DeepSSM: A Deep Learning Framework for Statistical Shape Modeling from Raw Images

Abstract

Statistical shape modeling is an important tool to characterize variation in anatomical morphology. Typical shapes of interest are measured using 3D imaging and a subsequent pipeline of registration, segmentation, and some extraction of shape features or projections onto some lower-dimensional shape space, which facilitates subsequent statistical analysis. Many methods for constructing compact shape representations have been proposed, but are often impractical due to the sequence of image preprocessing operations, which involve significant parameter tuning, manual delineation, and/or quality control by the users. We propose DeepSSM: a deep learning approach to extract a low-dimensional shape representation directly from 3D images, requiring virtually no parameter tuning or user assistance. DeepSSM uses a convolutional neural network (CNN) that simultaneously localizes the biological structure of interest, establishes correspondences, and projects these points onto a low-dimensional shape representation in the form of PCA loadings within a point distribution model. To overcome the challenge of the limited availability of training images with dense correspondences, we present a novel data augmentation procedure that uses existing correspondences on a relatively small set of processed images with shape statistics to create plausible training samples with known shape parameters. In this way, we leverage the limited CT/MRI scans (40-50) into thousands of images needed to train a deep neural net. After the training, the CNN automatically produces accurate low-dimensional shape representations for unseen images. We validate DeepSSM for three different applications pertaining to modeling pediatric cranial CT for characterization of metopic craniosynostosis, femur CT scans identifying morphologic deformities of the hip due to femoroacetabular impingement, and left atrium MRI scans for atrial fibrillation recurrence prediction.

Cascaded statistical shape model based segmentation of the full lower limb in CT

“…CNNs are versatile in their application domain, going way beyond image segmentation, as they can be trained for 2D to 3D registration purposes, disease recognition and quantification (Antony et al 2016). While such techniques are currently still being explored and developed, approaches such as SSIM segmentation techniques stand as valid alternatives as the computational cost for training is significantly lower and they are able to provide the necessary amount of labeled training data for advanced neural network applications (Bhalodia et al 2018).…”

The Use of Eye Tracking to Discern the Threshold at Which Metopic Orbitofrontal Deformity Attracts Attention

“…Currently, there is no gold standard for MCS diagnosis, and surgeons vary when deciding to intervene (Birgfeld et al, 2015; Yee et al, 2015; Jaskolka, 2017; Cho et al, 2018). Mathematical modeling has been used to objectively define pathologic versus physiologic closure of the metopic suture using angles (Kellogg et al, 2012), linear distance ratios (Gerety, 2017), shape analysis (Ruiz-Correa et al, 2008), and machine learning (Bhalodia et al, 2018; Cho et al, 2018). While objective analysis helps in defining the spectrum of severity, the inflection point between conservative management and surgery is still based on practitioner judgment.…”

A review of medical image data augmentation techniques for deep learning applications

“…This model can then be used to generate deformations within the range of plausible parameters. This approach has demonstrated an improvement in segmentation performance 123–127 …”