The use of foreign exam management to share diagnostic images (DI) and reports across disparate organizations has been well adopted across various provinces and territories in Canada. It is common that each site contributing to a Diagnostic Imaging Repository (DIR) will have a unique terminology lexicon local to that specific site. Deterministic terminology mappings are often applied to associate a relationship between the local to regional DI procedure codes and names.
In this presentation, we will examine the use of Machine Learning to create a probabilistic model to predict image type with 90% accuracy and offer a new unsupervised methodology that clusters the images based on similarities in their metadata.
- Identify the key challenges with semantic interoperability as it relates to sharing exams between sites with disparate terminology lexicons.
- Explain the process involved that built the Machine Learning model to function successfully.
- Understand how your organization can apply the use of this Machine Learning model and appreciate future opportunities this model may assist with.