AegisOSS/stage-2

Aegis Stage 2 — OncoTree Subtype Classifier (ResNet-50)

Aegis Stage 2 is a high-performance deep learning model trained on 10,986 real histopathology tiles across 8 mapped data sources to perform fine-grained cancer subtype categorization according to the MSK OncoTree Taxonomy.

• Architecture: ResNet-50 (timm)
• Total Parameters: 23,526,473 (~23.53 Million)
• INT8 ONNX Model Size: 22.66 MB
• Test Accuracy: 99.77% (425 / 426 test tiles correct)
• Macro-F1 Score: 99.67%
• License: Creative Commons Attribution 4.0 (CC-BY-4.0)

---

Performance Benchmarks (Unseen Test Set)

Evaluated across 426 unseen histopathology test tiles:

Subtype	Class Name	Support	Precision	Recall	F1-Score	Status
BRCA	Invasive Breast Carcinoma	171	1.0000	1.0000	1.0000
GB	Glioblastoma	68	1.0000	1.0000	1.0000
CRC	Colorectal Carcinoma	65	1.0000	1.0000	1.0000
LUAD	Lung Adenocarcinoma	60	1.0000	0.9833	0.9916
LUSC	Lung Squamous Cell	62	0.9841	1.0000	0.9920

---

Dataset Attribution & Citation

This model was trained on public histopathology research datasets:

• LC25000 (CC-BY-4.0)
• Dorsar Lung Cancer (MIT License)
• Breast Histopathology IDC (CC-BY-4.0)
• Brain Tumor Pathology (CC-BY-4.0)

---

Medical Research Disclaimer

Aegis is an open-source AI research prototype intended for educational, scientific evaluation, and research purposes only. It is not a certified medical device and must not be used for primary clinical diagnosis or treatment planning.

Citation

@software{aegis_stage1,
  title = {Aegis Stage 2: OncoTree Subtype Classifier ResNet-50},
  author = {Ranveer Soni},
  year = {2026},
  url = {https://huggingface.co/AegisOSS/stage-2},
  license = {CC-BY-4.0}
}