The reading workload for radiologists is increasing because the numbers of examinations and images per examination are increasing due to the technical progress on imaging modalities such as computed tomography and magnetic resonance imaging. A computer-assisted detection (CAD) system based on machine learning is expected to assist radiologists. The preliminary results of a multi-institutional study indicate that the performance of the CAD system for each institution improved using training data of other institutions. This indicates that transfer learning may be useful for developing the CAD systems among multiple institutions. In this paper, we focus on transfer learning without sharing training data due to the need to protect personal information in each institution. Moreover, we raise a problem of negative transfer in CAD system and propose an algorithm for inhibiting negative transfer. Our algorithm provides a theoretical guarantee for managing CAD software in terms of transfer learning and exhibits experimentally better performance compared to that of the current algorithm in cerebral aneurysm detection.

In this paper, we propose a blind frame-by-frame dereverberation using correlation coefficients considering periodicity of voiced speech. In the proposed method, the reference signal is derived by reducing speech periodicity from recovered signal of previous frame. The reference signal is used in order to estimate the impulse response in each frame. The recovered signal is obtained by using the estimated impulse response. From the investigation of performance evaluation, we show the proposed method is more effective than the conventional method in reducing reverberation without introducing speech distortion

The adaptive Volterra filter (AVF), based on the discrete Volterra series, has been proposed to identify a nonlinear system, remove nonlinear distortion, etc. In AVF algorithms, the fast transversal filter (FTF) algorithm (Junghsi Lee and Mathews, V.J., IEEE Trans. Acoust., Speech, Signal Process., vol.41, no.3, p.1087-102, 1993), which has O(N3) multiplications per iteration, has perfect convergence properties, where N represents the memory span of the nonlinear system model. However, in practical applications, we require more reduction in computational complexity. Until now, although the technique of dividing AVF by the order of the Volterra kernel has been proposed (Takeichi, K. and Furukawa, T., 16th Digital Signal Process. Symp., 2001), it can only reduce calculations in the second-order AVF by 20%. The paper presents a design method for a second-order parallel adaptive Volterra filter (PAVF) to decrease calculations further. Finally, to illustrate the effectiveness of the proposed method, the results of computer simulation with the FTF algorithm are presented