The highly sensitive automated detection and presentation of small lesions to physicians are expected to contribute to improving the accuracy and efficiency of diagnostic imaging. In a previous study, a fixed-size three-dimensional (3D) patch image was cut out of a test image, and lesions were detected and classified using a deep learning model for 3D images. However, there is a limitation in that the accuracy reduces as the lesion size decreases. Then, we assumed that using 2D images generated by multiscale patch sampling and minimum projection onto an orthogonal triplane, a deep learning model for 2D images used in general image recognition could detect small lesions with high accuracy. When the detection sensitivity of the proposed method was compared with that of the previous study's single scale equivalent by adjusting the false positive to be roughly 25 lesions per case, the detection sensitivity of the proposed method was higher, with an 0.72. These results indicate that the proposed method is benefical for realizing the computer-aided detection (CADe) of liver lesions from EOB-enhanced magnetic resonance imaging, which has not yet been established.

In this paper, we propose a relocalization method for keyframe-based SLAM that enables real-time and accurate recovery from tracking failures. To realize an AR-based application in a real world situation, not only accurate camera tracking but also fast and accurate relocalization from tracking failure is required. The previous keyframe-based relocalization methods have some drawbacks with regard to speed and accuracy. The proposed relocalization method selects two algorithms adaptively depending on the relative camera pose between a current frame and a target keyframe. In addition, it estimates a degree of false matches to speed up RANSAC-based model estimation. We present effectiveness of our method by an evaluation using public tracking dataset.