Dr. Donald M. O’Rourke Profile

Dr. Donald M. O’Rourke

at Univ. of Pennsylvania

SPIE Involvement:

Author

Publications (2)

SPIE Journal Paper | 30 August 2024

Predicting peritumoral glioblastoma infiltration and subsequent recurrence using deep-learning–based analysis of multi-parametric magnetic resonance imaging

Sunwoo Kwak, Hamed Akbari, Jose Garcia, Suyash Mohan, Yehuda Dicker, Chiharu Sako, Yuji Matsumoto, MacLean Nasrallah, Mahmoud Shalaby, Donald O’Rourke, Russel Shinohara, Fang Liu, Chaitra Badve, Jill Barnholtz-Sloan, Andrew Sloan, Matthew Lee, Rajan Jain, Santiago Cepeda, Arnab Chakravarti, Joshua Palmer, Adam P. Dicker, Gaurav Shukla, Adam Flanders, Wenyin Shi, Graeme Woodworth, Christos Davatzikos

JMI, Vol. 11, Issue 05, 054001, (August 2024) https://doi.org/10.1117/12.10.1117/1.JMI.11.5.054001

KEYWORDS: Tumors, Education and training, Deep learning, Voxels, Magnetic resonance imaging, Data modeling, Tissues, Resection, Brain, Cross validation

Read Abstract +

PurposeGlioblastoma (GBM) is the most common and aggressive primary adult brain tumor. The standard treatment approach is surgical resection to target the enhancing tumor mass, followed by adjuvant chemoradiotherapy. However, malignant cells often extend beyond the enhancing tumor boundaries and infiltrate the peritumoral edema. Traditional supervised machine learning techniques hold potential in predicting tumor infiltration extent but are hindered by the extensive resources needed to generate expertly delineated regions of interest (ROIs) for training models on tissue most and least likely to be infiltrated.ApproachWe developed a method combining expert knowledge and training-based data augmentation to automatically generate numerous training examples, enhancing the accuracy of our model for predicting tumor infiltration through predictive maps. Such maps can be used for targeted supra-total surgical resection and other therapies that might benefit from intensive yet well-targeted treatment of infiltrated tissue. We apply our method to preoperative multi-parametric magnetic resonance imaging (mpMRI) scans from a subset of 229 patients of a multi-institutional consortium (Radiomics Signatures for Precision Diagnostics) and test the model on subsequent scans with pathology-proven recurrence.ResultsLeave-one-site-out cross-validation was used to train and evaluate the tumor infiltration prediction model using initial pre-surgical scans, comparing the generated prediction maps with follow-up mpMRI scans confirming recurrence through post-resection tissue analysis. Performance was measured by voxel-wised odds ratios (ORs) across six institutions: University of Pennsylvania (OR: 9.97), Ohio State University (OR: 14.03), Case Western Reserve University (OR: 8.13), New York University (OR: 16.43), Thomas Jefferson University (OR: 8.22), and Rio Hortega (OR: 19.48).ConclusionsThe proposed model demonstrates that mpMRI analysis using deep learning can predict infiltration in the peri-tumoral brain region for GBM patients without needing to train a model using expert ROI drawings. Results for each institution demonstrate the model’s generalizability and reproducibility.

Proceedings Article | 13 March 2019 Paper

Non-invasive transcriptomic classification of de novo Glioblastoma patients through multivariate quantitative analysis of baseline preoperative multimodal magnetic resonance imaging

Saima Rathore, Hamed Akbari, Spyridon Bakas, Jared Pisapia, Xiao Da, Donald O’Rourke, Christos Davatzikos

Proceedings Volume 10950, 1095043 (2019) https://doi.org/10.1117/12.2512837

KEYWORDS: Tumors, Magnetic resonance imaging, Tissues, Performance modeling, Feature extraction, Quantitative analysis, Brain, Image analysis, Neuroimaging, Machine learning

Read Abstract +

Glioblastoma, the most common primary malignant brain tumor, is genetically diverse and classified into four transcriptomic subtypes, i.e., classical, mesenchymal, proneural, neural. We sought a noninvasive robust quantitative imaging phenomic (QIP) signature associated with this transcriptomic classification of glioblastoma patients, derived from clinically-acquired imaging protocols and without the need for advanced genetic testing. This QIP signature was discovered and evaluated in a retrospective cohort of 112 pathology-proven de novo glioblastoma patients for whom basic preoperative multi-parametric MRI (mpMRI) data (T1, T1-Gd, T2, T2-FLAIR) were available, and compared with the tumor subtype as obtained through an RNA isoform-based classifier. Comprehensive and diverse QIP features capturing intensity distributions, volume, morphology, statistics, tumors’ anatomical location, and texture for each tumor sub-region, were multivariately integrated via support vector machines to construct our QIP signature. The performance/generalizability of the model was evaluated using 5-fold cross-validation. The overall accuracy of the proposed method was estimated equal to 71% for identifying the transcriptomic tumor subtype; 82.14% [AUC:0.82], 75.89% [AUC:0.78], 75.89% [AUC:0.81], and 88.39% [AUC:0.84] for predicting proneural, neural, mesenchymal and classical subtypes, respectively. The obtained QIP signature revealed a macroscopic biological insight of the complex tumor subtypes, including more pronounced presence of tissue with higher water content in neural subtype, larger enhancement component of the tumor in mesenchymal subtype, and overall smaller tumors in classical subtype. Our results indicate that quantitative analysis of imaging features extracted from clinically-acquired mpMRI yields prompt non-invasive biomarkers of the molecular profile of glioblastoma patients, important in influencing surgical decision-making, treatment planning, and assessment of inoperable tumors.

View contact details

UPDATE YOUR PROFILE

Is this your profile? Update it now.

Sign into your SPIE.org account

Don’t have a profile and want one?

Create an account on SPIE.org

Keywords/Phrases

Search In:

Publication Years