CD-Tron: Cognitive Decline Detection from EHR using Large Clinical Language Model

Model Name: CD-Tron

Model Description

CD-Tron is a fine-tuned large clinical language model based on GatorTron for the task of detecting cognitive decline from free-text clinical notes.

The model was fine-tuned on real-world clinical data, and synthetic data can be used for demonstration.

Intended Use

Task: Cognitive decline detection / screening
Input: Free-text clinical notes (EHR sections, progress notes, discharge summaries, etc.)
Output: Binary classification:
- 0 = No cognitive decline
- 1 = Cognitive decline detected

This model is for research purposes and proof-of-concept demonstration.

How to Use

Example code to load and run inference:

from transformers import AutoTokenizer, AutoModelForSequenceClassification

tokenizer = AutoTokenizer.from_pretrained("HAO-AI/cdtron-cognitive-decline")
model = AutoModelForSequenceClassification.from_pretrained("HAO-AI/cdtron-cognitive-decline")

text = "Patient presents with recent memory loss, confusion, and impaired attention..."
inputs = tokenizer(text, return_tensors="pt", truncation=True, padding=True, max_length=512)
outputs = model(**inputs)
prediction = outputs.logits.argmax(dim=1).item()
print("Predicted label:", prediction)

Citation

If you find this work useful, please cite:

@article{guan2025cd,
  title={CD-Tron: Leveraging large clinical language model for early detection of cognitive decline from electronic health records},
  author={Guan, Hao and Novoa-Laurentiev, John and Zhou, Li},
  journal={Journal of Biomedical Informatics},
  pages={104830},
  year={2025},
  publisher={Elsevier}
}