scispace - formally typeset
Search or ask a question
Author

Junfeng Lu

Bio: Junfeng Lu is an academic researcher from Fudan University. The author has contributed to research in topics: Glioma & Resting state fMRI. The author has an hindex of 15, co-authored 47 publications receiving 567 citations.

Papers
More filters
Journal ArticleDOI
TL;DR: This study proposes a multi-channel architecture of 3D convolutional neural networks (CNNs) for deep learning upon those metric maps, from which high-level predictive features are extracted for each individual patch of these maps.
Abstract: High-grade gliomas are the most aggressive malignant brain tumors. Accurate pre-operative prognosis for this cohort can lead to better treatment planning. Conventional survival prediction based on clinical information is subjective and could be inaccurate. Recent radiomics studies have shown better prognosis by using carefully-engineered image features from magnetic resonance images (MRI). However, feature engineering is usually time consuming, laborious and subjective. Most importantly, the engineered features cannot effectively encode other predictive but implicit information provided by multi-modal neuroimages. We propose a two-stage learning-based method to predict the overall survival (OS) time of high-grade gliomas patient. At the first stage, we adopt deep learning, a recently dominant technique of artificial intelligence, to automatically extract implicit and high-level features from multi-modal, multi-channel preoperative MRI such that the features are competent of predicting survival time. Specifically, we utilize not only contrast-enhanced T1 MRI, but also diffusion tensor imaging (DTI) and resting-state functional MRI (rs-fMRI), for computing multiple metric maps (including various diffusivity metric maps derived from DTI, and also the frequency-specific brain fluctuation amplitude maps and local functional connectivity anisotropy-related metric maps derived from rs-fMRI) from 68 high-grade glioma patients with different survival time. We propose a multi-channel architecture of 3D convolutional neural networks (CNNs) for deep learning upon those metric maps, from which high-level predictive features are extracted for each individual patch of these maps. At the second stage, those deeply learned features along with the pivotal limited demographic and tumor-related features (such as age, tumor size and histological type) are fed into a support vector machine (SVM) to generate the final prediction result (i.e., long or short overall survival time). The experimental results demonstrate that this multi-model, multi-channel deep survival prediction framework achieves an accuracy of 90.66%, outperforming all the competing methods. This study indicates highly demanded effectiveness on prognosis of deep learning technique in neuro-oncological applications for better individualized treatment planning towards precision medicine.

123 citations

Journal ArticleDOI
TL;DR: This study demonstrates the potential utility of combining awake craniotomy with iMRI; it is safe and reliable to perform awake surgery using a movable iMRI.

61 citations

Journal ArticleDOI
TL;DR: R-fMRI sensitivity and specificity are high for localizing hand motor area and even equivalent or slightly higher compared with T-f MRI, given its convenience for patients.
Abstract: Resting-state functional magnetic resonance imaging (R-fMRI) is a promising tool in clinical application, especially in presurgical mapping for neurosurgery. This study aimed to investigate the sensitivity and specificity of R-fMRI in the localization of hand motor area in patients with brain tumors validated by direct cortical stimulation (DCS). We also compared this technique to task-based blood oxygenation level-dependent (BOLD) fMRI (T-fMRI). R-fMRI and T-fMRI were acquired from 17 patients with brain tumors. The cortex sites of the hand motor area were recorded by DCS. Site-by-site comparisons between R-fMRI/T-fMRI and DCS were performed to calculate R-fMRI and T-fMRI sensitivity and specificity using DCS as a “gold standard”. R-fMRI and T-fMRI performances were compared statistically A total of 609 cortex sites were tested with DCS and compared with R-fMRI findings in 17 patients. For hand motor area localization, R-fMRI sensitivity and specificity were 90.91 and 89.41 %, respectively. Given that two subjects could not comply with T-fMRI, 520 DCS sites were compared with T-fMRI findings in 15 patients. The sensitivity and specificity of T-fMRI were 78.57 and 84.76 %, respectively. In the 15 patients who successfully underwent both R-fMRI and T-fMRI, there was no statistical difference in sensitivity or specificity between the two methods (p = 0.3198 and p = 0.1431, respectively) R-fMRI sensitivity and specificity are high for localizing hand motor area and even equivalent or slightly higher compared with T-fMRI. Given its convenience for patients, R-fMRI is a promising substitute for T-fMRI for presurgical mapping

45 citations

Journal ArticleDOI
TL;DR: A new deep learning-based OS prediction method for GBM patients, which can derive tumor genotype-related features from pre-operative multimodal magnetic resonance imaging (MRI) brain data and feed them to OS prediction, achieves the highest OS prediction accuracy compared to other state-of-the-art methods.
Abstract: Glioblastoma (GBM) is the most common and deadly malignant brain tumor. For personalized treatment, an accurate pre-operative prognosis for GBM patients is highly desired. Recently, many machine learning-based methods have been adopted to predict overall survival (OS) time based on the pre-operative mono- or multi-modal imaging phenotype. The genotypic information of GBM has been proven to be strongly indicative of the prognosis; however, this has not been considered in the existing imaging-based OS prediction methods. The main reason is that the tumor genotype is unavailable pre-operatively unless deriving from craniotomy. In this paper, we propose a new deep learning-based OS prediction method for GBM patients, which can derive tumor genotype-related features from pre-operative multimodal magnetic resonance imaging (MRI) brain data and feed them to OS prediction. Specifically, we propose a multi-task convolutional neural network (CNN) to accomplish both tumor genotype and OS prediction tasks jointly . As the network can benefit from learning tumor genotype-related features for genotype prediction, the accuracy of predicting OS time can be prominently improved. In the experiments, multimodal MRI brain dataset of 120 GBM patients, with as many as four different genotypic/molecular biomarkers, are used to evaluate our method. Our method achieves the highest OS prediction accuracy compared to other state-of-the-art methods.

45 citations

Journal ArticleDOI
TL;DR: An automated method for identifying language network in brain tumor subjects using ICA on rs-fMRI-based pre-surgical functional mapping is developed and successfully applied to brain tumor patients.
Abstract: As a noninvasive and “task-free” technique, resting-state functional magnetic resonance imaging (rs-fMRI) has been gradually applied to pre-surgical functional mapping. Independent component analysis (ICA)-based mapping has shown advantage, as no a priori information is required. We developed an automated method for identifying language network in brain tumor subjects using ICA on rs-fMRI. In addition to standard processing strategies, we applied a discriminability-index-based component identification algorithm to identify language networks in three different groups. The results from the training group were validated in an independent group of healthy human subjects. For the testing group, ICA and seed-based correlation were separately computed and the detected language networks were assessed by intra-operative stimulation mapping to verify reliability of application in the clinical setting. Individualized language network mapping could be automatically achieved for all subjects from the two healthy groups except one (19/20, success rate = 95.0%). In the testing group (brain tumor patients), the sensitivity of the language mapping result was 60.9%, which increased to 87.0% (superior to that of conventional seed-based correlation [47.8%]) after extending to a radius of 1 cm. We established an automatic and practical component identification method for rs-fMRI-based pre-surgical mapping and successfully applied it to brain tumor patients.

43 citations


Cited by
More filters
Journal Article
TL;DR: Definition: To what extent does the study allow us to draw conclusions about a causal effect between two or more constructs?
Abstract: Definition: To what extent does the study allow us to draw conclusions about a causal effect between two or more constructs? Issues: Selection, maturation, history, mortality, testing, regression towrd the mean, selection by maturation, treatment by mortality, treatment by testing, measured treatment variables Increase: Eliminate the threats, above all do experimental manipulations, random assignment, and counterbalancing.

2,006 citations

Journal ArticleDOI
TL;DR: Recurrent tumors can be offered re-intervention, participation in clinical trials, anti-angiogenic agent or local electric field therapy, without an evident impact on survival, while molecular-targeted therapies, immunotherapy and gene therapy are promising tools currently under research.
Abstract: Glioblastoma (GBM) is the most common and lethal tumor of the central nervous system. The natural history of treated GBM remains very poor with 5-year survival rates of 5 %. Survival has not significantly improved over the last decades. Currently, the best that can be offered is a modest 14-month overall median survival in patients undergoing maximum safe resection plus adjuvant chemoradiotherapy. Prognostic factors involved in survival include age, performance status, grade, specific markers (MGMT methylation, mutation of IDH1, IDH2 or TERT, 1p19q codeletion, overexpression of EGFR, etc.) and, likely, the extent of resection. Certain adjuncts to surgery, especially cortical mapping and 5-ALA fluorescence, favor higher rates of gross total resection with apparent positive impact on survival. Recurrent tumors can be offered re-intervention, participation in clinical trials, anti-angiogenic agent or local electric field therapy, without an evident impact on survival. Molecular-targeted therapies, immunotherapy and gene therapy are promising tools currently under research.

434 citations

ComponentDOI
12 Aug 2014-PLOS ONE

394 citations

Journal ArticleDOI
TL;DR: New insights are reviewed into the functional connectivity underlying the sensorimotor, visuospatial, language and sociocognitive systems, and the clinical implications of this paradigmatic shift from localizationism to hodotopy, in the context of brain surgery, neurology, neurorehabilitation and psychiatry, are discussed.
Abstract: At present, direct electrical stimulation (DES) is the only technique that allows directin vivomapping of white matter tracts in humans. In this Review, Hugues Duffau discusses the insights into functional connectivity that have been gained from DES during awake surgery for brain lesions. In addition, the author considers the clinical implications of a paradigmatic shift from a localizationist model to a hodotopical model of cerebral processing. Despite advances in the new science of connectomics, which aims to comprehensively map neural connections at both structural and functional levels, techniques to directly study the function of white matter tracts in vivo in humans have proved elusive. Direct electrical stimulation (DES) mapping of the subcortical fibres offers a unique opportunity to investigate the functional connectivity of the brain. This original method permits real-time anatomo-functional correlations, especially with regard to neural pathways, in awake patients undergoing brain surgery. In this article, the goal is to review new insights, gained from axonal DES, into the functional connectivity underlying the sensorimotor, visuospatial, language and sociocognitive systems. Interactions between these neural networks and multimodal systems, such as working memory, attention, executive functions and consciousness, can also be investigated by axonal stimulation. In this networking model of conation and cognition, brain processing is not conceived as the sum of several subfunctions, but results from the integration and potentiation of parallel—though partially overlapping—subnetworks. This hodotopical account, supported by axonal DES, improves our understanding of neuroplasticity and its limitations. The clinical implications of this paradigmatic shift from localizationism to hodotopy, in the context of brain surgery, neurology, neurorehabilitation and psychiatry, are discussed.

309 citations

Journal ArticleDOI
TL;DR: This paper highlights the common pathophysiology attributes of glioblastoma, surgical options for diagnosis/treatment, current thoughts of extent of resection of tumor, and post-operative (neo)adjuvant treatment.
Abstract: This manuscript discusses the current surgical management of glioblastoma. This paper highlights the common pathophysiology attributes of glioblastoma, surgical options for diagnosis/treatment, current thoughts of extent of resection (EOR) of tumor, and post-operative (neo)adjuvant treatment. Glioblastoma is not a disease that can be cured with surgery alone, however safely performed maximal surgical resection is shown to significantly increase progression free and overall survival while maximizing quality of life. Upon invariable tumor recurrence, re-resection also is shown to impact survival in a select group of patients. As adjuvant therapy continues to improve survival, the role of surgical resection in the treatment of glioblastoma looks to be further defined.

239 citations