Lazy loaded image
Logs for Semantic-Relation Constrained Whole-Slide Image Processing
00 min
Oct 1, 2023
Dec 18, 2024
type
status
date
slug
summary
tags
category
icon
password
The following survey papers are (partially) quoted to illustrate the concepts of these papers:
Lin et al. Dual-path network with synergistic grouping loss and evidence driven risk stratification for whole slide cervical image analysis. MIA 2021. Cheng, et al. Robust whole slide image analysis for cervical cancer screening using deep learning. Nature Communications 2021. Zhu, et al. Hybrid AI-assistive diagnostic model permits rapid TBS classification of cervical liquid-based thin-layer cell smears. Nature Communications 2021. Zhao, et al. Clinical-Inspired Cytological Whole Slide Image Screening with Just Slide-Level Labels. Axiv 2023. Stegmüller, et al. Self-Supervised Learning-Based Cervical Cytology Diagnostics in Low-Data Regime and Low-Resource Setting. arXiv 2023.
arXiv, 2306.03407
Many thanks to the original authors.
 
My current research field is whole-slide images (WSI) processing, and this page will provide my thoughts for processing problems. This page would be served as my discovery logs and should not be treated seriously.
 
Update: [2024.12.18] The paper is accepted by IEEE TMI! It was a long wait for 839 days.
 

Sect. 0 Survey

Sect. 0.1 Cytology WSI Analysis Survey

In fact, due to the sparsity of its publicly available data, there are few discoveries in DL in cytology whole-slide images (WSI). Most of them can be categorized into the following subfields:
 
notion image
 
(A one-stage design and a two-stage design paper are redacted since the confidential request)
 
As we may see, current methods are limiting themselves in two-stage networks, i.e., first feed the patches into a detector, and then aggregate the attributes, and recently, several one-stage methods borrowing concepts from histopathology are on the trend. This blog is intended to dig fine-grained methods to explore MIL in cytology methods.
 

Sect. 0.2 Histopathology WSI Analysis Survey

(Neglected specific methods introduction. Namely, MAXMIL/MeanMIL/ABMIL/CLAM/DSMIL/TransMIL/HIPT/DTFD/PMIL/S4MIL/ICMIL/HEAT/PatchGCN/BCL/ILRA-MIL… etc.)
Comments:
1. Current field of histopathology WSI analysis contains lots of attention-based aggregators MIL methods.
  1. There exists high class unbalance problem even within a single WSI.
notion image
3. (From our experiment) Current MIL methods exists biased attention distribution problem, meaning the core problem of MIL is building the accurate mapping of the malicious patches and the benign patches. Also, the highly class imbalance within a single WSI contributes to this degradation.
notion image
  1. To the best of my knowledge, all those mentioned methods can be roughly divided into the following fields:
    1. Basic MILs: Mean-pooling/Max-pooling/Attention-based pooling (ABMIL)
    2. Small Attention Mechanism Enhancement: CLAM/TransMIL/PMIL/DTFD
    3. Leveraging New Architecture: HEAT/PatchGCN (GNNs), ILRA-MIL (Low Rank Properties), S4MIL (State Space Models)
    4. Leveraging Multi-resolution Correlations: DSMIL/HIPT
    5. Leveraging Iterative Refinement: ICMIL/BCL

Sect. 1 Proposal & Possible Directions

  1. Current model have severe attention biased problem, meaning there is still room to improve the mapping of the malicious patches and the slide label.
  1. Self-challenge is a training technique for hard problem. (https://link.springer.com/chapter/10.1007/978-3-030-58536-5_8)
  1. No method is discovering the coarse-to-fine relationship. i.e., finer level class has dependencies on coarser level class.
  1. No MIL classification methods for fine-grained cytology classification.
 
For current project, I attempt to and have discovered Directions 3&4.
 

Sect 1.1 Failed Attempts

Current coarse-to-fine design seems to related to the term hierarchical classification. Three possible sub-directions are presented as follows:
 
  1. Tree-based classifier/loss:
notion image
Ref. Kim, Sangwon, Jaeyeal Nam, and Byoung Chul Ko. "Vit-net: Interpretable vision transformers with neural tree decoder." International Conference on Machine Learning. PMLR, 2022.
 
b. Transformer Prompting:
notion image
Ref. Wang, Wenhao, et al. "TransHP: Image Classification with Hierarchical Prompting." Advances in Neural Information Processing Systems 36 (2023).
 
Neither does these methods work on the cytology dataset based on ABMIL backbone.
 
c. Utilize Classifiers in different hierarchies (Sort like ensemble?)
This direction works in my attempts.
 

Sect 2. Method

notion image
A pretty simple method is formed like this. By using coarse-grained labels as guidance for constrain, this proposed network can serve as a fine-grained MIL method for cervical WSI classification.
 
This method comprises of two stages:
  • Instance-level Hierarchy Learning
    • In this stage, the attention learned from each individual patches are forced to align through class hierarchy.
  • Bag-level Hierarchy Learning
    • In this stage, the class probabilities are aligned through class hierarchy.
 
To further improve the effectiveness of fine-grained classification, we employed supervised contrastive learning and dynamic weighting to balance two branches. Why is this? Since supervised contrastive learning are inherent imbalance label dealer (https://arxiv.org/abs/2004.11362), and the latter is pure of heuristic. By plugging these modules, a patchwork is formed — But it works.
 
I hate to mention it, but I really have to say this is my first work of my PhD studies. Therefore it has the meaning of my worked PhD training.
 
It really is not a big deal compared to other “hot” or “high technical contribution” papers, but means a lot to me.
 
 
 
 
Prev Post
Logs for Casual Deep Learning
Next Post
[Restricted Access] Franticness Research

Comments
Loading...