# Imbalance-Aware Self-Supervised Learning for 3D Radiomic Representations Source: [Imbalance-Aware Self-Supervised Learning for 3D Radiomic Representations]() Collection index: [llms.txt]() Data updated: 2026-01-04T05:04:57.981Z Authors: Hongwei Li, Fei-Fei Xue, Krishna Chaitanya, Shengda Luo, Ivan Ezhov, Benedikt Wiestler, Jianguo Zhang, Bjoern Menze Published: 2021-03-06T18:17:03.000Z ## Abstract Radiomic representations can quantify properties of regions of interest in medical image data. Classically, they account for pre-defined statistics of shape, texture, and other low-level image features. Alternatively, deep learning-based representations are derived from supervised learning but require expensive annotations from experts and often suffer from overfitting and data imbalance issues. In this work, we address the challenge of learning representations of 3D medical images for an effective quantification under data imbalance. We propose a \\emph{self-supervised} representation learning framework to learn high-level features of 3D volumes as a complement to existing radiomics features. Specifically, we demonstrate how to learn image representations in a self-supervised fashion using a 3D Siamese network. More importantly, we deal with data imbalance by exploiting two unsupervised strategies: a) sample re-weighting, and b) balancing the composition of training batches. When combining our learned self-supervised feature with traditional radiomics, we show significant improvement in brain tumor classification and lung cancer staging tasks covering MRI and CT imaging modalities. ## Recorded benchmark results A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. - [Brain Tumor Classification on BraTS full labels (cross-validation)](): Brain tumor classification; BraTS full labels (cross-validation); Sensitivity: 92. - [Lung Cancer Staging on Lung cancer staging full labels (cross-validation)](): Lung cancer staging; Lung cancer staging full labels (cross-validation); Overall Accuracy: 53.8. - [Lung cancer staging on Lung cancer staging 50% labels (cross-validation)](): Lung cancer staging; Lung cancer staging 50% labels (cross-validation); Overall Accuracy: 51.9. - [Brain Tumor Classification on BraTS 50% labels (val)](): Brain tumor classification; BraTS 50% labels (val); Sensitivity: 84.8. ## Original sources and implementation [Original paper]() [PDF]() [Implementation]()