An AI tool to measure degrees of misalignment of lower limb rotation in pre-surgical planning assessments .
CT Scans
Lower limb rotation
Graveyard
Current practice of manually drawing lines over the long bones and the joints on CT images in PACS to figure the degree
of misalignment is time-consuming (15-20 minutes per case). An AI tool would automatically create these lines and
angles, saving time and reducing variability.
Clinical lead: Dr. Christopher Tang
Project Plan
1. Meeting of all persons involved to determine AI specifications. 4. Model training
For radiologist (Chris)
:
- Confirm +/3 to 5 degree precision with peers
- Confirm with surgeons how angles are used to guide knee surgery
- Confirm 3 bone ankle slice is the correct slice to analyse.
- Review software requirements spec
2. Setting technical and system requirements for AI model.
For CSC (Anil)
- Perform the risk assessment
- Create software design spec from software requirements spec
- Attempt to build non-A.I software with “dicom_server” and eventually “AI Deployment Engine”
- Investigate the AI route
3. Dataset curation (retrospective).
- Collect Patient ID’s for 100-200 patients, analysed by different radiologists
5.Model testing
6. Implementation
7. Audit
References:
Website used to help calculate rotational angles is here.