Multi-task learning for concurrent survival prediction and semi-supervised segmentation of gliomas in brain MRI

Article Computer Science, Interdisciplinary Applications

A Fully Automated Multimodal MRI-Based Multi-Task Learning for Glioma Segmentation and IDH Genotyping

Jianhong Cheng et al.

Summary: Accurate prediction of isocitrate dehydrogenase (IDH) mutation and glioma segmentation is crucial for computer-aided diagnosis using preoperative multimodal magnetic resonance imaging (MRI). Existing methods often address these tasks separately and overlook the correlation between them. Furthermore, the limited availability of IDH mutation data hampers modeling. In this paper, we propose a fully automated multimodal MRI-based multi-task learning framework that simultaneously performs glioma segmentation and IDH genotyping. Our framework combines a hybrid CNN-Transformer encoder with a multi-scale classifier to extract shared spatial and global information for segmentation and genotyping. A multi-task learning loss is designed to balance the two tasks, while an uncertainty-aware pseudo-label selection method improves the accuracy of IDH genotyping with semi-supervised learning. Experimental results on a multi-institutional public dataset demonstrate the promising performance of our proposed method compared to single-task learning and existing state-of-the-art methods. The availability of unlabeled data further enhances the performance of glioma segmentation and IDH genotyping. The source codes of our framework are publicly available.

IEEE TRANSACTIONS ON MEDICAL IMAGING (2022)

Add to Collection

Article Computer Science, Interdisciplinary Applications

SimCVD: Simple Contrastive Voxel-Wise Representation Distillation for Semi-Supervised Medical Image Segmentation

Chenyu You et al.

Summary: Automated segmentation in medical image analysis is a challenging task. This work introduces a simple contrastive distillation framework, SimCVD, that improves the accuracy of image segmentation through unsupervised training and structural distillation. Experimental results on two datasets demonstrate the effectiveness of SimCVD.

IEEE TRANSACTIONS ON MEDICAL IMAGING (2022)

Add to Collection

Article Computer Science, Artificial Intelligence

Multi-Task Learning for Dense Prediction Tasks: A Survey

Simon Vandenhende et al.

Summary: With the advent of deep learning, dense prediction tasks have significantly improved. Recent multi-task learning techniques have shown promising results by jointly tackling multiple tasks. This survey provides a comprehensive view on state-of-the-art deep learning approaches for multi-task learning in computer vision, with a focus on dense prediction tasks.

IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE (2022)

Add to Collection

Proceedings Paper Computer Science, Artificial Intelligence

An Empirical Study of Training End-to-End Vision-and-Language Transformers

Zi-Yi Dou et al.

Summary: In this paper, the authors propose the Meter framework, a multimodal end-to-end Transformer framework, for designing and pre-training fully Transformer-based vision-and-language models. The experiments show that the Meter framework achieves good performance on the VQAv2 dataset and outperforms previous models.

2022 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR 2022) (2022)

Add to Collection

Article Medicine, General & Internal

Deep learning features from diffusion tensor imaging improve glioma stratification and identify risk groups with distinct molecular pathway activities

Jing Yan et al.

Summary: A deep learning signature (DLS) from diffusion tensor imaging (DTI) was developed for predicting overall survival in patients with infiltrative gliomas, with five key biological pathways identified. The DLS was associated with survival, independent predictor, and improved survival prediction when incorporated into existing risk system. Therapies targeting neuron-to-brain tumor synaptic communication may be more effective in high-risk glioma defined by DTI-derived DLS.

EBIOMEDICINE (2021)

Add to Collection

Review Radiology, Nuclear Medicine & Medical Imaging

Performance of machine learning algorithms for glioma segmentation of brain MRI: a systematic literature review and meta-analysis

Evi J. van Kempen et al.

Summary: The study provides an overview and meta-analysis of various machine learning algorithms for automated segmentation of gliomas, showing promising accuracy. However, high heterogeneity between studies and publication bias were observed as potential challenges.

EUROPEAN RADIOLOGY (2021)

Add to Collection

Review Oncology

The 2021 WHO Classification of Tumors of the Central Nervous System: a summary

David N. Louis et al.

Summary: The fifth edition of the WHO Classification of Tumors of the Central Nervous System integrates molecular diagnostics into CNS tumor classification, introduces different approaches to tumor nomenclature and grading, and emphasizes the importance of integrated diagnoses and layered reports.

NEURO-ONCOLOGY (2021)

Add to Collection

Article Oncology