Shape analysis for assessment of progression in spinal deformities

Adolescent idiopathic scoliosis (AIS) is a three-dimensional structural spinal deformation. It is the most common type of scoliosis. It can be visually detected as a lateral curvature in the postero-anterior plane. This condition starts in early puberty, affecting between 1-4% of the adolescent population between 10-18 years old, affecting in majority female. In severe cases (0.1% of population with AIS) the patient will require a surgical treatment. To date, the diagnosis of AIS relies on the quantification of the major curvature observed on posteroanterior and sagittal radiographs. \n \nRadiographs in standing position are the common imaging modality used in clinical settings to diagnose AIS. The assessment of the deformation is carried out using the Cobb angle method. This angle is calculated in the postero-anterior plane, and it is formed between a line drawn parallel to the superior endplate of the upper vertebra included in the scoliotic curve and a line drawn parallel to the inferior endplate of the lower vertebra of the same curve. Patients that present a Cobb angle of more than 10°, are diagnosed with AIS. \n \nThe gold standard to classify curve deformations is the Lenke classification method. This paradigm is widely accepted in the clinical community. It divides spines with scoliosis into six types and provides treatment recommendations depending on the type. This method is limited to the analysis of the spine in the 2D space, since it relies on the observation of radiographs and Cobb angle measurements. \n \nOn the one hand, when clinicians are treating patients with AIS, one of the main concerns is to determine whether the deformation will progress through time. Knowing beforehand of how the shape of the spine is going to evolve would aid to guide treatments strategies. On the other hand, however, patients at higher risks of progression require to be monitored more frequently, which results in constant exposure to radiation. Therefore, there is a need for an alternative radiation-free technology to reduce the use of radiographs and alleviate the perils of other health issues derived from current imaging modalities. \n \nThis thesis presents a framework designed to characterize and model the variation of the shape of the spine throughout AIS. This framework includes three contributions: 1) two measurement techniques for computing 3D descriptors of the spine, and a classification method to categorize spine deformations, 2) a method to simulate the variation of the shape of the spine through time, and 3) a protocol to generate a 3D model of the spine from a volume reconstruction produced from ultrasound images. \n \nIn our first contribution, we introduced two measurement techniques to characterize the shape of the spine in the 3D space, leave-n-out, and fan leave-n-out angles. In addition, a dynamic ensemble method was presented as an automated alternative to classify spinal deformations. Our measurement techniques were designed for computing the 3D descriptors and to be easy to use in a clinical setting. Also, the classification method contributes by assisting clinicians to identify patient-specific descriptors, which could help improving the classification in borderline curve deformations and, hence, suggests the proper management strategies. \n \nIn order to observe how the shape of the spine progresses through time, in our second contribution, we designed a method to visualize the shape’s variation from the first visit up to 18 months, for every three months. Our method is trained with modes of variation, computed using independent component analysis from 3D model reconstructions of the spine of patients with AIS. Each of the modes of variation can be visualized for interpretation. This contribution could aid clinicians to identify which spine progression pattern might be prone to progression. \n \nFinally, our third contribution addresses the necessity of a radiation-free image modality for assessing and monitoring patients with AIS. We proposed a protocol to model a spine by identifying the spinous processes on a volume reconstruction. This reconstruction was computed from ultrasound images acquired from the external geometry of the subject. Our acquisition protocol documents a setup for image acquisition, as well as some recommendations to take into account depending on the body composition of the subjects to be scanned. We believe that this protocol could contribute to reduce the use of radiographs during the assessment and monitoring of patients with AIS.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.