PRIOR: Prototype Representation Joint Learning from Medical Images and Reports

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Cheng, Pujin, Lin, Li, Lyu, Junyan, Huang, Yijin, Luo, Wenhan, Tang, Xiaoying
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914708526727168
author Cheng, Pujin
Lin, Li
Lyu, Junyan
Huang, Yijin
Luo, Wenhan
Tang, Xiaoying
author_facet Cheng, Pujin
Lin, Li
Lyu, Junyan
Huang, Yijin
Luo, Wenhan
Tang, Xiaoying
contents Contrastive learning based vision-language joint pre-training has emerged as a successful representation learning strategy. In this paper, we present a prototype representation learning framework incorporating both global and local alignment between medical images and reports. In contrast to standard global multi-modality alignment methods, we employ a local alignment module for fine-grained representation. Furthermore, a cross-modality conditional reconstruction module is designed to interchange information across modalities in the training phase by reconstructing masked images and reports. For reconstructing long reports, a sentence-wise prototype memory bank is constructed, enabling the network to focus on low-level localized visual and high-level clinical linguistic features. Additionally, a non-auto-regressive generation paradigm is proposed for reconstructing non-sequential reports. Experimental results on five downstream tasks, including supervised classification, zero-shot classification, image-to-text retrieval, semantic segmentation, and object detection, show the proposed method outperforms other state-of-the-art methods across multiple datasets and under different dataset size settings. The code is available at https://github.com/QtacierP/PRIOR.
format Preprint
id arxiv_https___arxiv_org_abs_2307_12577
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle PRIOR: Prototype Representation Joint Learning from Medical Images and Reports
Cheng, Pujin
Lin, Li
Lyu, Junyan
Huang, Yijin
Luo, Wenhan
Tang, Xiaoying
Computer Vision and Pattern Recognition
Contrastive learning based vision-language joint pre-training has emerged as a successful representation learning strategy. In this paper, we present a prototype representation learning framework incorporating both global and local alignment between medical images and reports. In contrast to standard global multi-modality alignment methods, we employ a local alignment module for fine-grained representation. Furthermore, a cross-modality conditional reconstruction module is designed to interchange information across modalities in the training phase by reconstructing masked images and reports. For reconstructing long reports, a sentence-wise prototype memory bank is constructed, enabling the network to focus on low-level localized visual and high-level clinical linguistic features. Additionally, a non-auto-regressive generation paradigm is proposed for reconstructing non-sequential reports. Experimental results on five downstream tasks, including supervised classification, zero-shot classification, image-to-text retrieval, semantic segmentation, and object detection, show the proposed method outperforms other state-of-the-art methods across multiple datasets and under different dataset size settings. The code is available at https://github.com/QtacierP/PRIOR.
title PRIOR: Prototype Representation Joint Learning from Medical Images and Reports
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2307.12577