Comparison of unsupervised classification methods for brain tumor segmentation using multi-parametric MRI

Advances in the Use of Deep Learning for the Analysis of Magnetic Resonance Image in Neuro-Oncology

Machine Learning is entering a phase of maturity, but its medical applications still lag behind in terms of practical use. The field of oncological radiology (and neuro-oncology in particular) is at the forefront of these developments, now boosted by the success of Deep-Learning methods for the analysis of medical images. This paper reviews in detail some of the most recent advances in the use of Deep Learning in this field, from the broader topic of the development of Machine-Learning-based analytical pipelines to specific instantiations of the use of Deep Learning in neuro-oncology; the latter including its use in the groundbreaking field of ultra-low field magnetic resonance imaging.

Is pre-radiotherapy metabolic heterogeneity of glioblastoma predictive of progression-free survival?

BACKGROUND AND PURPOSE

All glioblastoma subtypes share the hallmark of aggressive invasion, meaning that it is crucial to identify their different components if we are to ensure effective treatment and improve survival. Proton MR spectroscopic imaging (MRSI) is a noninvasive technique that yields metabolic information and is able to identify pathological tissue with high accuracy. The aim of the present study was to identify clusters of metabolic heterogeneity, using a large MRSI dataset, and determine which of these clusters are predictive of progression-free survival (PFS).

MATERIALS AND METHODS

MRSI data of 180 patients acquired in a pre-radiotherapy examination were included in the prospective SPECTRO-GLIO trial. Eight features were extracted for each spectrum: Cho/NAA, NAA/Cr, Cho/Cr, Lac/NAA, and the ratio of each metabolite to the sum of all the metabolites. Clustering of data was performed using a mini-batch k-means algorithm. The Cox model and logrank test were used for PFS analysis.

RESULTS

Five clusters were identified as sharing similar metabolic information and being predictive of PFS. Two clusters revealed metabolic abnormalities. PFS was lower when Cluster 2 was the dominant cluster in patients' MRSI data. Among the metabolites, lactate (present in this cluster and in Cluster 5) was the most statistically significant predictor of poor outcome.

CONCLUSION

Results showed that pre-radiotherapy MRSI can be used to reveal tumor heterogeneity. Groups of spectra, which have the same metabolic information, reflect the different tissue components representative of tumor burden proliferation and hypoxia. Clusters with metabolic abnormalities and high lactate are predictive of PFS.

Factorizer: A scalable interpretable approach to context modeling for medical image segmentation

Convolutional Neural Networks (CNNs) with U-shaped architectures have dominated medical image segmentation, which is crucial for various clinical purposes. However, the inherent locality of convolution makes CNNs fail to fully exploit global context, essential for better recognition of some structures, e.g., brain lesions. Transformers have recently proven promising performance on vision tasks, including semantic segmentation, mainly due to their capability of modeling long-range dependencies. Nevertheless, the quadratic complexity of attention makes existing Transformer-based models use self-attention layers only after somehow reducing the image resolution, which limits the ability to capture global contexts present at higher resolutions. Therefore, this work introduces a family of models, dubbed Factorizer, which leverages the power of low-rank matrix factorization for constructing an end-to-end segmentation model. Specifically, we propose a linearly scalable approach to context modeling, formulating Nonnegative Matrix Factorization (NMF) as a differentiable layer integrated into a U-shaped architecture. The shifted window technique is also utilized in combination with NMF to effectively aggregate local information. Factorizers compete favorably with CNNs and Transformers in terms of accuracy, scalability, and interpretability, achieving state-of-the-art results on the BraTS dataset for brain tumor segmentation and ISLES'22 dataset for stroke lesion segmentation. Highly meaningful NMF components give an additional interpretability advantage to Factorizers over CNNs and Transformers. Moreover, our ablation studies reveal a distinctive feature of Factorizers that enables a significant speed-up in inference for a trained Factorizer without any extra steps and without sacrificing much accuracy. The code and models are publicly available at https://github.com/pashtari/factorizer.

Decoding the Heterogeneity of Malignant Gliomas by PET and MRI for Spatial Habitat Analysis of Hypoxia, Perfusion, and Diffusion Imaging: A Preliminary Study

Background

Tumor heterogeneity poses major clinical challenges in high-grade gliomas (HGGs). Quantitative radiomic analysis with spatial tumor habitat clustering represents an innovative, non-invasive approach to represent and quantify tumor microenvironment heterogeneity. To date, habitat imaging has been applied mainly on conventional magnetic resonance imaging (MRI), although virtually extendible to any imaging modality, including advanced MRI techniques such as perfusion and diffusion MRI as well as positron emission tomography (PET) imaging.

Objectives

This study aims to evaluate an innovative PET and MRI approach for assessing hypoxia, perfusion, and tissue diffusion in HGGs and derive a combined map for clustering of intra-tumor heterogeneity.

Materials and Methods

Seventeen patients harboring HGGs underwent a pre-operative acquisition of MR perfusion (PWI), Diffusion (dMRI) and 18F-labeled fluoroazomycinarabinoside (18F-FAZA) PET imaging to evaluate tumor vascularization, cellularity, and hypoxia, respectively. Tumor volumes were segmented on fluid-attenuated inversion recovery (FLAIR) and T1 post-contrast images, and voxel-wise clustering of each quantitative imaging map identified eight combined PET and physiologic MRI habitats. Habitats’ spatial distribution, quantitative features and histopathological characteristics were analyzed.

Results

A highly reproducible distribution pattern of the clusters was observed among different cases, particularly with respect to morphological landmarks as the necrotic core, contrast-enhancing vital tumor, and peritumoral infiltration and edema, providing valuable supplementary information to conventional imaging. A preliminary analysis, performed on stereotactic bioptic samples where exact intracranial coordinates were available, identified a reliable correlation between the expected microenvironment of the different spatial habitats and the actual histopathological features. A trend toward a higher representation of the most aggressive clusters in WHO (World Health Organization) grade IV compared to WHO III was observed.

Conclusion

Preliminary findings demonstrated high reproducibility of the PET and MRI hypoxia, perfusion, and tissue diffusion spatial habitat maps and correlation with disease-specific histopathological features.

Magnetic Resonance Features of Acquired Immune Deficiency Syndrome Involving Central Nervous System Diseases by Intelligent Fuzzy C-Means Clustering (FCM) Algorithm

This study was aimed to explore the application of fuzzy C-means (FCM) algorithm in MR images of acquired immune deficiency syndrome (AIDS) patients. Sixty AIDS patients with central nervous disease were selected as the research object. A method of brain MR image segmentation based on FCM clustering optimization was proposed, and FCM was optimized based on the neighborhood pixel correlation of gray difference. The correlation was introduced into the objective function to obtain more accurate pixel membership and segmentation features of the image. The segmented image can retain the original image information. The proposed algorithm can clearly distinguish gray matter from white matter in images. The average time of image segmentation was 0.142 s, the longest time of level set algorithm was 2.887 s, and the running time of multithreshold algorithm was 1.708 s. FCM algorithm had the shortest running time, and the average time was significantly better than other algorithms (P < 0.05). FCM image segmentation efficiency was above 90%, and patients can clearly display the location of lesions after MRI imaging examination. In summary, FCM algorithm can effectively combine the spatial neighborhood information of the brain image, segment the BRAIN MR image, analyze the characteristics of AIDS patients from different directions, and provide effective treatment for patients.

Extraction of artefactual MRS patterns from a large database using non-negative matrix factorization

Despite the success of automated pattern recognition methods in problems of human brain tumor diagnostic classification, limited attention has been paid to the issue of automated data quality assessment in the field of MRS for neuro-oncology. Beyond some early attempts to address this issue, the current standard in practice is MRS quality control through human (expert-based) assessment. One aspect of automatic quality control is the problem of detecting artefacts in MRS data. Artefacts, whose variety has already been reviewed in some detail and some of which may even escape human quality control, have a negative influence in pattern recognition methods attempting to assist tumor characterization. The automatic detection of MRS artefacts should be beneficial for radiology as it guarantees more reliable tumor characterizations, as well as the development of more robust pattern recognition-based tumor classifiers and more trustable MRS data processing and analysis pipelines. Feature extraction methods have previously been used to help distinguishing between good and bad quality spectra to apply subsequent supervised pattern recognition techniques. In this study, we apply feature extraction differently and use a variant of a method for blind source separation, namely Convex Non-Negative Matrix Factorization, to unveil MRS signal sources in a completely unsupervised way. We hypothesize that, while most sources will correspond to the different tumor patterns, some of them will reflect signal artefacts. The experimental work reported in this paper, analyzing a combined short and long echo time ¹ H-MRS database of more than 2000 spectra acquired at 1.5T and corresponding to different tumor types and other anomalous masses, provides a first proof of concept that points to the possible validity of this approach.

Fuzzy C-Means Algorithm-Based ARM-Linux-Embedded System Combined with Magnetic Resonance Imaging for Progression Prediction of Brain Tumors

The aim of this research was to analyze the application of fuzzy C-means (FCM) algorithm-based ARM-Linux-embedded system in magnetic resonance imaging (MRI) images for prediction of brain tumors. The optimized FCM (OFCM) algorithm was proposed based on kernel function, and the ARM-Linux-embedded imaging system was designed under ARM9 chip and Linux recorder, which were applied in MRI images of brain tumor patients. It was found that the sensitivity, specificity, and accuracy of the OFCM algorithm (90.46%, 88.97%, and 97.46%) were greater obviously than those of the deterministic C-means clustering algorithm (80.38%, 77.98%, and 85.24%) and the traditional FCM algorithm (83.26%, 79.56%, and 86.45%), and the difference was statistically substantial (P < 0.05). The ME and running time of the OFCM algorithm decreased sharply in contrast to those of the deterministic C-means clustering algorithm and the traditional FCM algorithm (P < 0.05). There were great differences in fraction anisotropy (FA) and mean diffusion (MD) of tumor parenchymal area, surrounding edema area, and normal white matter area (P < 0.05). FA of stage III+IV was smaller than those of stage I and II (P < 0.05), while the apparent diffusion coefficient (ADC) of stage III+IV was greater than that of stage I and II (P < 0.05). In conclusion, the poor update data processing and low data clustering efficiency of FCM were solved by OFCM. Moreover, computational efficiency of ARM-Linux-embedded imaging system was improved, so as to better realize the prediction of brain tumor patients through ARM-Linux-embedded system based on adaptive FCM incremental clustering algorithm.

3D AGSE-VNet: an automatic brain tumor MRI data segmentation framework

BACKGROUND

Glioma is the most common brain malignant tumor, with a high morbidity rate and a mortality rate of more than three percent, which seriously endangers human health. The main method of acquiring brain tumors in the clinic is MRI. Segmentation of brain tumor regions from multi-modal MRI scan images is helpful for treatment inspection, post-diagnosis monitoring, and effect evaluation of patients. However, the common operation in clinical brain tumor segmentation is still manual segmentation, lead to its time-consuming and large performance difference between different operators, a consistent and accurate automatic segmentation method is urgently needed. With the continuous development of deep learning, researchers have designed many automatic segmentation algorithms; however, there are still some problems: (1) The research of segmentation algorithm mostly stays on the 2D plane, this will reduce the accuracy of 3D image feature extraction to a certain extent. (2) MRI images have gray-scale offset fields that make it difficult to divide the contours accurately.

METHODS

To meet the above challenges, we propose an automatic brain tumor MRI data segmentation framework which is called AGSE-VNet. In our study, the Squeeze and Excite (SE) module is added to each encoder, the Attention Guide Filter (AG) module is added to each decoder, using the channel relationship to automatically enhance the useful information in the channel to suppress the useless information, and use the attention mechanism to guide the edge information and remove the influence of irrelevant information such as noise.

RESULTS

We used the BraTS2020 challenge online verification tool to evaluate our approach. The focus of verification is that the Dice scores of the whole tumor, tumor core and enhanced tumor are 0.68, 0.85 and 0.70, respectively.

CONCLUSION

Although MRI images have different intensities, AGSE-VNet is not affected by the size of the tumor, and can more accurately extract the features of the three regions, it has achieved impressive results and made outstanding contributions to the clinical diagnosis and treatment of brain tumor patients.

Morphological active contour model for automatic brain tumor extraction from multimodal magnetic resonance images

BACKGROUND

Brain tumor extraction from magnetic resonance (MR) images is challenging due to variations in the location, shape, size and intensity of tumors. Manual delineation of brain tumors from MR images is time-consuming and prone to human errors.

METHOD

In this paper, we present a method for automatic tumor extraction from multimodal MR images. Brain tumors are first detected using k-means clustering. A morphological region-based active contour model is then used for tumor extraction using an initial contour defined based on the boundary of the detected brain tumor regions. The contour evolution for tumor extraction was performed using successive application of morphological operators. In our model, a Gaussian distribution was used to model local image intensities. The spatial correlation between neighboring voxels was also modeled using Markov random field.

RESULTS

The proposed method was evaluated on BraTS 2013 dataset including patients with high-grade and low-grade tumors. In comparison with other active contour based methods, the proposed method yielded better performance on tumor segmentation with mean Dice similarity coefficients of 0.9179 ( ± 0.025) and 0.8910 ( ± 0.042) obtained on high-grade and low-grade tumors, respectively.

CONCLUSION

The proposed method achieved higher accuracies for brain tumor extraction in comparison to other contour-based methods.

AUCseg: An Automatically Unsupervised Clustering Toolbox for 3D-Segmentation of High-Grade Gliomas in Multi-Parametric MR Images

The segmentation of high-grade gliomas (HGG) using magnetic resonance imaging (MRI) data is clinically meaningful in neurosurgical practice, but a challenging task. Currently, most segmentation methods are supervised learning with labeled training sets. Although these methods work well in most cases, they typically require time-consuming manual labeling and pre-trained models. In this work, we propose an automatically unsupervised segmentation toolbox based on the clustering algorithm and morphological processing, named AUCseg. With our toolbox, the whole tumor was first extracted by clustering on T2-FLAIR images. Then, based on the mask acquired with whole tumor segmentation, the enhancing tumor was segmented on the post-contrast T1-weighted images (T1-CE) using clustering methods. Finally, the necrotic regions were segmented by morphological processing or clustering on T2-weighted images. Compared with K-means, Mini-batch K-means, and Fuzzy C Means (FCM), the Gaussian Mixture Model (GMM) clustering performs the best in our toolbox. We did a multi-sided evaluation of our toolbox in the BraTS2018 dataset and demonstrated that the whole tumor, tumor core, and enhancing tumor can be automatically segmented using default hyper-parameters with Dice score 0.8209, 0.7087, and 0.7254, respectively. The computing time of our toolbox for each case is around 22 seconds, which is at least 3 times faster than other state-of-the-art unsupervised methods. In addition, our toolbox has an option to perform semi-automatic segmentation via manually setup hyper-parameters, which could improve the segmentation performance. Our toolbox, AUCseg, is publicly available on Github. (https://github.com/Haifengtao/AUCseg).

Psychological analysis of classroom learning based on face recognition and neural network

With the rapid development of deep learning and parallel computing, deep learning neural network based on big data has been applied to the field of facial nerve recognition. This innovative operation has attracted extensive attention of scholars. The reason why the application of neural network is realized lies in deep learning, which can reduce the error and change the weight by means of back propagation and error optimization, so as to extract more key points and features. In spite of this, data collection and key points extraction is still a very complex problem. This paper mainly aims at the above problems, studies the way of deep learning and information extraction and its internal structure, and optimizes its application to classroom learning, so as to provide effective help for the realization of distance education.

Data mining and XBRL integration in management accounting information based on artificial intelligence

In today’s society, the application of information technology is becoming more and more extensive. At the same time, management accounting, as an important branch of modern accounting, also ushered in new development opportunities, and the research of data mining also pays more attention to the combination of theory and practice. Therefore, data mining can provide some technical support for the implementation of strategic management accounting. Because the most important thing of management accounting informatization is to process, calculate and transmit the business information of the enterprise through the corresponding information processing platform through the use of computer technology, and provide the corresponding data to the management of the major companies in order to better analyze and make decisions and perfect the future development strategy of the enterprise, so the screening of the corresponding technology is more important in the process of management accounting informatization. Based on the development of artificial intelligence technology and the combination of data mining and XBRL technology, this paper discusses the new strategies of contemporary management accounting development.

Application research of automobile modeling optimization design based on virtual reality technology

Automobile styling design is an important part of the design chain. In the traditional automobile modeling evaluation, the process of project evaluation is more in-depth, and designers exchange ideas. Different designers have different evaluations of automobile styling. The evaluation process lasts a long time, which leads to the design cycle being too long and the efficiency of automobile modeling evaluation is greatly reduced. The introduction of virtual reality in automobile modeling evaluation can effectively optimize the evaluation process and promote the rapid adjustment of the model on the basis of development. From the virtual reality system based on mechanical engineering, we only need the parameters of the car model to observe the actual situation through VR technology, and use the measurement tools to directly and accurately evaluate the driver’s field of vision. Through the application of virtual reality technology in the automobile design stage, the interactive and network-based remote research on automobile modeling will also make the automobile design process more convenient, easier to communicate with designers, and reduce the development cycle and cost of automobile design.

An improved framework for brain tumor analysis using MRI based on YOLOv2 and convolutional neural network

Brain tumor is a group of anomalous cells. The brain is enclosed in a more rigid skull. The abnormal cell grows and initiates a tumor. Detection of tumor is a complicated task due to irregular tumor shape. The proposed technique contains four phases, which are lesion enhancement, feature extraction and selection for classification, localization, and segmentation. The magnetic resonance imaging (MRI) images are noisy due to certain factors, such as image acquisition, and fluctuation in magnetic field coil. Therefore, a homomorphic wavelet filer is used for noise reduction. Later, extracted features from inceptionv3 pre-trained model and informative features are selected using a non-dominated sorted genetic algorithm (NSGA). The optimized features are forwarded for classification after which tumor slices are passed to YOLOv2-inceptionv3 model designed for the localization of tumor region such that features are extracted from depth-concatenation (mixed-4) layer of inceptionv3 model and supplied to YOLOv2. The localized images are passed toMcCulloch'sKapur entropy method to segment actual tumor region. Finally, the proposed technique is validated on three benchmark databases BRATS 2018, BRATS 2019, and BRATS 2020 for tumor detection. The proposed method achieved greater than 0.90 prediction scores in localization, segmentation and classification of brain lesions. Moreover, classification and segmentation outcomes are superior as compared to existing methods.

Multi-parametric MR Imaging Biomarkers Associated to Clinical Outcomes in Gliomas: A Systematic Review

PURPOSE

To systematically review evidence regarding the association of multiparametric biomarkers with clinical outcomes and their capacity to explain relevant subcompartments of gliomas.

MATERIALS AND METHODS

Scopus database was searched for original journal papers from January 1st, 2007 to February 20th, 2017 according to PRISMA. Four hundred forty-nine abstracts of papers were reviewed and scored independently by two out of six authors. Based on those papers we analyzed associations between biomarkers, subcompartments within the tumor lesion, and clinical outcomes. From all the articles analyzed, the twenty-seven papers with the highest scores were highlighted to represent the evidence about MR imaging biomarkers associated with clinical outcomes. Similarly, eighteen studies defining subcompartments within the tumor region were also highlighted to represent the evidence of MR imaging biomarkers. Their reports were critically appraised according to the QUADAS-2 criteria.

RESULTS

It has been demonstrated that multi-parametric biomarkers are prepared for surrogating diagnosis, grading, segmentation, overall survival, progression-free survival, recurrence, molecular profiling and response to treatment in gliomas. Quantifications and radiomics features obtained from morphological exams (T1, T2, FLAIR, T1c), PWI (including DSC and DCE), diffusion (DWI, DTI) and chemical shift imaging (CSI) are the preferred MR biomarkers associated to clinical outcomes. Subcompartments relative to the peritumoral region, invasion, infiltration, proliferation, mass effect and pseudo flush, relapse compartments, gross tumor volumes, and highrisk regions have been defined to characterize the heterogeneity. For the majority of pairwise cooccurrences, we found no evidence to assert that observed co-occurrences were significantly different from their expected co-occurrences (Binomial test with False Discovery Rate correction, α=0.05). The co-occurrence among terms in the studied papers was found to be driven by their individual prevalence and trends in the literature.

CONCLUSION

Combinations of MR imaging biomarkers from morphological, PWI, DWI and CSI exams have demonstrated their capability to predict clinical outcomes in different management moments of gliomas. Whereas morphologic-derived compartments have been mostly studied during the last ten years, new multi-parametric MRI approaches have also been proposed to discover specific subcompartments of the tumors. MR biomarkers from those subcompartments show the local behavior within the heterogeneous tumor and may quantify the prognosis and response to treatment of gliomas.

Transferrin receptor 1 targeted optical imaging for identifying glioma margin in mouse models

OBJECTIVE

Optical molecular imaging technology that indiscriminately detects intracranial glioblastoma (GBM) can help neurosurgeons effectively remove tumor masses. Transferrin receptor 1 (TfR 1) is a diagnostic and therapeutic target in GBM. A TfR 1-targeted peptide, CRTIGPSVC (CRT), was shown to cross the blood brain barrier (BBB) and accumulate at high levels in GBM tissues. In this study, we synthesized a TfR 1-targeted near-infrared fluorescent (NIRF) probe, Cy5-CRT, for identifying the GBM tissue margin in mouse models.

METHODS

We initially confirmed the overexpression of TfR 1 in GBM and the tumor-specific homing ability of Cy5-CRT in subcutaneous and orthotopic GBM mouse models. We then examined the feasibility of Cy5-CRT for identifying the tumor margin in orthotopic GBM xenografts. Finally, we compared Cy5-CRT with the clinically used fluorescein sodium in identifying tumor margins.

RESULTS

Cy5-CRT specifically accumulated in GBM tissues and detected the tumor burden with exceptional contrast in mice with orthotopic GBM, enabling fluorescence-guided GBM resection under NIRF live imaging conditions. Importantly, Cy5-CRT recognized the GBM tissue margin more clearly than fluorescein sodium.

CONCLUSIONS

The TfR 1-targeted optical probe Cy5-CRT specifically differentiates tumor tissues from the surrounding normal brain with high sensitivity, indicating its potential application for the precise surgical removal of GBM.

Embedding MRI information into MRSI data source extraction improves brain tumour delineation in animal models

Glioblastoma is the most frequent malignant intra-cranial tumour. Magnetic resonance imaging is the modality of choice in diagnosis, aggressiveness assessment, and follow-up. However, there are examples where it lacks diagnostic accuracy. Magnetic resonance spectroscopy enables the identification of molecules present in the tissue, providing a precise metabolomic signature. Previous research shows that combining imaging and spectroscopy information results in more accurate outcomes and superior diagnostic value. This study proposes a method to combine them, which builds upon a previous methodology whose main objective is to guide the extraction of sources. To this aim, prior knowledge about class-specific information is integrated into the methodology by setting the metric of a latent variable space where Non-negative Matrix Factorisation is performed. The former methodology, which only used spectroscopy and involved combining spectra from different subjects, was adapted to use selected areas of interest that arise from segmenting the T2-weighted image. Results showed that embedding imaging information into the source extraction (the proposed semi-supervised analysis) improved the quality of the tumour delineation, as compared to those obtained without this information (unsupervised analysis). Both approaches were applied to pre-clinical data, involving thirteen brain tumour-bearing mice, and tested against histopathological data. On results of twenty-eight images, the proposed Semi-Supervised Source Extraction (SSSE) method greatly outperformed the unsupervised one, as well as an alternative semi-supervised approach from the literature, with differences being statistically significant. SSSE has proven successful in the delineation of the tumour, while bringing benefits such as 1) not constricting the metabolomic-based prediction to the image-segmented area, 2) ability to deal with signal-to-noise issues, 3) opportunity to answer specific questions by allowing researchers/radiologists define areas of interest that guide the source extraction, 4) creation of an intra-subject model and avoiding contamination from inter-subject overlaps, and 5) extraction of meaningful, good-quality sources that adds interpretability, conferring validation and better understanding of each case.

A potential field segmentation based method for tumor segmentation on multi-parametric MRI of glioma cancer patients

Background

Accurate segmentation of brain tumors is vital for the gross tumor volume (GTV) definition in radiotherapy. Functional MR images like apparent diffusion constant (ADC) and fractional anisotropy (FA) images can provide more comprehensive information for sensitive detection of the GTV. We synthesize anatomical and functional MRI for accurate and semi-automatic segmentation of GTVs and improvement of clinical efficiency.

Methods

Four MR image sets including T1-weighted contrast-enhanced (T1C), T2-weighted (T2), apparent diffusion constant (ADC) and fractional anisotropy (FA) images of 5 glioma patients were acquired and registered. A new potential field segmentation (PFS) method was proposed based on the concept of potential field in physics. For T1C, T2 and ADC images, global potential field segmentation (global-PFS) was used on user defined region of interest (ROI) for rough segmentation and then morphologically processed for accurate delineation of the GTV. For FA images, white matter (WM) was removed using local potential field segmentation (local-PFS), and then tumor extent was delineated with region growing and morphological methods. The individual segmentations of multi-parametric images were ensembled into a fused segmentation, considered as final GTV. GTVs were compared with manually delineated ground truth and evaluated with segmentation quality measure (Q), Dice’s similarity coefficient (DSC) and Sensitivity and Specificity.

Results

Experimental study with the five patients’ data and new method showed that, the mean values of Q, DSC, Sensitivity and Specificity were 0.80 (±0.07), 0.88 (±0.04), 0.92 (±0.01) and 0.88 (±0.05) respectively. The global-PFS used on ROIs of T1C, T2 and ADC images can avoid interferences from skull and other non-tumor areas. Similarity to local-PFS on FA images, it can also reduce the time complexity as compared with the global-PFS on whole image sets.

Conclusions

Efficient and semi-automatic segmentation of the GTV can be achieved with the new method. Combination of anatomical and functional MR images has the potential to provide new methods and ideas for target definition in radiotherapy.

A modality-adaptive method for segmenting brain tumors and organs-at-risk in radiation therapy planning

In this paper we present a method for simultaneously segmenting brain tumors and an extensive set of organs-at-risk for radiation therapy planning of glioblastomas. The method combines a contrast-adaptive generative model for whole-brain segmentation with a new spatial regularization model of tumor shape using convolutional restricted Boltzmann machines. We demonstrate experimentally that the method is able to adapt to image acquisitions that differ substantially from any available training data, ensuring its applicability across treatment sites; that its tumor segmentation accuracy is comparable to that of the current state of the art; and that it captures most organs-at-risk sufficiently well for radiation therapy planning purposes. The proposed method may be a valuable step towards automating the delineation of brain tumors and organs-at-risk in glioblastoma patients undergoing radiation therapy.

Brain Tumor Segmentation Based on Improved Convolutional Neural Network in Combination with Non-quantifiable Local Texture Feature

Accurate and reliable brain tumor segmentation is a critical component in cancer diagnosis. According to deep learning model, a novel brain tumor segmentation method is developed by integrating fully convolutional neural networks (FCNN) and dense micro-block difference feature (DMDF) into a unified framework so as to obtain segmentation results with appearance and spatial consistency. Firstly, we propose a local feature to describe the rotation invariant property of the texture. In order to deal with the change of rotation and scale in texture image, Fisher vector encoding method is used to analyze the texture feature, which can combine with the scale information without increasing the dimension of the local feature. The obtained local features have strong robustness to rotation and gray intensity variation. Then, the non-quantifiable local feature is fused to the FCNN to perform fine boundary segmentation. Since brain tumors occupy a small portion of the image, deconvolutional layers are designed with skip connections to obtain a high quality feature map. Compared with the traditional MRI brain tumor segmentation methods, the experimental results show that the segmentation accuracy and stability has been greatly improved. Average Dice index can be up to 90.98%. And the proposed method has very high real-time performance, where brain tumor image can segment within 1 s.

Brain abnormality detection using template matching

Magnetic resonance imaging (MRI) is a widely used imaging modality to evaluate brain disorders. MRI generates huge volumes of data, which consist of a sequence of scans taken at different instances of time. As the presence of brain disorders has to be evaluated on all magnetic resonance (MR) sequences, manual brain disorder detection becomes a tedious process and is prone to inter- and intra-rater errors. A technique for detecting abnormalities in brain MRI using template matching is proposed. Bias filed correction is performed on volumetric scans using N4ITK filter, followed by volumetric registration. Normalized cross-correlation template matching is used for image registration taking into account, the rotation and scaling operations. A template of abnormality is selected which is then matched in the volumetric scans, if found, the corresponding image is retrieved. Post-processing of the retrieved images is performed by the thresholding operation; the coordinates and area of the abnormality are reported. The experiments are carried out on the glioma dataset obtained from Brain Tumor Segmentation Challenge 2013 database (BRATS 2013). Glioma dataset consisted of MR scans of 30 real glioma patients and 50 simulated glioma patients. NVIDIA Compute Unified Device Architecture framework is employed in this paper, and it is found that the detection speed using graphics processing unit is almost four times faster than using only central processing unit. The average Dice and Jaccard coefficients for a wide range of trials are found to be 0.91 and 0.83, respectively.

Diffusion Kurtosis Imaging Reflects Glial Fibrillary Acidic Protein (GFAP), Topo IIα, and O⁶-Methylguanine-DNA Methyltransferase (MGMT) Expression in Astrocytomas

BACKGROUND Astrocytomas are the most common primary brain neoplasms. Biological indicators of astrocytomas can reflect its biological characteristics. The aim of this study was to assess the expression of the pathological glial fibrillary acidic protein (GFAP) Topo IIα and O⁶-methylguanine-DNA methyltransferase (MGMT) in astrocytomas using magnetic resonance (MR) diffusion kurtosis imaging (DKI) to evaluate the biological characteristics of astrocytomas. MATERIAL AND METHODS Sixty-six patients with pathologically proven astrocytomas were enrolled in this study. All patients underwent conventional MRI head scanning, DKI scanning, and enhanced scanning under the same conditions. Spearman's rank correlation analysis and Bonferroni correction were used to compare the values of DKI and the expression levels of GFAP, Topo IIα, and MGMT between the 2 groups. RESULTS Mean kurtosis (MK) values were negatively correlated with the expression of GFAP (r=-0.836; P=0.03). However, these were positively correlated with the expression of Topo IIα (r=0.896; P=0.01). Moreover, fractional anisotropy (FA) values were not correlated with the expression of GFAP (r=0.366; P=0.05), Topo IIα (r=-0.562; P=0.05), or MGMT (r=-0.153; P=0.10). CONCLUSIONS MK was significantly associated with the expression of GFAP and Topo IIα. To a certain extent, applying DKI may show the biological behavior of tumor cell differentiation, proliferation activity, invasion, and metastasis, and guide individual treatment.

Fully automated tissue segmentation of the prescription isodose region delineated through the Gamma knife plan for cerebral arteriovenous malformation (AVM) using fuzzy C-means (FCM) clustering

Background

Gamma knife radiosurgery (GKRS) is a common treatment for cerebral arterio-venous malformations (AVMs), particularly in cases where the malformation is deep-seated, large, or in eloquent areas of the brain. Unfortunately, these procedures can result in radiation injury to brain parenchyma. The fact that every AVM is unique in its vascular morphology makes it nearly impossible to exclude brain parenchyma from isodose radiation exposure during the formulation of a GKRS plan. Calculating the percentages of the various forms of tissue exposed to specific doses of radiation is crucial to understanding the clinical responses and causes of brain parenchyma injury following GKRS for AVM.

Methods

In this study, we developed a fully automated algorithm using unsupervised classification via fuzzy c-means clustering for the analysis of T2 weighted images used in a Gamma knife plan. This algorithm is able to calculate the percentages of nidus, brain tissue, and cerebrospinal fluid (CSF) within the prescription isodose radiation exposure region.

Results

The proposed algorithm was used to assess the treatment plan of 25 patients with AVM who had undergone GKRS. The Dice similarity index (SI) was used to determine the degree of agreement between the results obtained using the algorithm and a visually guided manual method (the gold standard) performed by an experienced neurosurgeon. In the nidus, the SI was (74.86 ± 1.30%) (mean ± standard deviation), the sensitivity was (83.05 ± 11.91)%, and the specificity was (86.73 ± 10.31)%. In brain tissue, the SI was (79.50 ± 6.01)%, the sensitivity was (73.05 ± 9.77)%, and the specificity was (85.53 ± 7.13)%. In the CSF, the SI was (69.57 ± 15.26)%, the sensitivity was (89.86 ± 5.87)%, and the specificity was (92.36 ± 4.35)%.

Conclusions

The proposed clustering algorithm provides precise percentages of the various types of tissue within the prescription isodose region in the T2 weighted images used in the GKRS plan for AVM. Our results shed light on the causes of brain radiation injury after GKRS for AVM. In the future, this system could be used to improve outcomes and avoid complications associated with GKRS treatment.

•

A novel image analytical method for the analysis of images of an AVM in a GKRS plan

•

Fuzzy c-means clustering was used for analyses of T2w images in the GKRS plan.

•

Automatic calculation of percentages of tissue inside the isodose line

•

Brain tissue percentages of the nidus of the AVM predict risk of complication.

•

Proposed method could be used to avoid complications associated after GKRS.

Brain tumor segmentation with Vander Lugt correlator based active contour

BACKGROUND AND OBJECTIVE

The manual segmentation of brain tumors from medical images is an error-prone, sensitive, and time-absorbing process. This paper presents an automatic and fast method of brain tumor segmentation.

METHODS

In the proposed method, a numerical simulation of the optical Vander Lugt correlator is used for automatically detecting the abnormal tissue region. The tumor filter, used in the simulated optical correlation, is tailored to all the brain tumor types and especially to the Glioblastoma, which considered to be the most aggressive cancer. The simulated optical correlation, computed between Magnetic Resonance Images (MRI) and this filter, estimates precisely and automatically the initial contour inside the tumorous tissue. Further, in the segmentation part, the detected initial contour is used to define an active contour model and presenting the problematic as an energy minimization problem. As a result, this initial contour assists the algorithm to evolve an active contour model towards the exact tumor boundaries. Equally important, for a comparison purposes, we considered different active contour models and investigated their impact on the performance of the segmentation task. Several images from BRATS database with tumors anywhere in images and having different sizes, contrast, and shape, are used to test the proposed system. Furthermore, several performance metrics are computed to present an aggregate overview of the proposed method advantages.

RESULTS

The proposed method achieves a high accuracy in detecting the tumorous tissue by a parameter returned by the simulated optical correlation. In addition, the proposed method yields better performance compared to the active contour based methods with the averages of Sensitivity=0.9733, Dice coefficient = 0.9663, Hausdroff distance = 2.6540, Specificity = 0.9994, and faster with a computational time average of 0.4119 s per image.

CONCLUSIONS

Results reported on BRATS database reveal that our proposed system improves over the recently published state-of-the-art methods in brain tumor detection and segmentation.

Supervised learning based multimodal MRI brain tumour segmentation using texture features from supervoxels

BACKGROUND

Accurate segmentation of brain tumour in magnetic resonance images (MRI) is a difficult task due to various tumour types. Using information and features from multimodal MRI including structural MRI and isotropic (p) and anisotropic (q) components derived from the diffusion tensor imaging (DTI) may result in a more accurate analysis of brain images.

METHODS

We propose a novel 3D supervoxel based learning method for segmentation of tumour in multimodal MRI brain images (conventional MRI and DTI). Supervoxels are generated using the information across the multimodal MRI dataset. For each supervoxel, a variety of features including histograms of texton descriptor, calculated using a set of Gabor filters with different sizes and orientations, and first order intensity statistical features are extracted. Those features are fed into a random forests (RF) classifier to classify each supervoxel into tumour core, oedema or healthy brain tissue.

RESULTS

The method is evaluated on two datasets: 1) Our clinical dataset: 11 multimodal images of patients and 2) BRATS 2013 clinical dataset: 30 multimodal images. For our clinical dataset, the average detection sensitivity of tumour (including tumour core and oedema) using multimodal MRI is 86% with balanced error rate (BER) 7%; while the Dice score for automatic tumour segmentation against ground truth is 0.84. The corresponding results of the BRATS 2013 dataset are 96%, 2% and 0.89, respectively.

CONCLUSION

The method demonstrates promising results in the segmentation of brain tumour. Adding features from multimodal MRI images can largely increase the segmentation accuracy. The method provides a close match to expert delineation across all tumour grades, leading to a faster and more reproducible method of brain tumour detection and delineation to aid patient management.

Radiomics in radiooncology - Challenging the medical physicist

PURPOSE

Noticing the fast growing translation of artificial intelligence (AI) technologies to medical image analysis this paper emphasizes the future role of the medical physicist in this evolving field. Specific challenges are addressed when implementing big data concepts with high-throughput image data processing like radiomics and machine learning in a radiooncology environment to support clinical decisions.

METHODS

Based on the experience of our interdisciplinary radiomics working group, techniques for processing minable data, extracting radiomics features and associating this information with clinical, physical and biological data for the development of prediction models are described. A special emphasis was placed on the potential clinical significance of such an approach.

RESULTS

Clinical studies demonstrate the role of radiomics analysis as an additional independent source of information with the potential to influence the radiooncology practice, i.e. to predict patient prognosis, treatment response and underlying genetic changes. Extending the radiomics approach to integrate imaging, clinical, genetic and dosimetric data ('panomics') challenges the medical physicist as member of the radiooncology team.

CONCLUSIONS

The new field of big data processing in radiooncology offers opportunities to support clinical decisions, to improve predicting treatment outcome and to stimulate fundamental research on radiation response both of tumor and normal tissue. The integration of physical data (e.g. treatment planning, dosimetric, image guidance data) demands an involvement of the medical physicist in the radiomics approach of radiooncology. To cope with this challenge national and international organizations for medical physics should organize more training opportunities in artificial intelligence technologies in radiooncology.

Tumor Tissue Detection using Blood-Oxygen-Level-Dependent Functional MRI based on Independent Component Analysis

Accurate delineation of gliomas from the surrounding normal brain areas helps maximize tumor resection and improves outcome. Blood-oxygen-level-dependent (BOLD) functional MRI (fMRI) has been routinely adopted for presurgical mapping of the surrounding functional areas. For completely utilizing such imaging data, here we show the feasibility of using presurgical fMRI for tumor delineation. In particular, we introduce a novel method dedicated to tumor detection based on independent component analysis (ICA) of resting-state fMRI (rs-fMRI) with automatic tumor component identification. Multi-center rs-fMRI data of 32 glioma patients from three centers, plus the additional proof-of-concept data of 28 patients from the fourth center with non-brain musculoskeletal tumors, are fed into individual ICA with different total number of components (TNCs). The best-fitted tumor-related components derived from the optimized TNCs setting are automatically determined based on a new template-matching algorithm. The success rates are 100%, 100% and 93.75% for glioma tissue detection for the three centers, respectively, and 85.19% for musculoskeletal tumor detection. We propose that the high success rate could come from the previously overlooked ability of BOLD rs-fMRI in characterizing the abnormal vascularization, vasomotion and perfusion caused by tumors. Our findings suggest an additional usage of the rs-fMRI for comprehensive presurgical assessment.

A Five-Level Wavelet Decomposition and Dimensional Reduction Approach for Feature Extraction and Classification of MR and CT Scan Images

This paper presents a two-dimensional wavelet based decomposition algorithm for classification of biomedical images. The two-dimensional wavelet decomposition is done up to five levels for the input images. Histograms of decomposed images are then used to form the feature set. This feature set is further reduced using probabilistic principal component analysis. The reduced set of features is then fed into either K nearest neighbor algorithm or feed-forward artificial neural network, to classify images. The algorithm is compared with three other techniques in terms of accuracy. The proposed algorithm has been found better up to 3.3%, 12.75%, and 13.75% on average over the first, second, and third algorithm, respectively, using KNN and up to 6.22%, 13.9%, and 14.1% on average using ANN. The dataset used for comparison consisted of CT Scan images of lungs and MR images of heart as obtained from different sources.