An estimated 18,000 people in the United States will die of brain and spinal cord tumors in 2020. To help doctors differentiate between the severity of brain cancer, the president of neurosurgery at Yale School, Drs. There is an international team of researchers led by Murat Guyal.
The Nixdorff-German Professor of Medicine, and Neurosurgery, constructed a machine learning model that uses complex mathematics to visualize different types of brain tumors in the brain. The model is designed to “learn” from this collected data to make predictions and help doctors diagnose the stage of brain cancer faster and more accurately.
To test their artificial learning method, the team used 229 patients with brain tumors with a spectrum likely to become incurable with low-grade gliomas, which are relatively slow-growing tumors that cause brain damage. Are produced from glial cells – a highly aggressive counterpart to glioblastoma gliomas.
“Our machine learning models, used to differentiate the types of tumors, were very accurate,” Dr. Said Hang Cao, a medical student at Xiangya Hospital who works with Gannell and the lead author of the study published in European Radiology.
The researchers compiled data from a public tumor machine resonance imaging (MRI) database, called The Cancer Imaging Archive. Board-certified neuro-radiologists then identified and selected glioma cases, which the researchers used to model them.
The team found that there is a significant difference in how cancers look, their volume and their locations in different areas of the brain. When taken together, the model can predict which tumors were low-grade gliomas or glioblastomas with a high degree of accuracy.
The time frame for using such a model in a clinical setting is not known at this time. Although it will now be possible to implement this as a stand-alone evaluation, the procedure is not yet integrated into the clinical evaluation of the patient. A clear set of standards should be established by the scientific community and then embraced by manufacturers of software and hardware used in radiology departments.
“This work is an important example of our understanding of brain tumors and the collaborative, multi-disciplinary effort we use to advance the field and provide the best care to patients with brain tumors,” co-author Dr. Jennifer Moliterino, Assistant Professor at the Yale School of Medicine and Clinical Program Leader of the Brain Tumor Program in Neurosurgery.
To establish a quantitative MR model that uses clinically relevant features of tumor location and tumor volume to differentiate lower grade glioma (LRGG, grade II and III) and glioblastoma (GBM, grade IV) is.
We extracted tumor location and tumor volume (tumor-enhancing, non-enhancing tumors, peritumor edema) in 229 The Cancer Genome Atlas (TCGA) -LGG and TCGA-GBM cases.
Through two sampling strategies, i.e., institution-based sampling and random sampling (10 times, 70% training set vs. 30% validation set), LASSO (least complete shrinkage and selection operator) regression and nine-machine learning method Repeat-based model set up and evaluation.
The principal component analysis of the 229 TCGA – LGG and TCGA – GBM cases suggested that the LRGG and GBM cases could be differentiated by the extracted features.
For the nine machine learning methods, the stack modeling and support vector machine achieved the highest performance (institution-based sample validation set, AUC> 0.900, classifier accuracy> 0.790; repeat random sampling, average validation set AUC> 0.930, classifier accuracy 0.8 0.8).
For the LASSO method, regression models based on tumor frontal lobe percentage and volume of enhancer and non-enhancing tumors achieved the highest performance (institution-based sample validation validation, AUC 0.909, classifier accuracy 0.830). The formula for the best performance of the LASSO model was established.
Computer-generated, clinically meaningful MRI characteristics of tumor location and component versions result in the separation of lower grade gliomas and glioblastomas with high performance (validation set AUC> 0.900, classifier accuracy> 0.790).