Quantitative evaluation of bones and ligaments around knee joint from magnetic resonance imaging (MRI) often requires the boundaries of selected structures to be manually traced using computer software. It may take several hours to delineate all structures of interest in a three-dimensional (3D) dataset used for the evaluation. Thus, providing automated tools, which can delineate knee anatomical structures can improve productivity and efficiency in radiology departments. In recent years, 3D deep convolutional neural networks (3D CNN) have been successfully used for segmentation of knee bones and cartilage. However, the key challenge is segmentation of the anterior cruciate ligament (ACL) and the posterior cruciate ligament (PCL), due to high variability of intensities in the areas of pathologies such as ligament tear. In this approach, an open source 3D CNN is adapted for segmentation of knee bones and ligaments in the knee MRI. The segmentation accuracy of ACL and PCL is improved further by atlas based segmentation technique. The atlas mask is non-rigidly aligned with the patient image based on composite of rigid and deformable vector field derived between the bone masks in the atlas and corresponding segmented bone masks in the patient image. The level set functions corresponding to particular objects of interest of the deformed atlas are used to refine segmentation of the corresponding objects in the patient image. The accuracy of the proposed method is assessed using Dice coefficient score for 50 manual segmentations of bone, cartilage and ligaments comprising of both normal and knee injury cases. Our results show that the proposed approach offers a viable alternative to manual contouring of knee MRI volume by a human reader with improved accuracy compared to the 3D CNN.
A potential drawback of computer-aided diagnosis (CAD) systems is that they tend to capture the noise characteristics along with signal variations due to a limited number of sources used in training. This leads to a decrease in performance on data from different sources. The variations in scanner settings, device manufacturers and sites pose a significant challenge to the learning capabilities of the CAD systems like chest radiographs, also called Chest X-rays (CXR). In the proposed work, we investigate if preprocessing transformations like global normalization along with local enhancements are good to tackle the variability of data from multiple sources on a supervised CXR classification system. We also propose a detail enhancement filter to enhance both finer structures and opacities in CXRs. With the proposed preprocessing improvement, experiments were performed on 13,000 images across 3 public and one private data source using Dense Convolutional Network (DenseNet). The sensitivity at equal error rate (mean ± sd) improved from 0.888 ± 0.043 to 0.931 ± 0.030 by applying a combination of global histogram equalization with the proposed detail enhancement filter when compared to the raw images. We conclude that the proposed transformations are effective in improving the learning of CXRs from different data sources.