Statistical shape modeling (SSM) enables quantitative analysis of anatomical shapes. SSM is widely used in biology and medicine to model anatomies and their shape variability within populations. The technological advancements of in vivo images have led to the development of various open-source tools that can automate statistical analysis of shapes. These tools are based on different modeling approaches and assumptions to accomplish the same objective. However, little work has been done in the systematic evaluation and validation of SSM tools in clinical applications that rely on morphometric quantifications.
In this thesis, we emphasize and demonstrate the importance of evaluation and validation of SSM tools in different clinical applications. To this end, SSM can help in assessing patient-specific anatomical information by relating it to population-level morphometrics. The clinical needs that are driven by patient-specific information, such as implant design and selection, motion tracking, surgical planning, and lesion screening, are considered for analysis. The shape models for these clinical needs are analyzed from three widely used, state-of-the-art SSM tools, namely, ShapeWorks, Deformetrica, and SPHARM-PDM. The shape models are evaluated and validated using intrinsic and extrinsic assessments. The evaluation and validation experiments show that SSM tools display different levels of consistencies in performance. ShapeWorks and Deformetrica models are more consistent compared to models from SPHARM-PDM due to the group-wise approach of estimating shape correspondences. Furthermore, ShapeWorks and Deformetrica shape models capture clinically relevant population-level variability compared to SPHARM-PDM models.
In this thesis, a literature survey is performed to identify the existing studies that performed evaluation and validation of SSM tools in clinical applications. The need for such an assessment is showcased using a proof-of-concept experiment. A clinical application-driven validation framework is proposed to compare the performance of SSM tools. The framework is tested on clinical needs to provide insights about the deployment of SSM in real clinical scenarios.