type
status
date
slug
summary
tags
category
icon
password
This serves as a log for using learning using privileged information (LUPI) paradigm to enhancing WSI diagnosis.
Cover page credits to Fengtao Zhou (https://cvzzz.com/).
Note: this page will provide my thoughts for processing problems. This page would be served as my discovery logs and should not be treated seriously.
Sect 1. Introduction
Precision medicine … survival, biomarker, molecular,…
Biomarker classification (how to transition?)
money vs (cleared)
Whole-slide Images serves as golden standard in diagnosis. Also it is low-cost, routine-based product for diagnosis. However, there are plenty of Currently, many methods are tackling WSI classification through MIL methods. However, they neglects the information from other modalities (i.e., genomics, reports, etc.).
Plenty of literature suggests to use multi-modal information and fuse the complementary information for better predictions. However, obtaining complete modalities is challenging in real-world scenario.
Sect 2. Results
Extensive …
Quantitative good…
After this we conduct several analysis…
External validations on…
Sect 3. Method
My method first reconstruct, then align at two stages…
reconstruct is done by cross attention
Sect 4. Discussions
pathways alternation…
foundation model…
- Author:Peter Kam
- URL:https://blog.petergam.top//article/multi-modal-distillation-guide
- Copyright:All articles in this blog, except for special statements, adopt BY-NC-SA agreement. Please indicate the source!
