FACTS: Fully Automatic CT Segmentation of a Hip Joint

Abstract

Extraction of surface models of a hip joint from CT data is a pre-requisite step for computer assisted diagnosis and planning (CADP) of periacetabular osteotomy (PAO). Most of existing CADP systems are based on manual segmentation, which is time-consuming and hard to achieve reproducible results. In this paper, we present a Fully Automatic CT Segmentation (FACTS) approach to simultaneously extract both pelvic and femoral models. Our approach works by combining fast random forest (RF) regression based landmark detection, multi-atlas based segmentation, with articulated statistical shape model (aSSM) based fitting. The two fundamental contributions of our approach are: (1) an improved fast Gaussian transform (IFGT) is used within the RF regression framework for a fast and accurate landmark detection, which then allows for a fully automatic initialization of the multi-atlas based segmentation; and (2) aSSM based fitting is used to preserve hip joint structure and to avoid penetration between the pelvic and femoral models. Taking manual segmentation as the ground truth, we evaluated the present approach on 30 hip CT images (60 hips) with a 6-fold cross validation. When the present approach was compared to manual segmentation, a mean segmentation accuracy of 0.40, 0.36, and 0.36 mm was found for the pelvis, the left proximal femur, and the right proximal femur, respectively. When the models derived from both segmentations were used to compute the PAO diagnosis parameters, a difference of 2.0 ± 1.5°, 2.1 ± 1.6°, and 3.5 ± 2.3% were found for anteversion, inclination, and acetabular coverage, respectively. The achieved accuracy is regarded as clinically accurate enough for our target applications.

Appendix

How to convert Volumetric Overlap Error (OE) to Dice Overlap Coefficient (DOC)?

Let $L_{1}$, $L_{2}$ $\subset {\mathbb{R}}^{3}$ denote the ground truth segmentation and the automatic segmentation, OE is calculated with the definition described in30:

$${\text{OE}} = 1 - \frac{{(|L_{1} \mathop \cap \nolimits L_{2} |)}}{{(|L_{1} \mathop \cup \nolimits L_{2} |)}}$$

(9)

And DOC is defined as

$${\text{DOC}} = 2 \times \frac{{|L_{1} \mathop \cap \nolimits L_{2} |}}{{(\left| {L_{1} } \right| + |L_{2} |)}}$$

(10)

Since we have

$$\left( {\left| {L_{1} \mathop \cup \nolimits L_{2} } \right|} \right) = (\left| {L_{1} } \right| + \left| {L_{2} } \right|) - |L_{1} \mathop \cap \nolimits L_{2} |$$

(11)

After some mathematic manipulations, one can derive following relationship between OE and DOC:

$${\text{DOC}} = 2 \times \frac{{(1.0 - {\text{OE}})}}{{(2.0 - {\text{OE}})}}$$

(12)

For example, in Ref.,11 Kainmueller et al. reported an average OE of 9.7% for segmenting right hip bone (RHB) (Table IV of Ref. 11). The corresponding DOC is then 94.90%.