Prospects for Theranostics in Neurosurgical Technology: Empowering Confocal Laser Endomicroscopy Diagnostics via Deep Learning

Confocal laser endomicroscopy (CLE) is an advanced optical fluorescence imaging technology that has potential to increase intraoperative precision, extend resection, and tailor surgery for malignant invasive brain tumors because of it subcellular dimension resolution. Manual examination and scrolling through hundreds to thousands of images acquired on the fly during surgery is laborious and time consuming. Therefore, we applied novel methods of deep learning neural networks and computer vision to create a model that could select diagnostic images for the pathologist's or surgeon's review. In this study, we present and assess deep learning models for automatic detection of the diagnostic CLE glioma images using a manually annotated dataset. We explored various training regimes and ensemble modeling effect on power of deep learning predictive models. Histological features from diagnostic CLE images were localized by visualization of shallow and deep neural activations. Results of the interrater comparison experiment confirmed that ensemble of deeply fine tuned models achieved promising agreement with the ground truth, established by the trained neuropathologist. Significant advances in speed and precision of automated diagnostic frame selection would augment the diagnostic potential of CLE, improve operative workflow and integration into brain tumor surgery. Such technology and bioinformatics analytics lend themselves to improved precision, personalization, and theranostics in brain tumor treatment.

READ FULL TEXT

Please sign up or login with your details

Forgot password? Click here to reset