Hugging Face's logo

2ms
/
EXAONE-Path-CRCMSI-Predictor

PyTorch
lg-ai
EXAONEPath-1.5
pathology
Model card Files
xet
Community
2
2ms cjs6211 commited on
Commit
88fde16
·
verified ·
1 Parent(s): 5b6672a

Update README.md (#1)

Browse files

- Update README.md (9e9f34302a407e2bd3ad6f823166f6692db2254a)


Co-authored-by: Jang <[email protected]>

Files changed (1) hide show
  1. README.md +21 -20
README.md CHANGED
@@ -7,15 +7,16 @@ tags:
7
  - EXAONEPath-1.5
8
  - pathology
9
  ---
10
- # EXAONE Path for CRCMSI – CRCMSI-centric Whole-Slide Image Classifier
11
- *A purpose-built upgrade of **EXAONE Path 1.5***
12
 
13
  ## Introduction
14
- **EXAONE Path for CRCMSI** is an **enhanced whole-slide image (WSI) classification framework** that retains the core architecture of EXAONE Path 1.5 while upgrading its internals for greater efficiency and richer multimodal integration.
 
15
 
16
  The pipeline still unfolds in two stages:
17
 
18
- 1. **Patch-wise feature extraction** – Each WSI is tiled into 256 × 256 px patches, which are embedded into 768-dimensional vectors using the frozen **[EXAONE Path v1.0](https://huggingface.co/LGAI-EXAONE/EXAONEPath)** encoder.
19
  2. **Slide-level aggregation** – The patch embeddings are aggregated using a Vision Transformer, producing a unified slide-level representation that a lightweight classification head transforms into task-specific probabilities.
20
 
21
  ---
@@ -69,19 +70,19 @@ python inference.py --svs_path ./samples/wsis/1/1.svs
69
 
70
  ### Model Performance Comparison
71
 
72
- | Metric (AUC) / Task | Titan (Conch v1.5 + iBot, image-text) | PRISM (virchow + perceiver, image-text) | CHIEF (CTransPath + CLAM, image-text, WSI-contrastive) | Prov-GigaPath (GigaPath + LongNet, image-only, mask-prediction) | UNI2-h + CLAM (image-only) | EXAONEPath V1.5 | **{MODEL_NAME}** |
73
- |------------------------------------|---------------------------------------|-----------------------------------------|--------------------------------------------------------|-----------------------------------------------------------------|---------------------------|------------------------|------------------------|
74
- | **CRC-MSI** | 0.9370 | 0.9432 | 0.9273 | 0.9541 | <u>0.9808</u> | 0.9537 | **0.9844** |
75
- | LUAD-TMB (cutoff 10) | 0.6901 | 0.6445 | 0.6501 | 0.6744 | 0.6686 | 0.6846 | 0.6842 |
76
- | LUAD-EGFR-mut | 0.8197 | 0.8152 | 0.7691 | 0.7623 | 0.8577 | 0.7607 | 0.8564 |
77
- | LUAD-KRAS-mut | 0.5405 | 0.6299 | 0.4676 | 0.5110 | 0.4690 | 0.5480 | 0.6038 |
78
- | BRCA-ER | 0.9343 | 0.8998 | 0.9115 | 0.9186 | 0.9454 | 0.9096 | 0.9278 |
79
- | BRCA-PR | 0.8804 | 0.8613 | 0.8470 | 0.8595 | 0.8770 | 0.8215 | 0.8430 |
80
- | BRCA-HER2 | 0.8046 | 0.8154 | 0.7822 | 0.7891 | 0.8322 | 0.7811 | 0.8050 |
81
- | BRCA-TP53 | 0.7879 | 0.8415 | 0.7879 | 0.7388 | 0.8080 | 0.6607 | 0.7656 |
82
- | BRCA-PIK3CA | 0.7577 | 0.8929 | 0.7015 | 0.7347 | 0.8571 | 0.7066 | 0.7908 |
83
- | RCC-PBRM1 | 0.6383 | 0.5570 | 0.5129 | 0.5270 | 0.5011 | 0.4445 | 0.5780 |
84
- | RCC-BAP1 | 0.7188 | 0.7690 | 0.7310 | 0.6970 | 0.7160 | 0.7337 | 0.7323 |
85
- | COAD-KRAS | 0.7642 | 0.7443 | 0.6989 | 0.8153 | 0.9432 | 0.6790 | 0.8693 |
86
- | COAD-TP53 | 0.8889 | 0.8160 | 0.7014 | 0.7118 | 0.7830 | 0.8785 | 0.8715 |
87
- | <span style="color:red">**Average**</span> | 0.7817 | 0.7869 | 0.7299 | 0.7457 | <u>0.7876</u> | 0.7356 | <span style="color:red">**0.7932**</span> |
 
7
  - EXAONEPath-1.5
8
  - pathology
9
  ---
10
+ <!--# EXAONE Path for CRCMSI – CRCMSI-centric Whole-Slide Image Classifier
11
+ *A purpose-built upgrade of **EXAONE Path 1.5***-->
12
 
13
  ## Introduction
14
+ <!--**EXAONE Path for CRCMSI** is an **enhanced whole-slide image (WSI) classification framework** that retains the core architecture of EXAONE Path 1.5 while upgrading its internals for greater efficiency and richer multimodal integration.-->
15
+ **EXAONE Path MSI** is an **enhanced whole-slide image (WSI) classification framework** that retains the core architecture of EXAONE Path while upgrading its internals for greater efficiency and richer multimodal integration.
16
 
17
  The pipeline still unfolds in two stages:
18
 
19
+ 1. **Patch-wise feature extraction** – Each WSI is tiled into 256 × 256 px patches, which are embedded into 768-dimensional vectors using the frozen **[EXAONE Path](https://huggingface.co/LGAI-EXAONE/EXAONEPath)** encoder.
20
  2. **Slide-level aggregation** – The patch embeddings are aggregated using a Vision Transformer, producing a unified slide-level representation that a lightweight classification head transforms into task-specific probabilities.
21
 
22
  ---
 
70
 
71
  ### Model Performance Comparison
72
 
73
+ | Metric (AUC) / Task | Titan (Conch v1.5 + iBot, image-text) | PRISM (virchow + perceiver, image-text) | CHIEF (CTransPath + CLAM, image-text, WSI-contrastive) | Prov-GigaPath (GigaPath + LongNet, image-only, mask-prediction) | UNI2-h + CLAM (image-only) | EXAONE Path MSI |
74
+ |------------------------------------|---------------------------------------|-----------------------------------------|--------------------------------------------------------|-----------------------------------------------------------------|---------------------------|------------------------|
75
+ | **CRC-MSI** | 0.9370 | 0.9432 | 0.9273 | 0.9541 | <u>0.9808</u> | **0.9844** |
76
+ <!--| LUAD-TMB (cutoff 10) | 0.6901 | 0.6445 | 0.6501 | 0.6744 | 0.6686 | 0.6846 | |
77
+ | LUAD-EGFR-mut | 0.8197 | 0.8152 | 0.7691 | 0.7623 | 0.8577 | 0.7607 | |
78
+ | LUAD-KRAS-mut | 0.5405 | 0.6299 | 0.4676 | 0.5110 | 0.4690 | 0.5480 | |
79
+ | BRCA-ER | 0.9343 | 0.8998 | 0.9115 | 0.9186 | 0.9454 | 0.9096 | |
80
+ | BRCA-PR | 0.8804 | 0.8613 | 0.8470 | 0.8595 | 0.8770 | 0.8215 | |
81
+ | BRCA-HER2 | 0.8046 | 0.8154 | 0.7822 | 0.7891 | 0.8322 | 0.7811 | |
82
+ | BRCA-TP53 | 0.7879 | 0.8415 | 0.7879 | 0.7388 | 0.8080 | 0.6607 | |
83
+ | BRCA-PIK3CA | 0.7577 | 0.8929 | 0.7015 | 0.7347 | 0.8571 | 0.7066 | |
84
+ | RCC-PBRM1 | 0.6383 | 0.5570 | 0.5129 | 0.5270 | 0.5011 | 0.4445 | |
85
+ | RCC-BAP1 | 0.7188 | 0.7690 | 0.7310 | 0.6970 | 0.7160 | 0.7337 | |
86
+ | COAD-KRAS | 0.7642 | 0.7443 | 0.6989 | 0.8153 | 0.9432 | 0.6790 | |
87
+ | COAD-TP53 | 0.8889 | 0.8160 | 0.7014 | 0.7118 | 0.7830 | 0.8785 | |
88
+ | <span style="color:red">**Average**</span> | 0.7817 | 0.7869 | 0.7299 | 0.7457 | <u>0.7876</u> | <span style="color:red">**0.7932**</span> |-->