Maximum Dice Similarity Coefficient Research Articles

Segmentation of Femoral Cartilage from Knee Ultrasound Images Using Mask R-CNN.

Segmentation of knee cartilage from Ultrasound (US) images is essential for various clinical tasks in diagnosis and treatment planning of Osteoarthritis. Moreover, the potential use of US imaging for guidance in robotic knee arthroscopy is presently being investigated. The femoral cartilage being the main organ at risk during the operation, it is paramount to be able to segment this structure, to make US guidance feasible. In this paper, we set forth a deep learning network, Mask R-CNN, based femoral cartilage segmentation in 2D US images for these types of applications. While the traditional imaging approaches showed promising results, they are mostly not real-time and involve human interaction. This being the case, in recent years, deep learning has paved its way into medical imaging showing commendable results. However, deep learning-based segmentation in US images remains unexplored. In the present study we employ Mask R-CNN on US images of the knee cartilage. The performance of the method is analyzed in various scenarios, with and without Gaussian filter preprocessing and pretraining the network with different datasets. The best results are observed when the images are preprocessed and the network is pretrained with COCO 2016 image dataset. A maximum Dice Similarity Coefficient (DSC) of 0.88 and an average DSC of 0.80 is achieved when tested on 55 images indicating that the proposed method has a potential for clinical applications.

Fully automatic detection of deep white matter T1 hypointense lesions in multiple sclerosis

A novel method is presented for fully automatic detection of candidate white matter (WM) T1 hypointense lesions in three-dimensional high-resolution T1-weighted magnetic resonance (MR) images. By definition, T1 hypointense lesions have similar intensity as gray matter (GM) and thus appear darker than surrounding normal WM in T1-weighted images. The novel method uses a standard classification algorithm to partition T1-weighted images into GM, WM and cerebrospinal fluid (CSF). As a consequence, T1 hypointense lesions are assigned an increased GM probability by the standard classification algorithm. The GM component image of a patient is then tested voxel-by-voxel against GM component images of a normative database of healthy individuals. Clusters (≥0.1 ml) of significantly increased GM density within a predefined mask of deep WM are defined as lesions. The performance of the algorithm was assessed on voxel level by a simulation study. A maximum dice similarity coefficient of 60% was found for a typical T1 lesion pattern with contrasts ranging from WM to cortical GM, indicating substantial agreement between ground truth and automatic detection. Retrospective application to 10 patients with multiple sclerosis demonstrated that 93 out of 96 T1 hypointense lesions were detected. On average 3.6 false positive T1 hypointense lesions per patient were found. The novel method is promising to support the detection of hypointense lesions in T1-weighted images which warrants further evaluation in larger patient samples.