Automatic Segmentation of Bone Marrow Lesions on MRI Using a Deep Learning Method.

Bone marrow lesion (BML) volume is a potential biomarker of knee osteoarthritis (KOA) as it is associated with cartilage degeneration and pain. However, segmenting and quantifying the BML volume is challenging due to the small size, low contrast, and various positions where the BML may occur. It is also time-consuming to delineate BMLs manually. In this paper, we proposed a fully automatic segmentation method for BMLs without requiring human intervention. The model takes intermediate weighted fat-suppressed (IWFS) magnetic resonance (MR) images as input, and the output BML masks are evaluated using both regular 2D Dice similarity coefficient (DSC) of the slice-level area metric and 3D DSC of the subject-level volume metric. On a dataset with 300 subjects, each subject has a sequence of 36 IWFS MR images approximately. We randomly separated the dataset into training, validation, and testing sets with a 70%/15%/15% split at the subject level. Since not every subject or image has a BML, we excluded the images without a BML in each subset. The ground truth of the BML was labeled by trained medical staff using a semi-automatic tool. Compared with the ground truth, the proposed segmentation method achieved a Pearson's correlation coefficient of 0.98 between the manually measured volumes and automatically segmented volumes, a 2D DSC of 0.68, and a 3D DSC of 0.60 on the testing set. Although the DSC result is not high, the high correlation of 0.98 indicates that the automatically measured BML volume is strongly correlated with the manually measured BML volume, which shows the potential to use the proposed method as an automatic measurement tool for the BML biomarker to facilitate the assessment of knee OA progression.

Automatic Measuring of Finger Joint Space Width on Hand Radiograph using Deep Learning and Conventional Computer Vision Methods.

Hand osteoarthritis (OA) severity can be assessed visually through radiographs using semi-quantitative grading systems. However, these grading systems are subjective and cannot distinguish minor differences. Joint space width (JSW) compensates for these disadvantages, as it quantifies the severity of OA by accurately measuring the distances between joint bones. Current methods used to assess JSW require users' interaction to identify the joints and delineate initial joint boundary, which is time-consuming. To automate this process and offer a more efficient and robust measurement for JSW, we proposed two novel methods to measure JSW: 1) The segmentation-based (SEG) method, which uses traditional computer vision techniques to calculate JSW; 2) The regression-based (REG) method, which is a deep learning approach employing a modified VGG-19 network to predict JSW. On a dataset with 3,591 hand radiographs, 10,845 DIP joints were cut as regions of interest and served as input to the SEG and REG methods. The bone masks of the ROI images generated by a U-Net model were sent as input in addition to the ROIs. The ground truth of JSW was labeled by a trained research assistant using a semi-automatic tool. Compared with the ground truth, the REG method achieved a correlation coefficient of 0.88 and mean square error (MSE) of 0.02 mm on the testing set; the SEG method achieved a correlation coefficient of 0.42 and MSE of 0.15 mm. Results show the REG method has promising performance in automatic JSW measurement and in general, Deep Learning approaches can facilitate the automatic quantification of distance features in medical images.