Enhancing Tumor Regions Research Articles

NIMG-60. SUBTYPING GLIOBLASTOMA FROM JOINT CLUSTERING OF IMAGING AND GENOMIC DATA

Abstract PURPOSE To comprehensively unravel the heterogeneity of glioblastoma, we developed a subtyping method using integrated imaging and genomic data. METHODS We assembled a retrospective cohort of 571 IDH-wildtype glioblastoma patients, obtaining pre-operative multi-parametric MRI (T1, T1-GD, T2, T2-FLAIR, DSC, DTI) scans (available for 462 patients) and targeted next-generation sequencing (NGS) data (available for 355 patients). We extracted 971 radiomic features from these MRI scans, selecting 12 significant dimensions through L21-norm minimization guided by 13 key driver genes from the five most frequently altered pathways in glioblastoma compared to healthy controls (RB1, P53, MAPK, PI3K, and RTK). Subtypes were identified using a joint clustering approach that integrated radiomics and genomics. RESULTS Our method identified three glioblastoma subtypes with distinct risk levels: high-risk (subtype 1), medium-risk (subtype 2), and low-risk (subtype 3), each characterized by their differential overall survival outcomes (Kaplan-Meier analysis, p < 0.05). The intensities of axial diffusivity in enhancing tumor regions and radial diffusivity in non-enhancing cores increased across three subtypes. All subtypes shared a common tendency for more frequent mutation co-occurrence pattern in [TP53,RB1] but exhibited unique molecular characteristics. For example, subtype 1 displayed a relatively high frequency of co-occurring mutations in [TP53,KDR] and [NOTCH2,MDM4]; subtype 2 had six co-occurring pairs [TP53,KDR], [PTPN11,NF1], [RB1,FUBP1], [PTEN,CIC], [PTEN,KRAS], and [NOTCH2,CDKN2A]; subtype 3 showed six co-occurring pairs [PTPN11,NF1], [PDGFRA,NF1], [NTRK1,NOTCH2], [MET,ATRX], [KRAS,KDR], and [FGFR2,DNMT3A] plus four significant patterns of mutual exclusivity in [TP53,PTPN11], [TP53,EGFR], [EGFR,RB1], and [EGFR,NF1]. CONCLUSION Our study discovered distinct glioblastoma subtypes, revealing complex patterns and structures across multiple data modalities without relying on prior clinical assumptions. This is promising to help enhance the precision of diagnosis and treatment of glioblastoma.

NIMG-79. A RADIOMIC-RISK SCORE FOR SURVIVAL PREDICTION IN GLIOMA: A LARGE MULTI-INSTITUTIONAL RETROSPECTIVE VALIDATION STUDY

Abstract Gliomas have dismal survival. Recently, radiomics has demonstrated success in developing non-invasive image-based biomarkers for survival prediction in gliomas. However, clinical applicability of radiomics-based biomarkers will ultimately require rigorous validation on large, multi-institutional data. This work attempts to evaluate a radiomic-based survival risk-assessment prognostic score on a multi-institutional cohort of Glioma patients. Our rationale is that optimizing and validating radiomic features that capture intensity-based textural variations and morphology-based changes (e.g., local curvature and global contour) on large multi-institutional cohorts can allow for robust risk assessment in Glioma. n=668 pre-operative T1w MRI scans for Glioma patients from 4 cohorts: University of California San Francisco “UCSF” (n=495), LUMIERE (n=86)), xCures (n=13), and Cleveland Clinic (CCF) (n=74) were analyzed. Subjects from UCSF, CCF, and xCures were used for training, whereas the LUMIERE cohort was used for hold-out testing. Following pre-processing, expert-vetted tumor segmentation into Edema (ED) and Enhancing tumor (ET) regions was performed. A total of n=1558 radiomic features were extracted from the tumor regions (779 per region), namely, 31 global contour, 20 curvature-based, 624 gradient co-occurrence statistics, and 104 Collage features. Cox regression model was employed to create a radiomic risk score (RRS) for every subject, which included radiomics features as well as on clinical and demographic variables (IDH, MGMT, sex, age), for survival analysis. RRS incorporated a total of 43 radiomic features, and the associated clinical and demographic variables (CDS), yielding significant differences between high and low risk groups in the training (RRS: p=0.043, HR = 2, CSD: p=0.0031, HR = 2.8) and testing sets (RRS: p=0.0096, HR = 1.8, CSD: p=0.0071, HR = 1.8). Combining RRS and CDS improved the OS prediction at group level (p=3.6e-6, 0.0011, HR=5.4, 2, for training and testing, respectively). Radiomics-based risk-stratification may be promising for reliable risk-stratification in glioma.

NIMG-55. DEVELOPING A COMPUTATIONAL PIPELINE TO FORECAST AND SPATIALLY MAP HIGH-GRADE GLIOMA RESPONSE TO RADIOTHERAPY USING BIOPHYSICAL MODELS AND DATA-DRIVEN STATISTICAL ANALYSIS

Abstract High-grade glioma (HGG) presents significant treatment challenges, particularly in identifying less responsive tumor areas to target with adaptive radiotherapy (RT). This study introduces a patient-specific computational pipeline designed to predict and spatially map HGG response to RT, focusing on regions demonstrating resilience to RT. Our computational pipeline is centered around a family of 3D biophysical models describing tumor cell invasion, proliferation, and response to RT which can be personalized for individual patients using longitudinal multiparametric MRI data. Utilizing longitudinal multiparametric MRI data reporting on cell density (diffusion weighted MRI) and the extent of disease (T2-FLAIR and contrast-enhanced T1-weighted MRIs) from 21 patients, ranging from baseline to three weeks into RT, we calibrated a family of biophysical models to forecast tumor cell density up to the final week of RT and one-month follow-up. By employing pareto tail analysis on empirical cumulative distribution function fits of voxel-wise changes in apparent diffusion coefficient maps (ΔADC) between the future images (end of RT, one-month follow-up) and pre-treatment images, we identified areas of increased tumor cellularity on both the ground truth and model predicted cellularity maps. The analysis differentiated tumor voxels into “non-responding” and “responding” categories, based on cutoffs determined through ROC analysis across various conditions—including assessments of both enhancing and non-enhancing tumor regions at different time points. Optimal ΔADC cutoffs for predicted and measured data were established, leading to high area under curve (AUC; ranging from 0.803 to 0.934) values and significant Youden’s index scores (ranging from 0.526 to 0.765) across all conditions. Our findings demonstrate the feasibility of this novel pipeline in delineating biologically relevant RT targets, offering a potential tool for enhancing the efficacy of adaptive RT strategies in treating HGG.

Estimation of Fractal Dimension and Segmentation of Brain Tumor with Parallel Features Aggregation Network

The accurate recognition of a brain tumor (BT) is crucial for accurate diagnosis, intervention planning, and the evaluation of post-intervention outcomes. Conventional methods of manually identifying and delineating BTs are inefficient, prone to error, and time-consuming. Subjective methods for BT recognition are biased because of the diffuse and irregular nature of BTs, along with varying enhancement patterns and the coexistence of different tumor components. Hence, the development of an automated diagnostic system for BTs is vital for mitigating subjective bias and achieving speedy and effective BT segmentation. Recently developed deep learning (DL)-based methods have replaced subjective methods; however, these DL-based methods still have a low performance, showing room for improvement, and are limited to heterogeneous dataset analysis. Herein, we propose a DL-based parallel features aggregation network (PFA-Net) for the robust segmentation of three different regions in a BT scan, and we perform a heterogeneous dataset analysis to validate its generality. The parallel features aggregation (PFA) module exploits the local radiomic contextual spatial features of BTs at low, intermediate, and high levels for different types of tumors and aggregates them in a parallel fashion. To enhance the diagnostic capabilities of the proposed segmentation framework, we introduced the fractal dimension estimation into our system, seamlessly combined as an end-to-end task to gain insights into the complexity and irregularity of structures, thereby characterizing the intricate morphology of BTs. The proposed PFA-Net achieves the Dice scores (DSs) of 87.54%, 93.42%, and 91.02%, for the enhancing tumor region, whole tumor region, and tumor core region, respectively, with the multimodal brain tumor segmentation (BraTS)-2020 open database, surpassing the performance of existing state-of-the-art methods. Additionally, PFA-Net is validated with another open database of brain tumor progression and achieves a DS of 64.58% for heterogeneous dataset analysis, surpassing the performance of existing state-of-the-art methods.

Single nucleus transcriptomics, pharmacokinetics, and pharmacodynamics of combined CDK4/6 and mTOR inhibition in a phase 0/1 trial of recurrent high-grade glioma.

Outcomes for adult patients with a high-grade glioma continue to be dismal and new treatment paradigms are urgently needed. To optimize the opportunity for discovery, we performed a phase 0/1 dose-escalation clinical trial that investigated tumor pharmacokinetics, pharmacodynamics, and single nucleus transcriptomics following combined ribociclib (CDK4/6 inhibitor) and everolimus (mTOR inhibitor) treatment in recurrent high-grade glioma. Patients with a recurrent high-grade glioma (n = 24) harboring 1) CDKN2A / B deletion or CDK4 / 6 amplification, 2) PTEN loss or PIK3CA mutations, and 3) wild-type retinoblastoma protein (Rb) were enrolled. Patients received neoadjuvant ribociclib and everolimus treatment and no dose-limiting toxicities were observed. The median unbound ribociclib concentrations in Gadolinium non-enhancing tumor regions were 170 nM (range, 65 - 1770 nM) and 634 nM (range, 68 - 2345 nM) in patients receiving 5 days treatment at the daily dose of 400 and 600 mg, respectively. Unbound everolimus concentrations were below the limit of detection (< 0.1 nM) in both enhancing and non-enhancing tumor regions at all dose levels. We identified a significant decrease in MIB1 positive cells suggesting ribociclib-associated cell cycle inhibition. Single nuclei RNAseq (snRNA) based comparisons of 17 IDH-wild-type on-trial recurrences to 31 IDH-wild-type standard of care treated recurrences data demonstrated a significantly lower fraction of cycling and neural progenitor-like (NPC-like) malignant cell populations. We validated the CDK4/6 inhibitor-directed malignant cell state shifts using three patient-derived cell lines. The presented clinical trial highlights the value of integrating pharmacokinetics, pharmacodynamics, and single nucleus transcriptomics to assess treatment effects in phase 0/1 surgical tissues, including malignant cell state shifts. ClinicalTrials.gov identifier: NCT03834740 .

DSC-PWI presurgical differentiation of grade 4 astrocytoma and glioblastoma in young adults: rCBV percentile analysis across enhancing and non-enhancing regions

PurposeThe presurgical discrimination of IDH-mutant astrocytoma grade 4 from IDH-wildtype glioblastoma is crucial for patient management, especially in younger adults, aiding in prognostic assessment, guiding molecular diagnostics and surgical planning, and identifying candidates for IDH-targeted trials. Despite its potential, the full capabilities of DSC-PWI remain underexplored. This research evaluates the differentiation ability of relative-cerebral-blood-volume (rCBV) percentile values for the enhancing and non-enhancing tumor regions compared to the more commonly used mean or maximum preselected rCBV values.MethodsThis retrospective study, spanning 2016–2023, included patients under 55 years (age threshold based on World Health Organization recommendations) with grade 4 astrocytic tumors and known IDH status, who underwent presurgical MR with DSC-PWI. Enhancing and non-enhancing regions were 3D-segmented to calculate voxel-level rCBV, deriving mean, maximum, and percentile values. Statistical analyses were conducted using the Mann-Whitney U test and AUC-ROC.ResultsThe cohort consisted of 59 patients (mean age 46; 34 male): 11 astrocytoma-4 and 48 glioblastoma. While glioblastoma showed higher rCBV in enhancing regions, the differences were not significant. However, non-enhancing astrocytoma-4 regions displayed notably higher rCBV, particularly in lower percentiles. The 30th rCBV percentile for non-enhancing regions was 0.705 in astrocytoma-4, compared to 0.458 in glioblastoma (p = 0.001, AUC-ROC = 0.811), outperforming standard mean and maximum values.ConclusionEmploying an automated percentile-based approach for rCBV selection enhances differentiation capabilities, with non-enhancing regions providing more insightful data. Elevated rCBV in lower percentiles of non-enhancing astrocytoma-4 is the most distinguishable characteristic and may indicate lowly vascularized infiltrated edema, contrasting with glioblastoma’s pure edema.

Estrogen α and β Receptor Expression in the Various Regions of Resected Glioblastoma Multiforme Tumors and in an In Vitro Model.

Glioblastoma multiforme (GBM) is a malignant tumor with a higher prevalence in men and a higher survival rate in transmenopausal women. It exhibits distinct areas influenced by changing environmental conditions. This study examines how these areas differ in the levels of estrogen receptors (ERs) which play an important role in the development and progression of many cancers, and whose expression levels are often correlated with patient survival. This study utilized two research models: an in vitro model employing the U87 cell line and a second model involving tumors resected from patients (including tumor core, enhancing tumor region, and peritumoral area). ER expression was assessed at both gene and protein levels, with the results validated using confocal microscopy and immunohistochemistry. Under hypoxic conditions, the U87 line displayed a decrease in ERβ mRNA expression and an increase in ERα mRNA expression. In patient samples, ERβ mRNA expression was lower in the tumor core compared to the enhancing tumor region (only in males when the study group was divided by sex). In addition, ERβ protein expression was lower in the tumor core than in the peritumoral area (only in women when the study group was divided by sex). Immunohistochemical analysis indicated the highest ERβ protein expression in the enhancing tumor area, followed by the peritumoral area, and the lowest in the tumor core. The findings suggest that ER expression may significantly influence the development of GBM, exhibiting variability under the influence of conditions present in different tumor areas.

NIMG-45. EXTERNAL EVALUATION OF A MACHINE LEARNING MODEL EMPLOYING RADIOMICS TO IDENTIFY REGIONS OF LOCAL RECURRENCE IN GLIOBLASTOMA FROM POSTOPERATIVE MRI

Abstract The standard surgical goal for operable glioblastomas worldwide is the complete excision of the enhancing tumor. Nevertheless, the peritumoral region harbors infiltrating cells leading to recurrence. Our study aims to evaluate a predictive model designed to predict potential regions of recurrence using voxel-based radiomics analysis on postoperative magnetic resonance imaging (MRI). This retrospective, multi-institutional study involved glioblastoma patients who had undergone complete resection. The training cohort comprised 49 patients from two Spanish institutions, with model evaluation using an external dataset of 27 patients from a Norwegian institution and the public Ivy-Gap dataset. We used follow-up MRI scans for ground truth definition of recurrence, while postoperative multiparametric MRI scans provided the data for extracting voxel-based radiomic features. Our method employed an unsupervised, hybrid approach, which combined a deep neural network and instance optimization, to align the MRI sequences for each patient, registering the follow-up with the postoperative scans. To generate ground truth labels, we segmented the peritumoral region in the postoperative scans, identified as the area showing T2/FLAIR signal alterations, in addition to the enhancing tumor visible in the registered follow-up scan. Overlapping voxels between the peritumoral and enhancing tumor regions were classified as recurrence, while non-overlapping voxels were labeled as nonrecurrence. Radiomic features were subsequently extracted from the postoperative peritumoral region. We trained four machine learning classifiers to predict recurrence. Among these, the Categorical Boosting (CatBoost) classifier demonstrated excellent performance on the test dataset, achieving an average area under the curve (AUC) of 0.83 ± 0.07, and an accuracy of 0.80 ± 0.12, based on voxel-wise comparison with the ground truth. Our study resulted in a method that accurately predicts the area of future tumor recurrence in glioblastoma patients' MRI scans. This innovation could potentially tailor surgical and radiotherapy treatment strategies to these specific areas, possibly extending patient survival.

Attention-guided multi-scale context aggregation network for multi-modal brain glioma segmentation.

Accurate segmentation of brain glioma is a critical prerequisite for clinical diagnosis, surgical planning and treatment evaluation. In current clinical workflow, physicians typically perform delineation of brain tumor subregions slice-by-slice, which is more susceptible to variabilities in raters and also time-consuming. Besides, even though convolutional neural networks (CNNs) are driving progress, the performance of standard models still have some room for furtherimprovement. To deal with these issues, this paper proposes an attention-guided multi-scale context aggregation network (AMCA-Net) for the accurate segmentation of brain glioma in the magnetic resonance imaging (MRI) images withmulti-modalities. AMCA-Net extracts the multi-scale features from the MRI images and fuses the extracted discriminative features via a self-attention mechanism for brain glioma segmentation. The extraction is performed via a series of down-sampling, convolution layers, and the global context information guidance (GCIG) modules are developed to fuse the features extracted for contextual features. At the end of the down-sampling, a multi-scale fusion (MSF) module is designed to exploit and combine all the extracted multi-scale features. Each of the GCIG and MSF modules contain a channel attention (CA) module that can adaptively calibrate feature responses and emphasize the most relevant features. Finally, multiple predictions with different resolutions are fused through different weightings given by a multi-resolution adaptation (MRA) module instead of the use of averaging or max-pooling to improve the final segmentationresults. Datasets used in this paper are publicly accessible, that is, the Multimodal Brain Tumor Segmentation Challenges 2018 (BraTS2018) and 2019 (BraTS2019). BraTS2018 contains 285 patient cases and BraTS2019 contains 335 cases. Simulations show that the AMCA-Net has better or comparable performance against that of the other state-of-the-art models. In terms of the Dice score and Hausdorff 95 for the BraTS2018 dataset, 90.4% and 10.2 mm for the whole tumor region (WT), 83.9% and 7.4 mm for the tumor core region (TC), 80.2% and 4.3 mm for the enhancing tumor region (ET), whereas the Dice score and Hausdorff 95 for the BraTS2019 dataset, 91.0% and 10.7 mm for the WT, 84.2% and 8.4 mm for the TC, 80.1% and 4.8 mm for the ET. The proposed AMCA-Net performs comparably well in comparison to several state-of-the-art neural net models in identifying the areas involving the peritumoral edema, enhancing tumor, and necrotic and non-enhancing tumor core of brain glioma, which has great potential for clinical practice. In future research, we will further explore the feasibility of applying AMCA-Net to other similar segmentationtasks.

Abstract LB067: A confident and operator-independent deep segmentation model to measure residual tumor volume in the follow-up MRIs for glioblastoma

Abstract Introduction: Magnetic resonance imaging (MRI) is the most common tool to examine glioblastoma. Preoperative MRI can be used to initial diagnosis, and surgery planning, while follow-up MRIs can be used to evaluate treatment responses, identify recurrency, and detect side effects. The follow-up MRIs are usually taken after the first-line therapy, such as maximal safe resection. In the past, radiologists manually segment the tumor regions from normal brain tissue on follow-up MRIs, which is time-consuming, error-prone, and challenging. Several deep-learning (DL) models have been developed utilizing preoperative images, but their performance has yet to be evaluated on follow-up MRIs. In this research, we built the largest follow-up MRI cohort (311 patients) to assess these DL models and their generalizability and performance on independent preoperative and follow-up images. We also made the first follow-up-based deep learning models for this specific task. Methods: All evaluation deep learning models (10 models) were trained by the Brain Tumor Segmentation challenge 2020 (BraTS’20) and evaluated by fifty pairs of preoperative and follow-up scans from our institution. The segmentation form our institution is evaluated by board certified radiologist. MRIs in the BraTS’20 dataset were all preoperative scans. After the evaluation, we randomly assigned 264 patients’ scans from our institution to the training dataset and 47 patients’ scans to the testing dataset. We compared three types of models in our follow-up deep learning model, including 1) UNet-3D, 2) UNet-3D+transfer-learning, and 3) UNet-3D+transfer-learning+baysian-learning. The benchmark for all models was the Dice similarity coefficient (DSC). DSC can measure the spatial overlap between model prediction and ground truth. The value of DSC is between zero to one. Zero means no overlap and one indicates complete overlap. Results: Our study demonstrates that the BraTS'20 trained models' performance decreased by 13.05% in independent preoperative MRI scans and 19.04% in follow-up MRI scans. The most significant mismatch regions were FLAIR hyperintense regions (3.68% drop in independent preoperative scans and 10.61% drop in independent follow-up scans) and Non-enhancing core (5.20% drop in independent preoperative scans and 11.99% drop in independent follow-up scans). Our best model can achieve the best DSC among three tumor regions compared to all evaluation models (FLAIR hyperintense regions: DSC 0.77 V.S. DSC 0.58; Enhancing tumor region: DSC 0.87 V.S. DSC 0.68; Non-enhancing tumor region: DSC 0.92 V.S. DSC 0.55). Conclusion: Maximal safe resection induced brain structure change, decreasing the performance of the preoperative-based DL model. Implementing a follow-up MRI-based segmentation model is essential to make accurate and generalizable results to address structural changes after maximal safe resection. Our follow-up DL model demonstrates the DSC score can be recovered. We commit to further developing the tool to assist radiologists in handling follow-up MRIs for glioblastoma patients. Citation Format: Kang Lin Hsieh, Tanjida Kabir, Luis Nunez-Rubiano, Yu-Chun Hsu, Yu Cai, Juan Rodriguez Quintero, Octavio Arevalo, Kangyi Zhao, Jackie Zhang, Jiguang Zhu Zhu, Roy Riascos, Xioaqian Jiang, Shayan Shams. A confident and operator-independent deep segmentation model to measure residual tumor volume in the follow-up MRIs for glioblastoma [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 2 (Clinical Trials and Late-Breaking Research); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(8_Suppl):Abstract nr LB067.

Selective Deeply Supervised Multi-Scale Attention Network for Brain Tumor Segmentation.

Brain tumors are among the deadliest forms of cancer, characterized by abnormal proliferation of brain cells. While early identification of brain tumors can greatly aid in their therapy, the process of manual segmentation performed by expert doctors, which is often time-consuming, tedious, and prone to human error, can act as a bottleneck in the diagnostic process. This motivates the development of automated algorithms for brain tumor segmentation. However, accurately segmenting the enhanced and core tumor regions is complicated due to high levels of inter- and intra-tumor heterogeneity in terms of texture, morphology, and shape. This study proposes a fully automatic method called the selective deeply supervised multi-scale attention network (SDS-MSA-Net) for segmenting brain tumor regions using a multi-scale attention network with novel selective deep supervision (SDS) mechanisms for training. The method utilizes a 3D input composed of five consecutive slices, in addition to a 2D slice, to maintain sequential information. The proposed multi-scale architecture includes two encoding units to extract meaningful global and local features from the 3D and 2D inputs, respectively. These coarse features are then passed through attention units to filter out redundant information by assigning lower weights. The refined features are fed into a decoder block, which upscales the features at various levels while learning patterns relevant to all tumor regions. The SDS block is introduced to immediately upscale features from intermediate layers of the decoder, with the aim of producing segmentations of the whole, enhanced, and core tumor regions. The proposed framework was evaluated on the BraTS2020 dataset and showed improved performance in brain tumor region segmentation, particularly in the segmentation of the core and enhancing tumor regions, demonstrating the effectiveness of the proposed approach. Our code is publicly available.

Androgen Receptor Expression in the Various Regions of Resected Glioblastoma Multiforme Tumors and in an In Vitro Model.

Glioblastoma multiforme (GBM) is a malignant glioma, difficult to detect and with the lowest survival rates among gliomas. Its greater incidence among men and its higher survival rate among premenopausal women suggest that it may be associated with the levels of androgens. As androgens stimulate the androgen receptor (AR), which acts as a transcription factor, the aim of this study was the investigate the role of AR in the progression of GBM. The study was conducted on tissues collected from three regions of GBM tumors (tumor core, enhancing tumor region, and peritumoral area). In addition, an in vitro experiment was conducted on U-87 cells under various culture conditions (necrotic, hypoxic, and nutrient-deficient), mimicking the conditions in a tumor. In both of the models, androgen receptor expression was determined at the gene and protein levels, and the results were confirmed by confocal microscopy and immunohistochemistry. AR mRNA expression was higher under nutrient-deficient conditions and lower under hypoxic conditions in vitro. However, there were no differences in AR protein expression. No differences in AR mRNA expression were observed between the tested tumor structures taken from patients. No differences in AR mRNA expression were observed between the men and women. However, AR protein expression in tumors resected from patients was higher in the enhancing tumor region and in the peritumoral area than in the tumor core. In women, higher AR expression was observed in the peritumoral area than in the tumor core. AR expression in GBM tumors did not differ significantly between men and women, which suggests that the higher incidence of GBM in men is not associated with AR expression. In the group consisting of men and women, AR expression varied between the regions of the tumor: AR expression was higher in the enhancing tumor region and in the peritumoral area than in the tumor core, showing a dependence on tumor conditions (hypoxia and insufficient nutrient supply).

MLRD-Net: 3D multiscale local cross-channel residual denoising network for MRI-based brain tumor segmentation

The precise segmentation of multimodal MRI images is the primary stage of tumor diagnosis and treatment. Current segmentation strategies often underutilize multiscale features, which can easily lead to loss of contextual information, reduction of low-level features and noise interference. To overcome these issues, a 3D multiscale local cross-channel residual denoising network (MLRD-Net) for an MRI-based brain tumor segmentation algorithm is proposed in this paper. Specifically, we employ encoder-decoder structure to connect local and global features, and enhance the receptive field of the network. Random slice operation has been conducted to enhance robustness. Then, residual blocks with pre-activation operation are developed in down-sampling stage, which effectively improves signal propagation along the network and alleviates network overfitting. Finally, the local cross-channel denoising mechanism is established to eliminate unimportant features without dimensionality reduction. Our proposal was evaluated in Brain Tumor Segmentation 2020 dataset (BraTS 2020), obtaining significantly improved results with mean Dice Similarity Coefficient metric of 0.91, 0.79, and 0.73 for the complete, tumor core, and enhancing tumor regions respectively. Besides, we conduct further practice on BraTS 2019, with the mean Dice Similarity Coefficient metric of 0.89, 0.80, and 0.75. Massive experiments demonstrate that our method is powerful and reliable. It increases little model complexity while achieving very competitive performance.

Novel 3D magnetic resonance fingerprinting radiomics in adult brain tumors: a feasibility study.

To test the feasibility of using 3D MRF maps with radiomics analysis and machine learning in the characterization of adult brain intra-axial neoplasms. 3D MRF acquisition was performed on 78 patients with newly diagnosed brain tumors including 33 glioblastomas (grade IV), 6 grade III gliomas, 12 grade II gliomas, and 27 patients with brain metastases. Regions of enhancing tumor, non-enhancing tumor, and peritumoral edema were segmented and radiomics analysis with gray-level co-occurrence matrices and gray-level run-length matrices was performed. Statistical analysis was performed to identify features capable of differentiating tumors based on type, grade, and isocitrate dehydrogenase (IDH1) status. Receiver operating curve analysis was performed and the area under the curve (AUC) was calculated for tumor classification and grading. For gliomas, Kaplan-Meier analysis for overall survival was performed using MRF T1 features from enhancing tumor region. Multiple MRF T1 and T2 features from enhancing tumor region were capable of differentiating glioblastomas from brain metastases. Although no differences were identified between grade 2 and grade 3 gliomas, differentiation between grade 2 and grade 4 gliomas as well as between grade 3 and grade 4 gliomas was achieved. MRF radiomics features were also able to differentiate IDH1 mutant from the wild-type gliomas. Radiomics T1 features for enhancing tumor region in gliomas correlated to overall survival (p < 0.05). Radiomics analysis of 3D MRF maps allows differentiating glioblastomas from metastases and is capable of differentiating glioblastomas from metastases and characterizing gliomas based on grade, IDH1 status, and survival. • 3D MRF data analysis using radiomics offers novel tissue characterization of brain tumors. • 3D MRF with radiomics offers glioma characterization based on grade, IDH1 status, and overall patient survival.

Cascaded layer-coalescing convolution network for brain tumor segmentation

Accurate segmentation of brain tumor regions from magnetic resonance images continues to be one of the active topics of research due to the high usability levels of the automation process. Faster processing helps clinicians in identification at initial stage of tumor and hence saves valuable time taken for manual image analysis. This work proposes a Cascaded Layer-Coalescing (CLC) model using convolution neural networks for brain tumor segmentation. The process includes three layers of convolution networks, each with cascading inputs from the previous layer and provides multiple outputs segmenting complete, core and enhancing tumor regions. The initial layer identifies complete tumor, coalesces the discriminative features and the input data, and passes it to the core tumor detection layer. The core tumor detection layer in- turn passes discriminative features to the enhancing tumor identification layer. The information injection through data coalescing voxels results in enhanced predictions and also in effective handling of data imbalance, which is a major contributor in model viewpoint. Experiments were performed with Brain Tumor Segmentation (BraTS) 2015 data. A comparison with existing literature works indicate improvements up to35% in sensitivity, 27% in PPV and 28% in Dice Score, indicating improvement in the segmentation process.

Cascade multiscale residual attention CNNs with adaptive ROI for automatic brain tumor segmentation

A brain tumor is one of the fatal cancer types which causes abnormal growth of brain cells. Earlier diagnosis of a brain tumor can play a vital role in its treatment; however, manual segmentation of the brain tumors from MRI images is a laborious and time-consuming task. Therefore, several automatic segmentation techniques have been proposed, but high inter-and intra-tumor variations in shape and texture make accurate segmentation of enhanced and core tumor regions challenging. The demand for a highly accurate and robust segmentation method persists; therefore, in this paper, we propose a novel fully automatic technique for brain tumor regions segmentation by using multiscale residual attention-UNet (MRA-UNet). MRA-UNet uses three consecutive slices as input to preserve the sequential information and employs multiscale learning in a cascade fashion, enabling it to exploit the adaptive region of interest scheme to segment enhanced and core tumor regions accurately. The study also investigates the application of postprocessing techniques (i.e., conditional random field and test time augmentation), which further helps to improve the overall performance. The proposed method has been rigorously evaluated on the most extensive publicly available datasets, i.e., BraTS2017, BraTS2019, and BraTS2020. Our approach achieved state-of-the-art results on the BraTS2020 dataset with the average dice score of 90.18%, 87.22%, and 86.74% for the whole tumor, tumor core, and enhanced tumor regions, respectively. A significant improvement has been demonstrated for the core and enhancing tumor region segmentation showing the proposed method’s evident effectiveness.

Brain Tumor Image Classification using CNN

Abstract: We present a method for segmenting and categorizing brain tumors in the challenge of content of brain tumor with segmentation is enrolled and skull is exposed for bar graph equivalent high-level contradiction refer amount. Preprocessing, segmentation, feature extraction, optimization, and classification are used to detect tumors. The tissue is then classified using preprocessed images. We utilized leave-one-out cross-validation to generate a Dice overlap of 88 for the whole tumor area, 75 for the core tumor region, and 95 for the enhancing tumor region, which is higher than the Dice overlap reported

Programmed Multi-Classification of Brain Tumor Images Using Deep Neural Network

In this paper, we propose a brain tumor segmentation and classification method for multi-modality magnetic resonance image scans. The data from multi-modal brain tumor segmentation challenge are utilized which are co-registered and skull stripped, and the histogram matching is performed with a reference volume of high contrast. We are detecting tumor by using preprocessing , segmentation, feature extraction ,optimization and lastly classification after that preprocessed images use to classify the tissue .We performed a leave-one out cross-validation and achieved 88 Dice overlap for the complete tumor region, 75 for the core tumor region and 95 for enhancing tumor region, which is higher than the Dice overlap reported.

Brain Tumor Classification using CNN

Abstract: In this paper, we propose a brain tumor segmentation and classification method for multi-modality magnetic resonance image scans. The data from multi-modal brain tumor segmentation challenge are utilized which are co-registered and skull stripped, and the histogram matching is performed with a reference volume of high contrast. We are detecting tumor by using preprocessing, segmentation, feature extraction, optimization and lastly classification after that preprocessed image use to classify the tissue. We performed a leave-one out cross-validation and achieved 88 Dice overlap for the complete tumor region, 75 for the core tumor region and 95 for enhancing tumor region, which is higher than the Dice overlap reported. Keywords: Machine Learning, CNN Algorithm, Deep Learning, Classification etc.

Improving geometric P-norm-based glioma segmentation through deep convolutional autoencoder encapsulation

Gliomas are the most common types of brain tumors. Their shapes are irregular and their boundaries are ambiguous, producing a tumor that is hard to detect. To improve the segmentation performance, we propose a P-Norm-Based Glioma Segmentation by encapsulating it within a Deep Convolutional Autoencoder (DCA). In this paper, we first develop a new robust Deep P-norm Generative Model (DPGM) that can extract and classify the feature by the encoder section and segment this latter by the decoder section. After that, we utilise a Deep Discriminative Model (DDM) which applies the conditional random fields to refine the segmentation result. Effectiveness is gained by gradient descent training on the set of pre-P-norm filtering and thresolding convolutional operations using a novel geometric P-norm-based loss function. This latter will be optimized by the RANSAC layer, which is geared toward optimizing the strength of the correct model relative to the most convincing false model. We evaluate the accuracy and robustness of our method within MRI sequences (T1, T2, T2c & FLAIR) from the Multimodal Brain Tumor Image Segmentation Challenge (BRATS 2013 Leaderboard, BRATS 2013 Challenge, BRATS 2015 and BRATS 2017). The experiments demonstrate a high-quality segmentation results of our DPGM-DDM architecture and the average Dice is as high as 97.2%, 97.6%, and 98.4% respectively for complete, core, and enhancing tumor regions. Overall, our deep proposed architecture is effective in segmenting gliomas in multimodal images or fair images, and has the potential in routine examinations of gliomas in daily clinical practice.