatic detection and segmentation of lesions in 18 F-FDG PET/CT imaging of patients with Hodgkin lymphoma using 3D dense U-Net.

The accuracy of automatic tumor segmentation in PET/computed tomography (PET/CT) images is crucial for the effective treatment and monitoring of Hodgkin lymphoma. This study aims to address the challenges faced by certain segmentation algorithms in accurately differentiating lymphoma from normal organ uptakes due to PET image resolution and tumor heterogeneity. Variants of the encoder-decoder architectures are state-of-the-art models for image segmentation. Among these kinds of architectures, U-Net is one of the most famous and predominant for medical image segmentation. In this study, we propose a fully automatic approach for Hodgkin lymphoma segmentation that combines U-Net and DenseNet architectures to reduce network loss for very small lesions, which is trained using the Tversky loss function. The hypothesis is that the fusion of these two deep learning models can improve the accuracy and robustness of Hodgkin lymphoma segmentation. A dataset with 141 samples was used to train our proposed network. Also, to test and evaluate the proposed network, we allocated two separate datasets of 20 samples. We achieved 0.759 as the mean Dice similarity coefficient with a median value of 0.767, and interquartile range (0.647-0.837). A good agreement was observed between the ground truth of test images against the predicted volume with precision and recall scores of 0.798 and 0.763, respectively. This study demonstrates that the integration of U-Net and DenseNet architectures, along with the Tversky loss function, can significantly enhance the accuracy of Hodgkin lymphoma segmentation in PET/CT images compared to similar studies.

Applications of artificial intelligence in interventional oncology: An up-to-date review of the literature.

Interventional oncology provides image-guided therapies, including transarterial tumor embolization and percutaneous tumor ablation, for malignant tumors in a minimally invasive manner. As in other medical fields, the application of artificial intelligence (AI) in interventional oncology has garnered significant attention. This narrative review describes the current state of AI applications in interventional oncology based on recent literature. A literature search revealed a rapid increase in the number of studies relevant to this topic recently. Investigators have attempted to use AI for various tasks, including automatic segmentation of organs, tumors, and treatment areas; treatment simulation; improvement of intraprocedural image quality; prediction of treatment outcomes; and detection of post-treatment recurrence. Among these, the AI-based prediction of treatment outcomes has been the most studied. Various deep and conventional machine learning algorithms have been proposed for these tasks. Radiomics has often been incorporated into prediction and detection models. Current literature suggests that AI is potentially useful in various aspects of interventional oncology, from treatment planning to post-treatment follow-up. However, most AI-based methods discussed in this review are still at the research stage, and few have been implemented in clinical practice. To achieve widespread adoption of AI technologies in interventional oncology procedures, further research on their reliability and clinical utility is necessary. Nevertheless, considering the rapid research progress in this field, various AI technologies will be integrated into interventional oncology practices in the near future.

Automated segmentation and source prediction of bone tumors using ConvNeXtv2 Fusion based Mask R-CNN to identify lung cancer metastasis

Lung cancer, which is a leading cause of cancer-related deaths worldwide, frequently metastasizes to the bones, significantly diminishing patients’ quality of life and complicating treatment strategies. This study aims to develop an advanced 3D Mask R-CNN model, enhanced with the ConvNeXt-V2 backbone, for the automatic segmentation of bone tumors and identification of lung cancer metastasis to support personalized treatment planning. Data were collected from two hospitals: Center A (106 patients) and Center B (265 patients). The data from Center B were used for training, while Center A’s dataset served as an independent external validation set. High-resolution CT scans with 1 mm slice thickness and no inter-slice gaps were utilized, and the regions of interest (ROIs) were manually segmented and validated by two experienced radiologists. The 3D Mask R-CNN model achieved a Dice Similarity Coefficient (DSC) of 0.856, a sensitivity of 0.921, and a specificity of 0.961 on the training set. On the test set, it achieved a DSC of 0.849, a sensitivity of 0.911, and a specificity of 0.931. For the classification task, the model attained an AUC of 0.865, an accuracy of 0.866, a sensitivity of 0.875, and a specificity of 0.835 on the training set, while achieving an AUC of 0.842, an accuracy of 0.836, a sensitivity of 0.847, and a specificity of 0.819 on the test set. These results highlight the model’s potential in improving the accuracy of bone tumor segmentation and lung cancer metastasis detection, paving the way for enhanced diagnostic workflows and personalized treatment strategies in clinical oncology.

Deep learning with uncertainty estimation for automatic tumor segmentation in PET/CT of head and neck cancers: impact of model complexity, image processing and augmentation

Objective. Target volumes for radiotherapy are usually contoured manually, which can be time-consuming and prone to inter- and intra-observer variability. Automatic contouring by convolutional neural networks (CNN) can be fast and consistent but may produce unrealistic contours or miss relevant structures. We evaluate approaches for increasing the quality and assessing the uncertainty of CNN-generated contours of head and neck cancers with PET/CT as input. Approach. Two patient cohorts with head and neck squamous cell carcinoma and baseline 18F-fluorodeoxyglucose positron emission tomography and computed tomography images (FDG-PET/CT) were collected retrospectively from two centers. The union of manual contours of the gross primary tumor and involved nodes was used to train CNN models for generating automatic contours. The impact of image preprocessing, image augmentation, transfer learning and CNN complexity, architecture, and dimension (2D or 3D) on model performance and generalizability across centers was evaluated. A Monte Carlo dropout technique was used to quantify and visualize the uncertainty of the automatic contours. Main results. CNN models provided contours with good overlap with the manually contoured ground truth (median Dice Similarity Coefficient: 0.75–0.77), consistent with reported inter-observer variations and previous auto-contouring studies. Image augmentation and model dimension, rather than model complexity, architecture, or advanced image preprocessing, had the largest impact on model performance and cross-center generalizability. Transfer learning on a limited number of patients from a separate center increased model generalizability without decreasing model performance on the original training cohort. High model uncertainty was associated with false positive and false negative voxels as well as low Dice coefficients. Significance. High quality automatic contours can be obtained using deep learning architectures that are not overly complex. Uncertainty estimation of the predicted contours shows potential for highlighting regions of the contour requiring manual revision or flagging segmentations requiring manual inspection and intervention.

One-click annotation to improve segmentation by a convolutional neural network for PET images of head and neck cancer patients

A convolutional neural network (CNN) can be used to perform fully automatic tumor segmentation from the positron emission tomography (PET) images of head and neck cancer patients but the predictions often contain false positive segmentation caused by the high concentration of the tracer substance in the human brain. A potential solution would be a one-click annotation in which a user points the location of the tumor by clicking the image. This information can then be given either directly to a CNN or an algorithm that fixes its predictions. In this article, we compare the fully automatic segmentation to four semi-automatic approaches by using 962 transaxial slices collected from the PET images of 100 head and neck cancer patients. According to our results, a semi-automatic segmentation method with information about the center of the tumor performs the best with a median Dice score of 0.708.

Automated brain tumor diagnostics: Empowering neuro-oncology with deep learning-based MRI image analysis.

Brain tumors, characterized by the uncontrolled growth of abnormal cells, pose a significant threat to human health. Early detection is crucial for successful treatment and improved patient outcomes. Magnetic Resonance Imaging (MRI) is the primary diagnostic tool for brain tumors, providing detailed visualizations of the brain's intricate structures. However, the complexity and variability of tumor shapes and locations often challenge physicians in achieving accurate tumor segmentation on MRI images. Precise tumor segmentation is essential for effective treatment planning and prognosis. To address this challenge, we propose a novel hybrid deep learning technique, Convolutional Neural Network and ResNeXt101 (ConvNet-ResNeXt101), for automated tumor segmentation and classification. Our approach commences with data acquisition from the BRATS 2020 dataset, a benchmark collection of MRI images with corresponding tumor segmentations. Next, we employ batch normalization to smooth and enhance the collected data, followed by feature extraction using the AlexNet model. This involves extracting features based on tumor shape, position, shape, and surface characteristics. To select the most informative features for effective segmentation, we utilize an advanced meta-heuristics algorithm called Advanced Whale Optimization (AWO). AWO mimics the hunting behavior of humpback whales to iteratively search for the optimal feature subset. With the selected features, we perform image segmentation using the ConvNet-ResNeXt101 model. This deep learning architecture combines the strengths of ConvNet and ResNeXt101, a type of ConvNet with aggregated residual connections. Finally, we apply the same ConvNet-ResNeXt101 model for tumor classification, categorizing the segmented tumor into distinct types. Our experiments demonstrate the superior performance of our proposed ConvNet-ResNeXt101 model compared to existing approaches, achieving an accuracy of 99.27% for the tumor core class with a minimum learning elapsed time of 0.53 s.

Renal tumor segmentation, visualization, and segmentation confidence using ensembles of neural networks in patients undergoing surgical resection.

To develop an automatic segmentation model for solid renal tumors on contrast-enhanced CTs and to visualize segmentation with associated confidence to promote clinical applicability. The training dataset included solid renal tumor patients from two tertiary centers undergoing surgical resection and receiving CT in the corticomedullary or nephrogenic contrast media (CM) phase. Manual tumor segmentation was performed on all axial CT slices serving as reference standard for automatic segmentations. Independent testing was performed on the publicly available KiTS 2019 dataset. Ensembles of neural networks (ENN, DeepLabV3) were used for automatic renal tumor segmentation, and their performance was quantified with DICE score. ENN average foreground entropy measured segmentation confidence (binary: successful segmentation with DICE score > 0.8 versus inadequate segmentation ≤ 0.8). N = 639/n = 210 patients were included in the training and independent test dataset. Datasets were comparable regarding age and sex (p > 0.05), while renal tumors in the training dataset were larger and more frequently benign (p < 0.01). In the internal test dataset, the ENN model yielded a median DICE score = 0.84 (IQR: 0.62-0.97, corticomedullary) and 0.86 (IQR: 0.77-0.96, nephrogenic CM phase), and the segmentation confidence an AUC = 0.89 (sensitivity = 0.86; specificity = 0.77). In the independent test dataset, the ENN model achieved a median DICE score = 0.84 (IQR: 0.71-0.97, corticomedullary CM phase); and segmentation confidence an accuracy = 0.84 (sensitivity = 0.86 and specificity = 0.81). ENN segmentations were visualized with color-coded voxelwise tumor probabilities and thresholds superimposed on clinical CT images. ENN-based renal tumor segmentation robustly performs in external test data and might aid in renal tumor classification and treatment planning. Ensembles of neural networks (ENN) models could automatically segment renal tumors on routine CTs, enabling and standardizing downstream image analyses and treatment planning. Providing confidence measures and segmentation overlays on images can lower the threshold for clinical ENN implementation. Ensembles of neural networks (ENN) segmentation is visualized by color-coded voxelwise tumor probabilities and thresholds. ENN provided a high segmentation accuracy in internal testing and in an independent external test dataset. ENN models provide measures of segmentation confidence which can robustly discriminate between successful and inadequate segmentations.

Retraction Note: Automatic segmentation and classification of liver tumor from CT image using feature difference and SVM based classifier-soft computing technique

Retraction Note: Automatic segmentation and classification of liver tumor from CT image using feature difference and SVM based classifier-soft computing technique

Integrating a deep neural network and Transformer architecture for the automatic segmentation and survival prediction in cervical cancer.

Automated tumor segmentation and survival prediction are critical to clinical diagnosis and treatment. This study aimed to develop deep-learning models for automatic tumor segmentation and survival prediction in magnetic resonance imaging (MRI) of cervical cancer (CC) by combining deep neural networks and Transformer architecture. This study included 406 patients with CC, each with comprehensive clinical information and MRI scans. We randomly divided patients into training, validation, and independent test cohorts in a 6:2:2 ratio. During the model training, we employed two architecture types: one being a hybrid model combining convolutional neural network (CNN) and ransformer (CoTr) and one of pure CNNs. For survival prediction, the hybrid model combined tumor image features extracted by segmentation models with clinical information. The performance of the segmentation models was evaluated using the Dice similarity coefficient (DSC) and 95% Hausdorff distance (HD95). The performance of the survival models was assessed using the concordance index. The CoTr model performed well in both contrast-enhanced T1-weighted (ceT1W) and T2-weighted (T2W) imaging segmentation tasks, with average DSCs of 0.827 and 0.820, respectively, which outperformed other the CNN models such as U-Net (DSC: 0.807 and 0.808), attention U-Net (DSC: 0.814 and 0.811), and V-Net (DSC: 0.805 and 0.807). For survival prediction, the proposed deep-learning model significantly outperformed traditional methods, yielding a concordance index of 0.732. Moreover, it effectively divided patients into low-risk and high-risk groups for disease progression (P<0.001). Combining Transformer architecture with a CNN can improve MRI tumor segmentation, and this deep-learning model excelled in the survival prediction of patients with CC as compared to traditional methods.

A Dynamic Context Encoder Network for Liver Tumor Segmentation.

Accurate segmentation of liver tumor regions in medical images is of great significance for clinical diagnosis and the planning of surgical treatments. Recent advancements in machine learning have shown that convolutional neural networks are powerful in such image processing while largely reducing human labor. However, the variable shape, fuzzy boundary, and discontinuous tumor region of liver tumors in medical images bring great challenges to accurate segmentation. The feature extraction capability of a neural network can be improved by expanding its architecture, but it inevitably demands more computing resources in training and hyperparameter tuning. This study presents a Dynamic Context Encoder Network (DCE-Net), which incorporates multiple new modules, such as the Involution Layer, Dynamic Residual Module, Context Extraction Module, and Channel Attention Gates, for feature extraction and enhancement. In the experiment, we used a liver tumor CT dataset of LiTS2017 to train and test the DCE-Net for liver tumor segmentation. The experimental results showed that the four evaluation indexes of the method, precision, recall, dice, and AUC, were 0.8961, 0.9711, 0.9270, and 0.9875, respectively. Furthermore, our ablation study reported that the accuracy and training efficiency of our network were markedly superior to the networks without involution or dynamic residual modules. Therefore, the DCE-Net proposed in this study has great potential for automatic segmentation of liver lesion tumors in the clinical diagnostic environment.

A Multi-Center, Multi-Parametric MRI Dataset of Primary and Secondary Brain Tumors

Brain metastases (BMs) and high-grade gliomas (HGGs) are the most common and aggressive types of malignant brain tumors in adults, with often poor prognosis and short survival. As their clinical symptoms and image appearances on conventional magnetic resonance imaging (MRI) can be astonishingly similar, their accurate differentiation based solely on clinical and radiological information can be very challenging, particularly for “cancer of unknown primary”, where no systemic malignancy is known or found. Non-invasive multiparametric MRI and radiomics offer the potential to identify these distinct biological properties, aiding in the characterization and differentiation of HGGs and BMs. However, there is a scarcity of publicly available multi-origin brain tumor imaging data for tumor characterization. In this paper, we introduce a multi-center, multi-origin brain tumor MRI (MOTUM) imaging dataset obtained from 67 patients: 29 with high-grade gliomas, 20 with lung metastases, 10 with breast metastases, 2 with gastric metastasis, 4 with ovarian metastasis, and 2 with melanoma metastasis. This dataset includes anonymized DICOM files alongside processed FLAIR, T1-weighted, contrast-enhanced T1-weighted, T2-weighted sequences images, segmentation masks of two tumor regions, and clinical data. Our data-sharing initiative is to support the benchmarking of automated tumor segmentation, multi-modal machine learning, and disease differentiation of multi-origin brain tumors in a multi-center setting.

Attention-guided context asymmetric fusion networks for the liver tumor segmentation of computed tomography images.

Liver tumor segmentation based on medical imaging is playing an increasingly important role in liver tumor research and individualized therapeutic decision-making. However, it remains a challenging in terms of the accuracy of automatic segmentation of liver tumors. Therefore, we aimed to develop a novel deep neural network for improving the results from the automatic segmentation of liver tumors. This paper proposes the attention-guided context asymmetric fusion network (AGCAF-Net), combining attention guidance and fusion context modules on the basis of a residual neural network for the automatic segmentation of liver tumors. According to the attention-guided context block (AGCB), the feature map is first divided into multiple small blocks, the local correlation between features is calculated, and then the global nonlocal fusion module (GNFM) is used to obtain the global information between pixels. Additionally, the context pyramid module (CPM) and asymmetric semantic fusion module (AFM) are used to obtain multiscale features and resolve the feature mismatch during feature fusion, respectively. Finally, we used the liver tumor segmentation benchmark (LiTS) dataset to verify the efficiency of our designed network. Our results showed that AGCAF-Net with AFM and CPM is effective in improving the accuracy of liver tumor segmentation, with the Dice coefficient increasing from 82.5% to 84.1%. The segmentation results of liver tumors by AGCAF-Net were superior to those of several state-of-the-art U-net methods, with a Dice coefficient of 84.1%, a sensitivity of 91.7%, and an average symmetric surface distance of 3.52. AGCAF-Net can obtain better matched and accurate segmentation in liver tumor segmentation, thus effectively improving the accuracy of liver tumor segmentation.