| Start Date: | 4/24/2023 | Start Time: | 2:00 PM |
| End Date: | 4/24/2023 | End Time: | 4:00 PM |
|
Event Description
BIOMED PhD Research Proposal
Title:
Machine Learning-based Classification and Automated Measurement of Pre-operative Clinical Indices in Early Onset Scoliosis (EOS) Patients
Speaker:
Girish Viraraghavan, PhD Candidate
School of Biomedical Engineering, Science and Health Systems
Drexel University
Advisor:
Sriram Balasubramanian, PhD
Associate Professor
School of Biomedical Engineering, Science and Health Systems
Drexel University
Details:
Early Onset Scoliosis (EOS) involves complex deformities of the spine and thorax in children under 10 years of age. EOS is believed to account for nearly 10% of all pediatric scoliosis cases, with true prevalence unknown. Etiology of spinal deformity (scoliosis) varies widely among EOS patients and can be classified into congenital scoliosis (structural abnormality of spine or thorax present at birth), idiopathic scoliosis (scoliosis without a known cause), neuromuscular scoliosis (abnormal muscle tone leading to scoliosis) and syndromic scoliosis (syndrome assisted scoliosis excluding congenital and neuromuscular scoliosis). Progressive spine deformity leads to a modified thoracic cage development, resulting in reduced size and shape of the thorax which ultimately affects normal lung development in growing children. Such reduced growth in EOS patients may ultimately lead to the inability of the thorax to support normal respiration and growth termed as thoracic insufficiency syndrome (TIS). Children with TIS can have further complications such as pulmonary hyperplasia (incomplete development of lungs) which can ultimately led to premature death.
Currently, the C-EOS classification system is the only available method for grouping EOS patients based on pre-operative clinical indices. While this system is both reliable and accurate, the cut-offs for major curve (Cobb) angle and kyphosis are based on arbitrary values. Additionally, the 48 subgroups in C-EOS limit meaningful analysis due to the small number of patients in each subgroup, making it difficult to correlate interventions with outcomes. While C-EOS has helped establish a standardized method to communicate different aspects of EOS deformity, it is not as widely used as the Lenke classification system for Adolescent Idiopathic Scoliosis (AIS) in guiding surgical treatment. Consequently, there is a need for data-driven grouping of the heterogeneous EOS patient population to generate a limited number of automated, meaningful subgroups based on pre-operative clinical indices of EOS patients.
Moreover, manual radiographic measurement of clinical indices is time-consuming and associated with significant inter-observer errors. Furthermore, the presence of vertebral, lung, and ribcage abnormalities in patients with EOS makes the identification of anatomical structures such as vertebrae challenging, resulting in greater measurement errors. Hence, there is a need for automated identification of anatomical structures like vertebrae and lungs using machine learning, which will ultimately aid in making accurate measurements of clinical indices from radiographs and help reduce the variability associated with manual measurements. Currently, no automated methods for identifying anatomical structures exist for the EOS population of clinical indices.
The overall goal of this proposal is to 1) introduce automated methods to help identify meaningful subgroups based on pre-operative clinical indices of EOS patients, and 2) identify anatomical structures, such as vertebrae and lungs, from patient radiographs to aid in accurate measurement of clinical indices. |
|
Audience:
Undergraduate StudentsGraduate StudentsFacultyStaff |
|
