Search bioRxiv⌕ Search

Biology subjects

Murugesan, G. K.

Publications and source records attributed to Murugesan, G. K..

9 recordsLinked to original sources
bioRxiv2023

Automatic Abdominal Multi Organ Segmentationusing Residual UNet

Automated segmentation of abdominal organs plays an important role in supporting computer-assisted diagnosis, radiotherapy, biomarker extraction, surgery navigation, and treatment planning. Segmenting multiple abdominal organs using a single algorithm would improve model development efficiency and accelerate model deployment into clinical workflows. To achieve broadly generalized performance, we trained a residual UNet using 500 CT/MRI scans collected from multi-center, multi-vendor, multi-phase, multi-disease patients, each with voxel-level annotation of 15 abdominal organs. Using the model trained on multimodality (CT/MRI), we achieved an average dice of 0.8990 in the held-out test dataset with only CT scans (N=100). An average dice of 0.8948 was achieved in the held-out test dataset with both CT and MRI scans (N=120. Our results demonstrate broad generalization of the model.

bioengineering↗
bioRxiv2023

Automatic Whole Body FDG PET/CT Lesion Segmentation using Residual UNet and Adaptive Ensemble

Multimodal Positron Emission Tomography/Computed Tomography (PET/CT) plays a key role in the diagnosis, staging, restaging, treatment response assessment, and radiotherapy planning of malignant tumors. The complementary nature of high-resolution anatomic CT and high sensitivity/specificity molecular PET imaging provides accurate assessment of disease status [14] In oncology, 18-fluorodeoxyglucose (FDG) PET/CT is the most widely used method to identify and analyze metabolically active tumors. In particular, FDG uptake allows for more accurate detection of both nodal and distant forms of metastatic disease. Accurate quantification and staging of tumors is the most important prognostic factor for predicting the survival of patients and for designing personalized patient management plans. [8,3] Analyzing PET/CT quantitatively by experienced medical imaging experts/radiologists is timeconsuming and error-prone. Automated quantitative analysis by deep learning algorithms to segment tumor lesions will enable accurate feature extraction, tumor staging, radiotherapy planning, and treatment response assessment. The AutoPET Challenge 2022 provided an opensource platform to develop and benchmark deep learning models for automated PET lesion segmentation by providing large open-source wholebody FDG-PET/CT data. Using the multimodal PET/CT data from 900 subjects with 1014 studies provided by the AutoPET MICCAI 2022 Challenge, we applied fivefold cross-validation on residual UNETs to automatically segment lesions. We then utilized the output from adaptive ensemble highly contributive models as the final segmentation. Our method achieved a 10th ranking with a dice score of 0.5541 in the heldout test dataset (N=150 studies).

bioengineering↗
bioRxiv2020

A Novel Fully Automated MRI-Based Deep Learning Method For Classification Of 1p/19q Co-Deletion Status In Brain Gliomas

BackgroundOne of the most important recent discoveries in brain glioma biology has been the identification of the isocitrate dehydrogenase (IDH) mutation and 1p/19q co-deletion status as markers for therapy and prognosis. 1p/19q co-deletion is the defining genomic marker for oligodendrogliomas and confers a better prognosis and treatment response than gliomas without it. Our group has previously developed a highly accurate deep-learning network for determining IDH mutation status using T2-weighted MRI only. The purpose of this study was to develop a similar 1p/19q deep-learning classification network. MethodsMulti-parametric brain MRI and corresponding genomic information were obtained for 368 subjects from The Cancer Imaging Archive (TCIA) and The Cancer Genome Atlas (TCGA). 1p/19 co-deletions were present in 130 subjects. 238 subjects were non co-deleted. A T2w image only network (1p/19q-net) was developed to perform 1p/19q co-deletion status classification and simultaneous single-label tumor segmentation using 3D-Dense-UNets. Threefold cross-validation was performed to generalize the network performance. ROC analysis was also performed. Dice-scores were computed to determine tumor segmentation accuracy. Results1p/19q-net demonstrated a mean cross validation accuracy of 93.46% across the 3 folds (93.4%, 94.35%, and 92.62%, standard dev=0.8) in predicting 1p/19q co-deletion status with a sensitivity and specificity of 0.90 {+/-}0.003 and 0.95 {+/-}0.01, respectively and a mean AUC of 0.95 {+/-}0.01. The whole tumor segmentation mean Dice-score was 0.80 {+/-} 0.007. ConclusionWe demonstrate high 1p/19q co-deletion classification accuracy using only T2-weighted MR images. This represents an important milestone toward using MRI to predict glioma histology, prognosis, and response to treatment. Keypoints1. 1p/19 co-deletion status is an important genetic marker for gliomas. 2. We developed a non-invasive, MRI based, highly accurate deep-learning method for the determination of 1p/19q co-deletion status that only utilizes T2 weighted MR images IMPORTANCE OF THE STUDYOne of the most important recent discoveries in brain glioma biology has been the identification of the isocitrate dehydrogenase (IDH) mutation and 1p/19q co-deletion status as markers for therapy and prognosis. 1p/19q co-deletion is the defining genomic marker for oligodendrogliomas and confers a better prognosis and treatment response than gliomas without it. Currently, the only reliable way to determine 1p/19q mutation status requires analysis of glioma tissue obtained either via an invasive brain biopsy or following open surgical resection. The ability to non-invasively determine 1p/19q co-deletion status has significant implications in determining therapy and predicting prognosis. We developed a highly accurate, deep learning network that utilizes only T2-weighted MR images and outperforms previously published imagebased methods. The high classification accuracy of our T2w image only network (1p/19q-net) in predicting 1p/19q co-deletion status marks an important step towards image-based stratification of brain gliomas. Imminent clinical translation is feasible because T2-weighted MR imaging is widely available and routinely performed in the assessment of gliomas.

neuroscience↗
bioRxiv2020

Brain Tumor IDH, 1p/19q, and MGMT Molecular Classification Using MRI-based Deep Learning: Effect of Motion and Motion Correction

Deep learning has shown promise for predicting glioma molecular profiles using MR images. Before clinical implementation, ensuring robustness to real-world problems, such as patient motion, is crucial. We sought to evaluate the effects of motion artifact on glioma marker classifier performance and develop a deep learning motion correction network to restore classification accuracies. T2w images and molecular information were retrieved from the TCIA and TCGA databases. Three-fold cross-validation was used to train and test the motion correction network on artifact-corrupted images. We then compared the performance of three glioma marker classifiers (IDH mutation, 1p/19q codeletion, and MGMT methylation) using motion-corrupted and motion-corrected images. Glioma marker classifier performance decreased markedly with increasing motion corruption. Applying motion correction effectively restored classification accuracy for even the most motion-corrupted images. For IDH classification, an accuracy of 99% was achieved, representing a new benchmark in non-invasive image-based IDH classification and exceeding the original performance of the network. Robust motion correction can enable high accuracy in deep learning MRI-based molecular marker classification rivaling tissue-based characterization. STATEMENT OF SIGNIFICANCEDeep learning networks have shown promise for predicting molecular profiles of gliomas using MR images. We demonstrate that patient motion artifact, which is frequently encountered in the clinic, can significantly impair the performance of these algorithms. The application of robust motion correction algorithms can restore the performance of these networks, rivaling tissue-based characterization.

neuroscience↗
bioRxiv2020

MRI-BASED DEEP LEARNING METHOD FOR DETERMINING METHYLATION STATUS OF THE O6 METHYLGUANINE DNA METHYLTRANSFERASE PROMOTER OUTPERFORMS TISSUE BASED METHODS IN BRAIN GLIOMAS

PURPOSEMethylation of the O6-Methylguanine-DNA Methyltransferase (MGMT) promoter results in epigenetic silencing of the MGMT enzyme and confers an improved prognosis and treatment response in gliomas. The purpose of this study was to develop a deep-learning network for determining the methylation status of the MGMT Promoter in gliomas using T2-w magnetic resonance images only. METHODSBrain MRI and corresponding genomic information were obtained for 247 subjects from The Cancer Imaging Archive (TCIA) and The Cancer Genome Atlas (TCGA). 163 subjects had a methylated MGMT promoter. A T2-w image only network (MGMT-net) was developed to determine MGMT promoter methylation status and simultaneous single label tumor segmentation. The network was trained using 3D-Dense-UNets. Three-fold cross-validation was performed to generalize the networks performance. Dice-scores were computed to determine tumor segmentation accuracy. RESULTSMGMT-net demonstrated a mean cross validation accuracy of 94.73% across the 3 folds (95.12%, 93.98%, and 95.12%, standard dev=0.66) in predicting MGMT methylation status with a sensitivity and specificity of 96.31% {+/-}0.04 and 91.66% {+/-}2.06, respectively and a mean AUC of 0.93 {+/-}0.01. The whole tumor segmentation mean Dice-score was 0.82 {+/-} 0.008. CONCLUSIONWe demonstrate high classification accuracy in predicting the methylation status of the MGMT promoter using only T2-w MR images that surpasses the sensitivity, specificity, and accuracy of invasive histological methods such as pyrosequencing, methylation-specific PCR, and immunofluorescence methods. This represents an important milestone toward using MRI to predict glioma histology, prognosis, and response to treatment.

neuroscience↗
bioRxiv2019

BrainNET: Inference of brain network topology using Machine Learning

ObjectiveTo develop a new fMRI network inference method, BrainNET, that utilizes an efficient machine learning algorithm to quantify contributions of various regions of interests (ROIs) in the brain to a specific ROI. MethodsBrainNET is based on Extremely Randomized Trees (ERT) to estimate network topology from fMRI data and modified to generate an adjacency matrix representing brain network topology, without reliance on arbitrary thresholds. Open source simulated fMRI data of fifty subjects in twenty-eight different simulations under various confounding conditions with known ground truth was used to validate the method. Performance was compared with correlation and partial correlation (PC). The real-world performance was then evaluated in a publicly available Attention-deficit/hyperactivity disorder (ADHD) dataset including 134 Typically Developing Children (mean age: 12.03, males: 83), 75 ADHD Inattentive (mean age: 11.46, males: 56) and 93 ADHD Combined (mean age: 11.86, males: 77) subjects. Network topologies in ADHD were inferred using BrainNET, correlation, and PC. Graph metrics were extracted to determine differences between the ADHD groups. ResultsBrainNET demonstrated excellent performance across all simulations and varying confounders in identifying true presence of connections. In the ADHD dataset, BrainNET was able to identify significant changes (p< 0.05) in graph metrics between groups. No significant changes in graph metrics between ADHD groups was identified using correlation and PC.

bioengineering↗
bioRxiv2019

Multidimensional and Multiresolution Ensemble Networks for Brain Tumor Segmentation

In this work, we developed multiple 2D and 3D segmentation models with multiresolution input to segment brain tumor components, and then ensembled them to obtain robust segmentation maps. This reduced overfitting and resulted in a more generalized model. Multiparametric MR images of 335 subjects from BRATS 2019 challenge were used for training the models. Further, we tested a classical machine learning algorithm (xgboost) with features extracted from the segmentation maps to classify subject survival range. Preliminary results on the BRATS 2019 validation dataset demonstrasted this method can achieve excellent performance with DICE scores of 0.898, 0.784, 0.779 for whole tumor, tumor core and enhancing tumor respectively and accuracy 34.5 % for survuval prediction.

neuroscience↗
bioRxiv2019

Fully Automated Brain Tumor Segmentation and Survival Prediction of Gliomas using Deep Learning and MRI

Tumor segmentation of magnetic resonance (MR) images is a critical step in providing objective measures of predicting aggressiveness and response to therapy in gliomas. It has valuable applications in diagnosis, monitoring, and treatment planning of brain tumors. The purpose of this work was to develop a fully automated deep learning method for brain tumor segmentation and survival prediction. Well curated brain tumor cases with multi-parametric MR Images from the BraTS2019 dataset were used. A three-group framework was implemented, with each group consisting of three 3D-Dense-UNets to segment whole tumor (WT), tumor core (TC) and enhancing tumor (ET). This method was implemented to decompose the complex multi-class segmentation problem into individual binary segmentation problems for each sub-component. Each group was trained using different approaches and loss functions. The output segmentations of a particular label from their respective networks from the 3 groups were ensembled and post-processed. For survival analysis, a linear regression model based on imaging texture features and wavelet texture features extracted from each of the segmented components was implemented. The networks were tested on the BraTS2019 validation dataset including 125 cases for the brain tumor segmentation task and 29 cases for the survival prediction task. The segmentation networks achieved average dice scores of 0.901, 0.844 and 0.801 for WT, TC and ET respectively. The survival prediction network achieved an accuracy score of 0.55 and mean squared error (MSE) of 119244. This method could be implemented as a robust tool to assist clinicians in primary brain tumor management and follow-up.

neuroscience↗