Lung Nodule CT Classification Model

I’m looking for a Deep Learning Engineer / Medical Imaging Researcher to work on a classification project involving lung nodule malignancy prediction.

The project uses:

LIDC/IDRI for CT-based nodule image classification.

NLST for clinical and demographic risk factor data.


The objective is to develop a research-level classification model that predicts whether a lung nodule is benign or malignant — achieving accuracy, AUC, and sensitivity close to published benchmarks such as NoduleX (Nature Scientific Reports, 2018).

You can modify preprocessing, model design, or training methods as needed — creativity and originality are encouraged.
AI tools (Claude, Cursor, Copilot, ChatGPT, etc.) may be used responsibly as long as the work is unique, well-documented, and high-quality.

Further technical details, performance targets, and references will be shared privately with shortlisted candidates.
---

Budget & Payment Structure

Total Fixed Budget: $400
Milestone-Based:

$50 → Enhancement Plan (short proposal on your implementation plan and paper references)

$100 → Dataset preprocessing and documentation

$250 → Final model, results, checkpoints, and report (after achieving or closely matching target metrics)



---

Expectations

Ability to handle both LIDC/IDRI and NLST datasets.

Experience in medical image classification and deep learning (CNNs, 3D CNNs, or ViTs).

Understanding of academic research papers and ability to apply their ideas in implementation.

Code must include checkpoints, metric tracking, and clear documentation.

Deliverables must be original (not direct copies of existing papers or codebases).



---

Reference Links (Context Only)

NoduleX paper (baseline concept):
https://www.nature.com/articles/s41598-018-27569-w

LIDC/IDRI dataset: https://www.cancerimagingarchive.net/collection/lidc-idri/

NLST dataset:
https://cdas.cancer.gov/nlst/



---

Screening Questions:

Dataset & Storage

1. Have you worked with LIDC/IDRI or NLST before?


2. How will you handle the size and storage of these datasets?



Model & Methods
3. What type of architecture (e.g., 3D CNN, Vision Transformer, hybrid) do you plan to use for classification?
4. How will you manage checkpoints and reproducibility?
5. Which framework do you prefer (PyTorch or TensorFlow), and why?

Research & Tools
6. Can you name 1–3 academic papers that are relevant to this project?
7. How do you typically use research papers to guide your implementations?
8. Will you use AI assistants (Claude, Cursor, Copilot, ChatGPT, etc.)? If yes, how will you ensure originality?

Experience & Reliability
9. Please share GitHub or Colab links showing your similar past work.
10. What GPU or server environment will you use for training (e.g., Colab Pro, GCP, local RTX)?
11. How often can you share progress (checkpoints, metrics, or short updates)?


---

Notes:

All payments go through escrow after verified milestone completion.

Additional technical details, references, and dataset handling plans will be shared privately with selected candidates.

Preference for applicants with medical imaging experience and solid understanding of research-based model design.

Bonus for originality, clean code, and strong results.