Deep Learning Model Could Accurately Classify Brai
Post# of 373
(Total Views: 264)
Posted On: 01/10/2022 4:51:38 PM
Deep Learning Model Could Accurately Classify Brain Tumors After One MRI Scan
Scientists from the School of Medicine at Washington University have created a deep-learning model that can classify brain tumors using an MRI scan. One of the researchers involved in the study, Satrajit Chakrabarty, stated that the research was the first to identify the absence or presence of a tumor from an MRI scan and directly determine the tumor class.
The most common types of intracranial tumors include acoustic neuroma, pituitary adenoma, meningioma, brain metastases, low-grade glioma and high-grade glioma. Each tumor type was documented via histopathology, which involves the surgical removal of tissue from the site of a cancer and assessing it under a microscope.
Chakrabarty believes that deep-learning and machine-learning approaches that use data from MRI scans could be used to automate the process of detecting and classifying brain tumors. The team of researchers came up with a dataset of 3D MRI scans, which was used to design their deep -earning model. The researchers named their model the convolutional neural network.
The dataset was comprised of data from various institutions, including Washington University, the Cancer Genome Glioblastoma Multiforme, the Brain Tumor Image Segmentation and the Cancer Genome Atlas Low Grade Glioma. They used the scans that were obtained to train their machine-learning model to differentiate between scans that showed tumors and healthy scans, as well as classify each tumor by its type.
After this, the scientists assessed the convolutional neural network’s performance, finding that the model could achieve an accuracy of almost 92% across all tumor classes as well as in healthy individuals, with sensitivities that ranged between 91% to 100%. They also found that the probability that patients with a positive screening test did have the illness ranged between 85% to 100% while the probability that individuals with a negative test didn’t have the ailment ranged between 98% to 100%.
Chakrabarty stated that their findings suggested that deep learning was an approach that held promise for automated brain tumor evaluation and classification. He explained that the convolutional neural network had demonstrated tremendous generalization capabilities on the test data and was highly accurate, even on a diverse dataset. He added that the model eliminated the laborious step of tumor segmentation, which is what is currently used to classify tumors.
The model’s codeveloper Dr. Aristeidis Sotiras, added that the model could also be used in neurological disorders or other types of brain tumors, which potentially offered a path to build up neuroradiology workflow.
The research was reported in “Radiology: Artificial Intelligence.”
Such studies demonstrating how to make it easier to obtain a positive diagnosis of brain tumors in a minimally invasive way provide a perfect complement to the work of companies such as CNS Pharmaceuticals Inc. (NASDAQ: CNSP) that seek to develop novel formulations targeting brain cancers in the continuum of care for cancer patients.
NOTE TO INVESTORS: The latest news and updates relating to CNS Pharmaceuticals Inc. (NASDAQ: CNSP) are available in the company’s newsroom at https://ibn.fm/CNSP
Please see full terms of use and disclaimers on the BioMedWire website applicable to all content provided by BMW, wherever published or re-published: http://BMW.fm/Disclaimer
Scientists from the School of Medicine at Washington University have created a deep-learning model that can classify brain tumors using an MRI scan. One of the researchers involved in the study, Satrajit Chakrabarty, stated that the research was the first to identify the absence or presence of a tumor from an MRI scan and directly determine the tumor class.
The most common types of intracranial tumors include acoustic neuroma, pituitary adenoma, meningioma, brain metastases, low-grade glioma and high-grade glioma. Each tumor type was documented via histopathology, which involves the surgical removal of tissue from the site of a cancer and assessing it under a microscope.
Chakrabarty believes that deep-learning and machine-learning approaches that use data from MRI scans could be used to automate the process of detecting and classifying brain tumors. The team of researchers came up with a dataset of 3D MRI scans, which was used to design their deep -earning model. The researchers named their model the convolutional neural network.
The dataset was comprised of data from various institutions, including Washington University, the Cancer Genome Glioblastoma Multiforme, the Brain Tumor Image Segmentation and the Cancer Genome Atlas Low Grade Glioma. They used the scans that were obtained to train their machine-learning model to differentiate between scans that showed tumors and healthy scans, as well as classify each tumor by its type.
After this, the scientists assessed the convolutional neural network’s performance, finding that the model could achieve an accuracy of almost 92% across all tumor classes as well as in healthy individuals, with sensitivities that ranged between 91% to 100%. They also found that the probability that patients with a positive screening test did have the illness ranged between 85% to 100% while the probability that individuals with a negative test didn’t have the ailment ranged between 98% to 100%.
Chakrabarty stated that their findings suggested that deep learning was an approach that held promise for automated brain tumor evaluation and classification. He explained that the convolutional neural network had demonstrated tremendous generalization capabilities on the test data and was highly accurate, even on a diverse dataset. He added that the model eliminated the laborious step of tumor segmentation, which is what is currently used to classify tumors.
The model’s codeveloper Dr. Aristeidis Sotiras, added that the model could also be used in neurological disorders or other types of brain tumors, which potentially offered a path to build up neuroradiology workflow.
The research was reported in “Radiology: Artificial Intelligence.”
Such studies demonstrating how to make it easier to obtain a positive diagnosis of brain tumors in a minimally invasive way provide a perfect complement to the work of companies such as CNS Pharmaceuticals Inc. (NASDAQ: CNSP) that seek to develop novel formulations targeting brain cancers in the continuum of care for cancer patients.
NOTE TO INVESTORS: The latest news and updates relating to CNS Pharmaceuticals Inc. (NASDAQ: CNSP) are available in the company’s newsroom at https://ibn.fm/CNSP
Please see full terms of use and disclaimers on the BioMedWire website applicable to all content provided by BMW, wherever published or re-published: http://BMW.fm/Disclaimer
(0)
(0)