Automatic brain anatomical substructure segmentation tools and holographic visualisation for diagnostic aid and decision support.
Objectives
The main objective is to create tools to support clinicians in the diagnosis, treatment and prognosis of brain diseases in which medical imaging is a criterion of great importance for diagnosis and prognosis.
Use case
An automated tool is used for the segmentation of solid tumours from medical images in DICOM/NIfTI format. Subsequently, a 3D image is generated, allowing for visualisation through smart glasses to facilitate diagnostic tasks.
Infrastructure
On Premise
Technology
Computer vision Machine learning and deep learning
Data
Private and anonymised dataset (IIS BIOARABA)
Resources
Deusto SEIDOR research staff for the definition of segmentation models in collaboration with IIS BIOARABA staff as dataset providers and validators. SEIDOR’s high-performance DPC infrastructure.
Difficulties and learning
The main challenge involved adhering to data security protocols, including obtaining project approval from the Research Ethics Committee, extracting medical images from the PACS, anonymising data, generating masks, etc. Data acquisition and dataset preparation can be a time-consuming process, often spanning several months. Therefore, identifying public datasets or generating synthetic data becomes crucial to advance the project until real data becomes available.