Image Analysis

The primary objective was to create an unifying image processing framework, which can handle all image processing related steps from the original radiological data and until the simulated patient-specific model. These steps comprise: image enhancement (e.g. noise removal); co-registration of all data into a same reference coordinate system; segmentation of different organs and their substructures (e.g. tumour, lesion, vessels, bronchi); meshing i.e. converting the binary segmentations into spatially discretized elements needed in numerical simulation. Previously, several separate tools and format conversions between them were needed and this has led to complex, time-consuming and error prone workflows. During this project we have built a unified single framework that provides seamless workflow for simpler, faster and safer way for the creation of patient specific models within the single web-based Environment.

Before simulation image analysis software is used for building patient specific 3-D models for FEM simulation. We have developed precise segmentation tools for relevant anatomical structures and accurate registration tools for aligning pre- interventional data into one common coordinate system.

Once the 3-D model is build user can use a virtual needle for the placement of heat sources into the FEM model. If the interventional data is available, the software tools can be used for extracting the true heat sources and aligning them with the model coordinate system.

After the simulation, we can use image analysis software for comparing the simulated lesion and the actual ground-truth lesion. We have developed tools for the accurate segmentation of the ablation zone, and precise registration of the ablation zone with the 3-D model.

Two examples of 3-D models produced with Go-Smart image processing tools. In left, a lung model with lung, lung vessels and bronchi. On right, a liver model with liver, liver vessels and the target tumour.