Image-Based Search in Radiology: Identification of Brain Tumor Subtypes within Databases using MRI-Based Radiomic Features

BACKGROUND AND PURPOSE: Existing neuroradiology reference materials do not cover the full range of primary brain tumor presentations, and text-based medical image search engines are limited by the lack of consistent structure in radiology reports. To address this, an image-based search approach is introduced here, leveraging an institutional database to find reference MRIs visually similar to presented query cases.

MATERIALS AND METHODS: 295 patients (mean age and SD, 51 ± 20 years) with primary brain tumors who underwent surgical and/or radiotherapeutic treatment between 2000 and 2021 were included in this retrospective study. Semi-automated convolutional neural network-based tumor segmentation was performed, and radiomic features were extracted. The dataset was split into reference and query subsets, and dimensionality reduction was applied to cluster reference cases. Radiomic features extracted from each query case were projected onto the clustered reference cases, and nearest neighbors were retrieved. Retrieval performance was evaluated using mean average precision at k, and the best-performing dimensionality reduction technique was identified. Expert readers independently rated visual similarity using a five-point Likert scale.

RESULTS: t-Distributed Stochastic Neighbor Embedding with six components was the highest-performing dimensionality reduction technique, with mean average precision at 5 ranging from 78% to 100% by tumor type. The top five retrieved reference cases showed high visual similarity Likert scores with corresponding query cases (76% ‘similar’ or ‘very similar’).

CONCLUSIONS: We introduce an image-based search method for exploring historical MR images of primary brain tumors and fetching reference cases closely resembling queried ones. Assessment involving comparison of tumor types and visual similarity Likert scoring by expert neuroradiologists validates the effectiveness of this method.

ABBREVIATIONS: PCA = Principal Component Analysis; t-SNE = t-Distributed Stochastic Neighbor Embedding; UMAP = Uniform Manifold Approximation and Projection; PHATE = Potential of Heat-Diffusion for Affinity-Based Trajectory Embedding; G/A = Glioblastoma and Astrocytoma CNS World Health Organization Grade 4; A/O = Astrocytoma and Oligodendroglioma CNS World Health Organization Grades 2-3; PA = Pilocytic Astrocytoma; MEN = Meningioma; mAP@k = Mean Average Precision at k; CNN = Convolutional Neural Network