Author:

Abstract:

Ultrasonography offers the opportunity to directly visualize in-vivo anatomical features. This can be exploited in numerous clinical applications. Ultrasonography is especially useful to provide information on the diagnosis and continuous evaluation of spastic cerebral palsy, since it allows an easy and fast data collection also during dynamic activity. However, most ultrasound (US) investigations are confined by the measurement range of the US probe, and hence, they are typically only used to visualize specific features, rather than an entire muscle and tendon. Current techniques to enlarge this measurement range and to record volumetric data are not yet comprehensively validated for specific clinical applications and do not allow simple and efficient implementation within a clinical environment. Therefore, the main goal of this thesis is to establish a 3D freehand ultrasound (3DfUS) framework capable of acquiring and processing images from the lower limb muscles and tendons in children with spastic cerebral palsy in order to obtain clinically-relevant parameters. 3DfUS combines a conventional 2D US probe with a pose sensor that tracks the 3D pose of the probe. 3DfUS enlarges the US field of view and is capable of capturing volumetric data. The corresponding acquisition is carried out with the examiner manually moving the probe (hence ‘freehand’), thereby allowing easier imaging along and about the various anatomical shapes. Multidisciplinary research studies are needed to reach the main goal. This main goal leads to three sub-goals, which compose the structure of the thesis. The first sub-goal is to create a framework for 3DfUS that is applicable in a clinical setting and has a verified validity. The framework developed is adaptable to hardware changes and is suitable to be used in day-to-day clinical practice. The custom software created for making the 3D reconstruction is made publicly available. The accuracy analysis for volume and length measurements using the established framework is lower than 3%. In addition, one of the more crucial aspects to ensure a valid and reproducible 3D reconstruction is a repeatable and accurate calibration procedure for the relative pose between the US probe and the 3D measurement system. This procedure is successfully extended to acquire volumetric data also larger than the US transducer size. The second sub-goal is to provide efficient and validated methods to the framework to be successfully applied in musculoskeletal static conditions. These are still lacking and are particularly crucial to be defined in pathological muscles. The combination of the hardware and software established in the first sub-goal is tuned for this condition by supporting the clinicians with an optimized procedure for extracting the clinically relevant parameters and with a resolution lower than the variations expected in children with spastic cerebral palsy. Moreover, for the data acquisition, a simple, yet innovative, US probe is conceived to reduce superficial muscle deformation that also causes muscle border mismatching between US sweeps. The reduced muscle deformation is also useful to help the operator in a more appropriate visual feedback during the acquisition. The third sub-goal is similar to the second one but with applications in musculoskeletal dynamic conditions. The results reveal higher discrepancies when defining the fascicles lengthening rather than the muscle-tendon junction displacement. This analysis also defined the corresponding level of resolutions, thereby supporting the clinicians in understanding the clinical impact of the measurements. An automated method to track this junction by utilising the optical flow approach is proposed and validated in children with spastic cerebral palsy. This method also provides a tool for a quick and effective supervision for correcting errors of the automatic method. In conclusion, this thesis has established a clinically applicable 3DfUS framework for extracting relevant muscle and tendon parameters. The established framework is validated to be used in children with spastic cerebral palsy and is currently applied for several clinical research studies. About 150 children have been analyzed by means of this framework. This framework is now also used for the ongoing clinically oriented multicentre research project ‘Treatment Algorithms based on Muscle and Tendon Architecture’ (TAMTA, FWO-TBM project). The framework can be used for other pathologies and for other muscles. Finally, the wide scope of this project has laid down the foundations for further developing and improving the framework. This could help disseminate the 3DfUS framework which is currently only performed in similar investigations in a limited number of research centres.