Abstract
Deep learning (DL) is a widely applied mathematical modeling technique. Classically, DL models utilize large volumes of training data, which are not available in many healthcare contexts. For patients with brain tumors, non-invasive diagnosis would represent a substantial clinical advance, potentially sparing patients from the risks associated with surgical intervention on the brain. Such an approach will depend upon highly accurate models built using the limited datasets that are available. Herein, we present a novel genetic algorithm (GA) that identifies optimal architecture parameters using feature embeddings from state-of-the-art image classification networks to identify the pediatric brain tumor, adamantinomatous craniopharyngioma (ACP). We optimized classification models for preoperative Computed Tomography (CT), Magnetic Resonance Imaging (MRI), and combined CT and MRI datasets with demonstrated test accuracies of 85.3%, 83.3%, and 87.8%, respectively. Notably, our GA improved baseline model performance by up to 38%. This work advances DL and its applications within healthcare by identifying optimized networks in small-scale data contexts. The proposed system is easily implementable and scalable for non-invasive computer-aided diagnosis, even for uncommon diseases.
Highlights
Deep learning (DL) is a widely applied mathematical modeling technique
Using twelve state-of-the-art networks that have publicly available deep learning models from the TensorFlow Hub library[11] trained on the ImageNet ILSVRC dataset (Fig. 1A)[4,12,13,14,15,16,17], we generated feature embedding vectors to be used in model training (Fig. 1B)
Using the pediatric brain tumor Adamantinomatous Craniopharyngioma as an example of a clinical entity with a small available dataset, we enhance the performance of a baseline Convolutional Neural Network using a series of optimization methodologies, including Transfer Learning, Data Augmentation, and Image Obfuscation
Summary
Deep learning (DL) is a widely applied mathematical modeling technique. Classically, DL models utilize large volumes of training data, which are not available in many healthcare contexts. CNNs and other variants have had great success in tasks such as image object recognition; speech recognition, translation, and generation; and medical diagnostics, genetics, and drug d iscovery[3] These applications have achieved remarkable success, to some extent by leveraging very large amounts of labeled training data. The success of TL has led to the development of publicly available pre-trained models derived from top ILSVRC solutions By using these pre-trained networks to generate feature embeddings for our dataset of interest, we enable our classifier to have access to the pattern recognition capabilities of these state-of-the-art architectures. Another technique commonly applied to image classification problems is data augmentation. TANDA was reported to yield synthetic data in which feature representations are distributed and invariant, helping disentangle the factors of variation between the two c lasses[7]
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.