MRI-based, patient-specific 3D bone imaging for orthopaedic surgery planning

To plan a surgery to fix broken or diseased bone, surgeons rely on a 2D image, a computed tomography (CT) scan, to visualise the patient's 3D bony anatomy. This is a difficult task for even the most experienced surgeon, leading to many misplanned surgeries.

The use of patient-specific 3D bone images derived from a patient's CT scan improves surgical planning accuracy. Therefore, there is a strong demand from healthcare professionals to utilise these 3D images in treating diseased and broken bones.

Medical imaging companies provide basic software with their CT scanners that can generate a 3D visualisation of the patient's bony anatomy. However, this software only provides a 3D representation of the 2D imaging data, known as a 3D volumetric render, showing all the anatomy within the scan. Therefore, it can't provide healthcare professionals with a patient-specific visualisation of the specific anatomy of interest. Furthermore, the 3D visual has to be manually generated and can only be viewed on a specific hospital computer.

AXIAL MEDICAL PRINTING LIMITED (Axial3D) currently provides the cloud-based software platform, Axial3D-Insight, which can automatically generate patient-specific 3D images of bone from CT scans that can be viewed on the go, on any device. Unlike the 3D volumetric rendering software, a process called segmentation is used where machine learning (ML) algorithms automatically label different types of anatomy (bone or soft tissue) within the scan. This creates an accurate, patient-specific, 3D visualisation of the specific anatomy type of interest without the need for manual input.

Medical imaging companies are actively looking to expand the capabilities of their imaging modalities to reduce patient exposure to ionising radiation during CT scanning. Conventionally, CT images bone and magnetic resonance imaging (MRI) images soft tissue. GE Healthcare has developed a new MRI imaging protocol, oZTEo, that can effectively image bone and soft tissue anatomy within the same scan. CT-like imaging of the bone is achieved without the need for ionising radiation and also the patient only needs to be scanned once, saving time and healthcare costs.

In this project, we will extend our current offering by developing new ML software that automatically segments a single bony anatomical region of interest such as the knee and the parts of this bony anatomical region such as the shinbone, thigh bone and kneecap (semantic segmentation) within these new MRI bone scans to generate a precise, patient-specific 3D image of the same for surgical planning use.