Deep learning and transfer learning for brain tumor detection and classification
Files
First author draft
Date
2023-04-11
Authors
Rustom, Faris
Parva, Pedram
Ogmen, Haluk
Yazdanbakhsh, Arash
Version
First author draft
OA Version
Citation
F. Rustom, P. Parva, H. Ogmen, A. Yazdanbakhsh. 2023. "Deep Learning and Transfer Learning for Brain Tumor Detection and Classification" https://doi.org/10.1101/2023.04.10.536226
Abstract
Convolutional neural networks (CNNs) are powerful tools that can be trained on image classification tasks and share many structural and functional similarities with biological visual systems and mechanisms of learning. In addition to serving as a model of biological systems, CNNs possess the convenient feature of transfer learning where a network trained on one task may be repurposed for training on another, potentially unrelated, task. In this retrospective study of public domain MRI data, we investigate the ability of neural networks to be trained on brain cancer imaging data while introducing a unique camouflage animal detection transfer learning step as a means of enhancing the network’s tumor detection ability. Training on glioma, meningioma, and healthy brain MRI data, both T1- and T2-weighted, we demonstrate the potential success of this training strategy for improving neural network classification accuracy both quantitatively with accuracy metrics and qualitatively with feature space analysis of the internal states of trained networks. In addition to animal transfer learning, similar improvements were noted as a result of transfer learning between MRI sequences, specifically from T1 to T2 data. Image sensitivity functions further this investigation by allowing us to visualize the most salient image regions from a network’s perspective while learning. Such methods demonstrate that the networks not only ‘look’ at the tumor itself when deciding, but also at the impact on the surrounding tissue in terms of compressions and midline shifts. These results suggest an approach to brain tumor MRIs that is comparatively similar to that of trained radiologists while also exhibiting a high sensitivity to subtle structural changes resulting from the presence of a tumor.
Summary Statement
Convolutional neural networks can be trained with transfer learning to accurately detect and classify brain tumor MRIs while simultaneously demonstrating an approach to imaging data resembling that of trained radiologists.
Key Results
-Transfer learning from camouflage animal recognition neural networks, and between T1-T2 MRI sequence, improved brain tumor detection and classification accuracy in retrospective study.
-Trained neural networks correctly identified oligoastrocytomas as an intermediate of astrocytomas and oligoastrocytomas, as visualized by feature space mapping.
-Image sensitivity maps of networks suggested a high sensitivity to midline shifts, tissue compressions, and symmetrical comparisons of brain regions for accurate tumor classification.
Description
License
The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY 4.0 International license.