Tumor Segmentation Research Articles

Augmented surgical decision-making for glioblastoma: integrating AI tools into education and practice.

Surgical decision-making for glioblastoma poses significant challenges due to its complexity and variability. This study investigates the potential of artificial intelligence (AI) tools in improving "decision-making processes" for glioblastoma surgery. A systematic review of literature identified 10 relevant studies, primarily focused on predicting resectability and surgery-related neurological outcomes. AI tools, especially rooted in radiomics and connectomics, exhibited promise in predicting resection extent through precise tumor segmentation and tumor-network relationships. However, they demonstrated limited effectiveness in predicting postoperative neurological due to dynamic and less quantifiable nature of patient-related factors. Recognizing these challenges, including limited datasets and the interpretability requirement in medical applications, underscores the need for standardization, algorithm optimization, and addressing variability in model performance and then further validation in clinical settings. While AI holds potential, it currently does not possess the capacity to emulate the nuanced decision-making process utilized by experienced neurosurgeons in the comprehensive approach to glioblastoma surgery.

New Approach for Brain Tumor Segmentation Based on Gabor Convolution and Attention Mechanism

New Approach for Brain Tumor Segmentation Based on Gabor Convolution and Attention Mechanism

MDSK‐Net: Multi‐scale dynamic segmentation kernel network for renal tumour endoscopic image segmentation

AbstractAutomatic segmentation of renal tumours during renal cell carcinoma surgery can help doctors accurately locate the tumour region, protect the tissues and organs around the kidneys, enhance surgical efficiency, and reduce the possibility of leakage and misdiagnosis. However, since general polyp endoscopic image segmentation models have many problems when facing the task of renal tumour segmentation, there needs to be more research on the segmentation of endoscopic images of renal tumours. This paper proposes a multi‐scale dynamic segmentation kernel network for endoscopic image segmentation of kidney tumours. First, a spatial receptive field module is proposed to augment the feature information and improve the performance of the whole network. Second, an enhanced cross‐attention module is offered to attenuate the effect of a high‐similarity segmentation background. Finally, a multi‐scale dynamic segmentation kernel module is introduced to gradually refine the segmentation results from small to large sizes to obtain more accurate tumour boundaries. Extensive experiments on the established kidney tumour endoscopic dataset and publicly available endoscopic datasets show that this method exhibits enhanced performance and generalization capabilities compared to existing techniques. On this renal tumour dataset, MDSK‐Net achieved excellent results of 94.1% and 90.1% on mDice and mIoU.

An automated two-stage approach to kidney and tumor segmentation in CT imaging.

The incidence of kidney tumors is progressively increasing each year. The precision of segmentation for kidney tumors is crucial for diagnosis and treatment. To enhance accuracy and reduce manual involvement, propose a deep learning-based method for the automatic segmentation of kidneys and kidney tumors in CT images. The proposed method comprises two parts: object detection and segmentation. We first use a model to detect the position of the kidney, then narrow the segmentation range, and finally use an attentional recurrent residual convolutional network for segmentation. Our model achieved a kidney dice score of 0.951 and a tumor dice score of 0.895 on the KiTS19 dataset. Experimental results show that our model significantly improves the accuracy of kidney and kidney tumor segmentation and outperforms other advanced methods. The proposed method provides an efficient and automatic solution for accurately segmenting kidneys and renal tumors on CT images. Additionally, this study can assist radiologists in assessing patients' conditions and making informed treatment decisions.

Resolving spatial response heterogeneity in glioblastoma.

Spatial intratumoral heterogeneity poses a significant challenge for accurate response assessment in glioblastoma. Multimodal imaging coupled with advanced image analysis has the potential to unravel this response heterogeneity. Based on automated tumor segmentation and longitudinal registration with follow-up imaging, we categorized contrast-enhancing voxels of 61 patients with suspected recurrence of glioblastoma into either true tumor progression (TP) or pseudoprogression (PsP). To allow the unbiased analysis of semantically related image regions, adjacent voxels with similar values of cerebral blood volume (CBV), FET-PET, and contrast-enhanced T1w were automatically grouped into supervoxels. We then extracted first-order statistics as well as texture features from each supervoxel. With these features, a Random Forest classifier was trained and validated employing a 10-fold cross-validation scheme. For model evaluation, the area under the receiver operating curve, as well as classification performance metrics were calculated. Our image analysis pipeline enabled reliable spatial assessment of tumor response. The predictive model reached an accuracy of 80.0% and a macro-weighted AUC of 0.875, which takes class imbalance into account, in the hold-out samples from cross-validation on supervoxel level. Analysis of feature importances confirmed the significant role of FET-PET-derived features. Accordingly, TP- and PsP-labeled supervoxels differed significantly in their 10th and 90th percentile, as well as the median of tumor-to-background normalized FET-PET. However, CBV- and T1c-related features also relevantly contributed to the model's performance. Disentangling the intratumoral heterogeneity in glioblastoma holds immense promise for advancing precise local response evaluation and thereby also informing more personalized and localized treatment strategies in the future.

MRI Brain Tumor Segmentation using Cuckoo Optimization and Ensemble CNNs

MRI Brain Tumor Segmentation using Cuckoo Optimization and Ensemble CNNs

A Novel Liver Tumor segmentation of Adverse Propagation Advanced Swin Transformer Network with Mask region-based convolutional neural networks

The diagnosis and treatment of liver diseases from computed tomography (CT) images is an indispensable task for segmentation of Liver & its tumours. Due to the uneven presence, fuzzy borders, diverse densities, shapes and sizes of lesions segmentation of liver & its tumour is a difficult task. At this point we mainly focused on deep learning algorithms for segmenting liver and its tumour from abdominal CT scan images thereafter minimising the time & energy used for a liver diseases diagnosis This study aims to classify and segment liver tumors using a novel deep learning-based model. A Mask region-based convolutional neural network (Mask R-CNN) model is proposed for multiorgan segmentation to aid esophageal radiation treatment. Due to the fact that organ boundaries may be fuzzy and organ shapes are various, original Mask R-CNN works well on natural image segmentation while leaves something to be desired on the multiorgan segmentation task. Addressing it, the advantages of this method are threefold: (1) a ROI (region of interest) generation method is presented in the RPN (region proposal network) which is able to utilize multiscale semantic features. (2) A prebackground classification subnetwork is integrated to the original mask generation branch to improve the precision of multiorgan segmentation. The segmented image is added to the Adverse Propagation Advanced Swin Transformer Network (APESTNet) to prevent overfitting. The proposed model is based on CT volume slices of patients with liver tumors and evaluated on the public 3D dataset IRCADB01.The proposed model dataset is split into training as 60% and testing as 40 % for classification. The proposed model's accuracy and recall are 99.23 % and 98.24 %, respectively.

Machine learning-based anatomical segmentation: A systematic review of methodologies and applications

Image segmentation has established itself as a cornerstone of modern medical imaging practices, achieving state-of-the-art accuracy and playing a pivotal role in critical applications such as pathology diagnosis, treatment planning, tumour segmentation and organ delineation to name a few. The integration of deep learning and neural network techniques has sparked a dynamic and constantly evolving landscape, propelling rapid advancements in image segmentation. This work aims to provide an all encompassing review of the general state-of-the-art, with a focus on developments made within the last few years. Over 100 papers presenting development in the field have been considered, primarily published between 2020-2023, separated into their architectural type. Within each type, significant works and conclusions have been highlighted. We conclude the paper with a discussion of the challenges facing the segmentation landscape as a whole, and how these challenges have been best circumnavigated to date.

From quantitative metrics to clinical success: assessing the utility of deep learning for tumor segmentation in breast surgery.

Preventing positive margins is essential for ensuring favorable patient outcomes following breast-conserving surgery (BCS). Deep learning has the potential to enable this by automatically contouring the tumor and guiding resection in real time. However, evaluation of such models with respect to pathology outcomes is necessary for their successful translation into clinical practice. Sixteen deep learning models based on established architectures in the literature are trained on 7318 ultrasound images from 33 patients. Models are ranked by an expert based on their contours generated from images in our test set. Generated contours from each model are also analyzed using recorded cautery trajectories of five navigated BCS cases to predict margin status. Predicted margins are compared with pathology reports. The best-performing model using both quantitative evaluation and our visual ranking framework achieved a mean Dice score of 0.959. Quantitative metrics are positively associated with expert visual rankings. However, the predictive value of generated contours was limited with a sensitivity of 0.750 and a specificity of 0.433 when tested against pathology reports. We present a clinical evaluation of deep learning models trained for intraoperative tumor segmentation in breast-conserving surgery. We demonstrate that automatic contouring is limited in predicting pathology margins despite achieving high performance on quantitative metrics.

Multimodal fully automated predictive model for therapeutic efficacy of first-line cancer immunotherapy based on clinical information and imaging modalities including brain MRI and chest CT images in advanced non-small cell lung cancer.

1555 Background: The integration of imaging and clinical data through artificial intelligence (AI) holds promise for more accurately predicting the therapeutic efficacy of immune checkpoint inhibitors (ICI) ± platinum-based chemotherapy in individual patients with advanced non-small cell lung cancer (NSCLC). However, tumor segmentation still relies entirely on radiologist evaluation and has not yet achieved fully automated tumor evaluation for image analysis. Furthermore, while central nervous system (CNS) metastasis is a critical prognostic factor in patients with advanced NSCLC, radiomics has primarily focused on chest imaging, with few studies incorporating brain MRI analysis. Methods: This study included patients with advanced NSCLC treated with ICI ± platinum-based chemotherapy as first-line therapy from March 2017 to August 2023. We developed an ensemble model combining a logistic regression model that analyzes clinical information with a deep learning model that extracts image features. Each sub-model was independent, allowing for individual adjustments and the ability to be attached or detached. Both models were trained to predict response, 12-month progression-free survival (PFS), and 12-month overall survival (OS) using cross-entropy loss. The dataset was randomly divided into training, validation, and test sets in a 5:3:2 ratio. This random assignment was repeated four times, and the average area under the curve (AUC) was calculated. The test sets were further classified into high- and low-risk groups based on the predicted results, and log-rank tests were performed on their survival curves. Results: The study included 232 patients, with 89 (38%) receiving ICI alone and 143 (62%) receiving ICI + platinum-based chemotherapy. In the entire population, the overall response rate (ORR) for ICI ± platinum-based chemotherapy was 49.5%, with a median PFS of 7.1 months (95% confidence interval [CI]: 5.8-9.1) and a median OS of 24.5 months (95% CI: 21.2-30.3). The model achieved an AUC of 0.70 (95% CI: 0.65- 0.76) for predicting response, 0.65 (95% CI: 0.59-0.71) for 12-month PFS, and 0.69 (95% CI: 0.48-0.90) for 12-month OS. Test sets were classified into high-risk (N = 55) and low-risk (N = 60) groups. The median PFS was 6.8 months in the high-risk group and 14.7 months in the low-risk group (p < 0.05). The median OS was 21.5 months in the high-risk group and 34.3 months in the low-risk group (p < 0.05). Conclusions: We have demonstrated that a multimodal, fully automated ensemble model has significant predictive accuracy for the efficacy and survival outcomes of ICI ± platinum-based chemotherapy in patients with advanced NSCLC.

Convolutional neural network and glioma genotyping: A systematic review.

e14038 Background: Gliomas are the most common primary intracranial tumors. Overall, there are 23 different glioma types, and the grading and prognosis of each type are determined by their specific mutations. Several identified mutations are IDH gene status, CDKN2A/B homozygote delegation, PTEN mutation, p53 mutation, TERT promotor mutation, H3K27M mutation, MGMT promotor methylation, trisomy 7/monozomy10, EGFR amplification, ATRX mutation, 1p/19q codeletion, and BRAF mutation. These specific mutations and genomic features can have their distinctive radiologic features seen in different magnetic resonance imaging (MRI) sequences. By using a Convolutional Neural Network (CNN) for feature extraction, the genotyping power of MRI for glioma could be increased. To address this, we systematically reviewed the use of different CNN-based models to detect the genetic context of glioma based on the MRI images. Methods: We conducted a systematic search in PubMed, Embase, Scopus, Web of Science, and Google Scholar. We included original peer-reviewed human studies (retrospective or prospective), focusing on the use of CNN in individuals with glioma. We excluded animal studies, studies with no CNN-based models, reviews, book chapters, conference abstracts, and case reports. JBI checklists were utilized for quality assessment. Results: The initial search resulted in 973 articles. after removing duplicates, 563 studies entered the screening process by title and abstract. 59 studies met our inclusion criteria (56 retrospective and 3 prospective cohort studies) with a total of 26434 glioma patients included (studies ranged from 21 – 2648 individuals). 31 studies used CNN methods only in different stages of image processing (tumor segmentation, feature extraction, image encoder) and 28 studies used CNN as a classifier and mutation predictor. ResNet was the most commonly used CNN-based model (n = 14). U-net (n = 7) and DenseNet models (n = 4) were other commonly used CNN-based models. Included studies targeted IDH status (n=44), 1p/19q codeletion (n=18), MGMT promotor methylation (n=11), ATRX mutation (n=5), TERT promotor mutation (n=5), CDKN2A/B homozygote deletion (n=3), PTEN mutation (n=3), p53 mutation (n=3), H3K27M mutation (n=2), trisomy 7/monozomy10 (n=2), EGFR amplification (n=2), and BRAF mutation (n=1). Highly severe methodological heterogeneity among these studies excluded the possibility of meta-analysis. Conclusions: Current studies have shown that, as a non-invasive method, using CNN-based models on multiple MRI sequences increases the power of early detection and prediction of glioma genetic types before pathological confirmation and has promising results. However, DL performance varies among mutations. Combining multiple DL models with clinical and radiomic features in different and larger databases can further increase the final results.

Prediction of pCR with pretreatment MRI radiomics in triple negative breast cancer treated with neoadjuvant chemo-immunotherapy.

589 Background: Neoadjuvant chemo-immunotherapy (NACI) has become the new standard of care for stage II-III triple-negative breast cancers (TNBC). This approach is linked to high rates of pathological complete responses (pCR), yet factors influencing responses to this regimen remain poorly characterized. Our objective was to investigate whether pretreatment MRI perfusion-based radiomics could serve as a predictive tool for pCR to NACI in patients with TNBC. Methods: This prospective study enrolled women with early-stage TNBC undergoing NACI at two different centers from August 2021 to July 2023. Tumor segmentation and pharmacokinetic analyses of pretreatment multi-vendor dynamic-contrast-enhanced (DCE)-MRI were conducted using the standard perfusion Tofts model. Radiomics features were extracted from the Ktrans, Ve, and Slopemax maps. Unsupervised correlation analysis of reproducible features and least absolute shrinkage and selector operation were employed to construct a radiomics score. The performance of the radiomic score was evaluated using the area under the receiver operating characteristic curve (AUC). A multivariable logistic regression model was employed to establish a multimodal nomogram. Performance metrics included discrimination, calibration, and clinical usefulness, assessed on the external validation set. Results: The training set comprised 112 patients from center 1 (mean age, 52 years ± 11 [SD]), and the external validation set included 83 participants from center 2 (mean age, 49 years ± 11 [SD]). Overall, the pCR rate was 70% (137/195). In the validation set, the radiomic score achieved an AUC of 0.80 (95% CI: 0.70, 0.89) for predicting pCR. A nomogram incorporating the radiomics score, tumor grade, and Ki67 yielded an AUC of 0.86 (95% CI: 0.78, 0.94) and demonstrated superior predictive efficacy compared to the histological prediction model (AUC, 0.76 [95% CI: 0.66, 0.87]; P = .02) in the validation set. Calibration curves of the nomogram exhibited good concordance between predicted and observed pCR in both sets. The radiomic score correlated with tumor size, washout enhancement curve, androgen receptor (AR), and combined positive score (CPS) PD-L1 levels (all P < 0.05). Conclusions: PretreatmentMRI-based radiomics is a reliable predictor of pCR to NACI and might help customizing treatment choice for early-stage TNBC patients.

Automated segmentation of liver tumors from computed tomographic scans

The precision of liver tumor segmentation heavily depends on the doctor's expertise, hence it is required to produce an algorithm for automatic liver tumor segmentation to reduce the manual intervention in assessing liver disease identification. We propose a CNN-based UNet architecture designed to segment liver tumors from CT images of size 128×128. In this model, modifications were made to the encoder, decoder, and bridge paths to enhance feature extraction efficiency. The performance of the modified UNet was evaluated against an existing segmentation method using the same CT image size. The comparison focused on the Dice similarity coefficient and accuracy. Our proposed method demonstrated a high Dice similarity coefficient of 75.37 % and an accuracy of 99.75 % on the 3Dircadb dataset. These results indicate that our modified UNet achieved superior segmentation metrics compared to state-of-the-art methods, showcasing its effectiveness in liver tumor segmentation.

A Two-Stage Generative Model with CycleGAN and Joint Diffusion for MRI-based Brain Tumor Detection.

Accuratedetection and segmentation of brain tumors is critical for medical diagnosis. However, current supervised learning methods require extensively annotated images and the state-of-the-art generative models used in unsupervised methods often have limitations in covering the whole data distribution. In this paper, we propose a novel framework Two-Stage Generative Model (TSGM) that combines Cycle Generative Adversarial Network (CycleGAN) and Variance Exploding stochastic differential equation using joint probability (VE-JP) to improve brain tumor detection and segmentation. The CycleGAN is trained on unpaired data to generate abnormal images from healthy images as data prior. Then VE-JP is implemented to reconstruct healthy images using synthetic paired abnormal images as a guide, which alters only pathological regions but not regions of healthy. Notably, our method directly learned the joint probability distribution for conditional generation. The residual between input and reconstructed images suggests the abnormalities and a thresholding method is subsequently applied to obtain segmentation results. Furthermore, the multimodal results are weighted with different weights to improve the segmentation accuracy further. We validated our method on three datasets, and compared with other unsupervised methods for anomaly detection and segmentation. The DSC score of 0.8590 in BraTs2020 dataset, 0.6226 in ITCS dataset and 0.7403 in In-house dataset show that our method achieves better segmentation performance and has better generalization.

Use of artificial intelligence–based digital pathology to predict outcomes for immune checkpoint inhibitor therapy in advanced gastro-esophageal cancer.

4013 Background: Accurate prediction of response and survival outcomes with anti-PD-1/PD-L1 immune checkpoint inhibition (ICI) remains a significant challenge in gastro-esophageal cancers. In this study, we use an artificial intelligence (AI)-based single-cell analysis of digitized whole-slide H&E images (WSIs) to predict objective response and survival benefit of ICI in two independent cohorts of gastro-esophageal cancer patients. Methods: WSIs were obtained from 82 ICI-treated advanced gastroesophageal cancer patients (gastric, esophageal, and gastro-esophageal junction adenocarcinoma and squamous cell carcinoma) at Stanford University for discovery, and 189 ICI-treated advanced gastric adenocarcinoma patients at Southern Medical University for external validation. Deep learning models were deployed for automated tumor area detection and segmentation of cell nuclei within WSIs. We developed a fully automated cell annotation approach by leveraging multiplex immunofluorescence and trained a deep learning model to classify nuclei into four cell types from H&E (tumor cells, lymphocytes, neutrophils, and macrophages). A total of 66 features were computed to quantify cell composition and cell-cell interactions within the tumor microenvironment. Treatment outcomes were assessed using progression-free survival (PFS) and best objective response per the Response Evaluation Criteria in Solid Tumors (v1.1), with statistical significance reported at the 95% confidence level. Results: In the training cohort, a spatial feature characterizing lymphocyte and neutrophil interaction had the strongest association with PFS (hazard ratio = 0.44, 95% CI 0.24-0.79, P = 0.0046). In the validation cohort, the spatial biomarker positive population had significantly improved PFS compared to the spatial biomarker negative population (hazard ratio = 0.41, 95% CI 0.27-0.61, P < 0.0001; median time to event: 19 months vs 9 months). For predicting objective response in the validation cohort, a multivariate model combining the spatial features achieved AUROC = 0.81 compared to AUROC = 0.65 for PD-L1 CPS (P = 0.0014), while combining the multivariate model with PD-L1 CPS achieved AUROC = 0.84. Conclusions: A single-cell computational pathology approachidentifies spatial biomarkers with predictive utility for determining ICI treatment outcomes in advanced gastro-esophageal cancer. Further validation of these findings is being pursued in additional cohorts.

FCM and CBAC based Brain Tumor Identification and Segmentation

A brain tumor are an abnormal growth of cells within the brain, forming a mass that can be either cancerous (malignant) or non-cancerous (benign). Despite their differences, both types of tumors can pose serious health risks. As these tumors grow, they can increase intracranial pressure, leading to potential brain damage. This increased pressure can result in various symptoms such as headaches, seizures, vision problems, and changes in cognitive function. The potential for life-threatening consequences makes early detection and treatment crucial. The objective of the research is to develop a system or algorithm capable of accurately identifying the presence of brain tumors within medical imaging data (CT or MRI scans) and subsequently segmenting the tumor regions from the surrounding healthy brain tissue. This research aims at building an automated multi stage reliable system for classifying MRI images as tumor or non-tumor images. However, the research aims to diagnose brain tumor by extracting the tumor region accurately. The main contribution of this work is to automatically segment the tumor region from the MRI brain images, using Fuzzy C-Means (FCM) Clustering and the Content-Based Active Contour (CBAC) method. The CBAC method helps to resolve the issues of saddle points and broken edges in the extracted tumor region.

SurvNet: A low-complexity convolutional neural network for survival time classification of patients with glioblastoma

Background and objectiveMalignant primary brain tumors cause the greatest number of years of life lost than any other cancer. Grade 4 glioma is particularly devastating: The median survival without any treatment is less than six months and with standard-of-care treatment is only 14.6 months. Accurate identification of the overall survival time of patients with brain tumors is of profound importance in many clinical applications. Automated image analytics with magnetic resonance imaging (MRI) can provide insights into the prognosis of patients with brain tumors. MethodsIn this paper, We propose SurvNet, a low-complexity deep learning architecture based on the convolutional neural network to classify the overall survival time of patients with brain tumors into long-time and short-time survival cohorts. Through the incorporation of diverse MRI modalities as inputs, we facilitate deep feature extraction at various anatomical sites, thereby augmenting the precision of predictive modeling. We compare SurvNet with the Inception V3, VGG 16 and ensemble CNN models on pre-operative magnetic resonance image datasets. We also analyzed the effect of segmented brain tumors and training data on the system performance. ResultsSeveral measures, such as accuracy, precision, and recall, are calculated to examine the perfor-mance of SurvNet on three-fold cross-validation. SurvNet with T1 MRI modality achieved a 62.7 % accuracy, compared with 52.9 % accuracy of the Inception V3 model, 58.5 % accuracy of the VGG 16 model, and 54.9 % of the ensemble CNN model. By increasing the MRI input modalities, SurvNet becomes more accurate and achieves 76.5 % accuracy with four MRI modalities. Combining the segmented data, SurvNet achieved the highest accuracy of 82.4 %. ConclusionsThe research results show that SurvNet achieves higher metrics such as accuracy and f1-score than the comparisons. Our research also proves that by using multiparametric MRI modalities, SurvNet is able to learn more image features and performs a better classification accuracy. We can conclude that SurvNet with the complete scenario, i.e., segmented data and four MRI modalities, achieved the best accuracy, showing the validity of segmentation information during the survival time prediction process.

Redefining brain tumor segmentation: a cutting-edge convolutional neural networks-transfer learning approach

Medical image analysis has witnessed significant advancements with deep learning techniques. In the domain of brain tumor segmentation, the ability to precisely delineate tumor boundaries from magnetic resonance imaging (MRI) scans holds profound implications for diagnosis. This study presents an ensemble convolutional neural network (CNN) with transfer learning, integrating the state-of-the-art Deeplabv3+ architecture with the ResNet18 backbone. The model is rigorously trained and evaluated, exhibiting remarkable performance metrics, including an impressive global accuracy of 99.286%, a high-class accuracy of 82.191%, a mean intersection over union (IoU) of 79.900%, a weighted IoU of 98.620%, and a Boundary F1 (BF) score of 83.303%. Notably, a detailed comparative analysis with existing methods showcases the superiority of our proposed model. These findings underscore the model’s competence in precise brain tumor localization, underscoring its potential to revolutionize medical image analysis and enhance healthcare outcomes. This research paves the way for future exploration and optimization of advanced CNN models in medical imaging, emphasizing addressing false positives and resource efficiency.

GAIR-U-Net: 3D guided attention inception residual u-net for brain tumor segmentation using multimodal MRI images

Deep learning technologies have led to substantial breakthroughs in the field of biomedical image analysis. Accurate brain tumor segmentation is an essential aspect of treatment planning. Radiologists agree that manual segmentation is a difficult and time-consuming task that frequently causes delays in the diagnosing process. While U-Net-based methods have been widely used for brain tumor segmentation, many challenges persist, particularly when dealing with tumors of varying sizes, locations, and shapes. Additionally, segmenting tumor regions with structures requires a comprehensive model, which can increase computational complexity and potentially cause gradient vanishing issues. This study presents a novel method called 3D Guided Attention-based deep Inception Residual U-Net (GAIR-U-Net) to address these challenges. This model combines attention mechanisms, an inception module, and residual blocks with dilated convolution to enhance feature representation and spatial context understanding. The backbone of the model is the U-Net model, which leverages the power of inception and residual connections to capture intricate patterns and hierarchical features while expanding the model’s width in three-dimensional space without significantly increasing computational complexity. The attention mechanisms play a role in focusing on important regions and areas while downgrading irrelevant details. The dilated convolutions in the network help in learning both local and global information, improving accuracy and adaptability in segmenting tumors. All the experiments in this study were carried out on multimodal MRI scans that include (T1-weighted, T1-ce, T2-weighted, and FLAIR sequences) from the BraTS 2020 dataset. The presented model is trained and tested on the same dataset, which exhibited promising performance compared to previous methods. On the BraTS 2020 validation dataset, the proposed model obtained a dice score of 0.8796, 0.8634, and 0.8441 for whole tumor (WT), tumor core (TC), and enhancing tumor (ET), respectively. These results demonstrate the model’s efficacy in precisely segmenting brain tumors across various modalities. Comparative analyses underscore the model’s versatility in handling tumor shape variations, size, and location, making it a promising solution for clinical applications.

Overall survival prediction and risk stratification in patients with neuroblastoma through machine learning in the large multi-institutional PRIMAGE cohort.

10054 Background: Neuroblastoma (NB) is the most prevalent solid cancer in childhood in which imaging plays a pivotal role at every step of the patient's journey. This investigation aimed to develop a machine learning model using clinical, molecular (MycN amplification), and magnetic resonance (MR) radiomics features at diagnosis to predict patient’s overall survival (OS) and improve their risk stratification. Methods: A database comprising clinical, molecular, and International Neuroblastoma Risk Group (INRG) staging on 513 patients with accessible MR imaging (discovery cohort) was employed for model training, validation, and testing. An additional 22 patients from non-discovery cohort hospitals served as an external validation group. Tumor segmentations of NB were manually and semi-automatically performed on corresponding T2-weighted MR images by an experienced radiologist. In total, 107 radiomics features were extracted and harmonized across MR scan manufacturers and magnetic field strengths using the nested ComBat methodology to correct both batch effects. These radiomics features, combined with clinical and molecular data, were utilized as input for the models. A nested cross-validation approach was employed for model development to determine the optimal preprocessing, machine learning algorithm, and model configuration. Results: The discovery cohort yielded a C-index of 0.788 ± 0.049 in the test partitions, with a random survival forest (RSF) exhibiting the best performance. In the validation cohort, a C-index of 0.934 was achieved. Interpretability analysis identified lesion heterogeneity, size, and molecular variables as crucial factors in OS prediction. The model demonstrated superior predictive performance and patient stratification compared to conventional staging systems in both cohorts. Conclusions: The RSF predictive model exhibited high performance, emphasizing the significant contribution of radiomics features and alignment with established clinical and molecular variables. The model stratified NB patients into low-, intermediate-, and high-risk categories and suggests that radiomics features potentially could improve current risk stratification systems. Additional external validation is warranted as this may present new evidence for enhancing patient care and clinical decision-making for NB patients.