ROKS: Predicting pain in cancer patients with bone metastases

The overarching objective of this project is to build a machine learning pipeline to combine radiomics, natural language processing (NLP), and patient reported outcomes (PROs) to predict cancer pain using the radiography images of cancer patients with bone metastasis.

In this project, PhD student Hossein Naseri will extract quantitative imaging features (such as tumor location, shape, size, intensity) from patient medical images (CT scans) using radiomics techniques and attempt to correlate them with physician-reported pain measures and PROs. Physician-reported measures will be extracted from electronic medical records using NLP. PROs will be collected via Opal.

Hossein will work towards achieving this objective by working on four specific tasks:

  1. Quantify physician-reported pain by applying NLP to retrospectively-collected clinical notes of patients with bone metastasis.
  2. Extract imaging features from CT and cone-beam CT scans using radiomics algorithms.
  3. Use NLP-quantified pain scores and extracted radiomic features to train a machine learning algorithm to predict pain from the CT scans of patients with bone metastasis.
  4. Validate our machine learning model using prospectively-collected PROs data collected from patients using the Opal app.
Hossein Naseri
Hossein Naseri
PhD Student
John Kildea
John Kildea
Assistant Professor of Medical Physics