Diseased thyroid tissue classification in OCT images using deep learning: Towards surgical decision support.

Intraoperative guidance tools for thyroid surgery based on optical coherence tomography (OCT) could aid distinguish between normal and diseased tissue. However, OCT images are difficult to interpret, thus, real-time automatic analysis could support the clinical decision-making. In this study, several deep learning models were investigated for thyroid disease classification on 2D and 3D OCT data obtained from ex vivo specimens of 22 patients undergoing surgery and diagnosed with several thyroid pathologies. Additionally, two open-access datasets were used to evaluate the custom models. On the thyroid dataset, the best performance was achieved by the 3D vision transformer model with a Matthew's correlation coefficient (MCC) of 0.79 (accuracy = 0.90) for the normal-versus-abnormal classification. On the open-access datasets, the custom models achieved the best performance (MCC > 0.88, accuracy > 0.96). Results obtained for the normal-versus-abnormal classification suggest OCT, complemented with deep learning-based analysis, as a tool for real-time automatic diseased tissue identification in thyroid surgery.

Inflation of test accuracy due to data leakage in deep learning-based classification of OCT images.

In the application of deep learning on optical coherence tomography (OCT) data, it is common to train classification networks using 2D images originating from volumetric data. Given the micrometer resolution of OCT systems, consecutive images are often very similar in both visible structures and noise. Thus, an inappropriate data split can result in overlap between the training and testing sets, with a large portion of the literature overlooking this aspect. In this study, the effect of improper dataset splitting on model evaluation is demonstrated for three classification tasks using three OCT open-access datasets extensively used, Kermany's and Srinivasan's ophthalmology datasets, and AIIMS breast tissue dataset. Results show that the classification performance is inflated by 0.07 up to 0.43 in terms of Matthews Correlation Coefficient (accuracy: 5% to 30%) for models tested on datasets with improper splitting, highlighting the considerable effect of dataset handling on model evaluation. This study intends to raise awareness on the importance of dataset splitting given the increased research interest in implementing deep learning on OCT data.

Does Anatomical Contextual Information Improve 3D U-Net-Based Brain Tumor Segmentation?

Effective, robust, and automatic tools for brain tumor segmentation are needed for the extraction of information useful in treatment planning. Recently, convolutional neural networks have shown remarkable performance in the identification of tumor regions in magnetic resonance (MR) images. Context-aware artificial intelligence is an emerging concept for the development of deep learning applications for computer-aided medical image analysis. A large portion of the current research is devoted to the development of new network architectures to improve segmentation accuracy by using context-aware mechanisms. In this work, it is investigated whether or not the addition of contextual information from the brain anatomy in the form of white matter (WM), gray matter (GM), and cerebrospinal fluid (CSF) masks and probability maps improves U-Net-based brain tumor segmentation. The BraTS2020 dataset was used to train and test two standard 3D U-Net (nnU-Net) models that, in addition to the conventional MR image modalities, used the anatomical contextual information as extra channels in the form of binary masks (CIM) or probability maps (CIP). For comparison, a baseline model (BLM) that only used the conventional MR image modalities was also trained. The impact of adding contextual information was investigated in terms of overall segmentation accuracy, model training time, domain generalization, and compensation for fewer MR modalities available for each subject. Median (mean) Dice scores of 90.2 (81.9), 90.2 (81.9), and 90.0 (82.1) were obtained on the official BraTS2020 validation dataset (125 subjects) for BLM, CIM, and CIP, respectively. Results show that there is no statistically significant difference when comparing Dice scores between the baseline model and the contextual information models (p > 0.05), even when comparing performances for high and low grade tumors independently. In a few low grade cases where improvement was seen, the number of false positives was reduced. Moreover, no improvements were found when considering model training time or domain generalization. Only in the case of compensation for fewer MR modalities available for each subject did the addition of anatomical contextual information significantly improve (p < 0.05) the segmentation of the whole tumor. In conclusion, there is no overall significant improvement in segmentation performance when using anatomical contextual information in the form of either binary WM, GM, and CSF masks or probability maps as extra channels.