CR (Computed Radiography) is characterized by high sensitivity and wide dynamic range. Moreover, it has the advantage of being able to transfer exposed images directly to a computer-aided detection (CAD) system which is not possible using conventional film digitizer systems. This paper proposes a high-performance clustered microcalcification detection system for CR mammography. Before detecting and classifying candidate regions, the system preprocesses images with a normalization step to take into account various imaging conditions and to enhance microcalcifications with weak contrast. Large-scale experiments using images taken under various imaging conditions at seven hospitals were performed. According to analysis of the experimental results, the proposed system displays high performance. In particular, at a true positive detection rate of 97.1%, the false positive clusters average is only 0.4 per image. The introduction of geometrical features of each microcalcification for identifying true microcalcifications contributed to the performance improvement. One of the aims of this study was to develop a system for practical use. The results indicate that the proposed system is promising.

Face detection from cluttered images is very challenging due to the wide variety of faces and the complexity of image backgrounds. In this paper, we propose a neural network based approach for locating frontal views of human faces in cluttered images. We use a radial basis function network (RBFN) for separation of face and non-face patterns, and the complexity of RBFN is reduced by principal component analysis (PCA). The influence of the number of hidden units and the configuration of basis functions on the detection performance was investigated. To further improve the performance, we integrate the distance from feature subspace into the RBFN. The proposed method has achieved high detection rate and low false positive rate on testing a large number of images.

This paper presents a novel liver segmentation algorithm that achieves higher performance than conventional algorithms in the segmentation of cases with unusual liver shapes and/or large liver lesions. An L1 norm was introduced to the mean squared difference to find the most relevant cases with an input case from a training dataset. A patient-specific probabilistic atlas was generated from the retrieved cases to compensate for livers with unusual shapes, which accounts for liver shape more specifically than a conventional probabilistic atlas that is averaged over a number of training cases. To make the above process robust against large pathological lesions, we incorporated a novel term based on a set of “lesion bases” proposed in this study that account for the differences from normal liver parenchyma. Subsequently, the patient-specific probabilistic atlas was forwarded to a graph-cuts-based fine segmentation step, in which a penalty function was computed from the probabilistic atlas. A leave-one-out test using clinical abdominal CT volumes was conducted to validate the performance, and proved that the proposed segmentation algorithm with the proposed patient-specific atlas reinforced by the lesion bases outperformed the conventional algorithm with a statistically significant difference.

The skeleton and the skeleton function of an object are important representations for shape analysis and recognition. They contain enough information to recognize an object and to reconstruct its original shape. However, they are sensitive to distortion caused by rotation and noise. This paper presents another approach for binary object representation called a modified exoskeleton(mES) that combines the previously defined exoskeleton with the use of symmetric object whose dominant property is rotation invariant. The mES is the skeleton of a circular background around the object that preserves the skeleton properties including significant information about the object for use in object recognition. Then the matching algorithm for object recognition based on the mES is presented. We applied the matching algorithm to evaluate the mES against the skeleton obtained from using 4-neighbor distance transformation on a set of artificial objects, and the experimental results reveal that the mES is more robust to distortion caused by rotation and noise than the skeleton and that the matching algorithm is capable of recognizing objects effectively regardless of their size and orientation.

This paper presents a new analysis result of two-dimensional four-phase abdominal CT images using variational Bayesian mixture of ICA. The four-phase CT images are assumed to be comprised of several exclusive areas, and each area is generated by a set of corresponding independent components. ICA mixture analysis results show that the CT images could be divided into a set of clinically and anatomically meaningful components. Initial analysis of the independent components shows its promising prospects in medical image processing and computer-aided diagnosis.

This paper presents an ensemble learning algorithm for liver tumour segmentation from a CT volume in the form of U-Boost and extends the loss functions to improve performance. Five segmentation algorithms trained by the ensemble learning algorithm with different loss functions are compared in terms of error rate and Jaccard Index between the extracted regions and true ones.

During the development of a human embryo, the position of eyes moves medially and caudally in the viscerocranium. A statistical model of this process can play an important role in embryology by facilitating qualitative analyses of change. This paper proposes an algorithm to construct a spatiotemporal statistical model for the eyeballs of a human embryo. The proposed modeling algorithm builds a statistical model of the spatial coordinates of the eyeballs independently for each Carnegie stage (CS) by using principal component analysis (PCA). In the process, a q-Gaussian distribution with a model selection scheme based on the Aaike information criterion is used to handle a non-Gaussian distribution with a small sample size. Subsequently, it seamlessly interpolates the statistical models of neighboring CSs, and we present 10 interpolation methods. We also propose an estimation algorithm for the CS using our spatiotemporal statistical model. A set of images of eyeballs in human embryos from the Kyoto Collection was used to train the model and assess its performance. The modeling results suggested that information geometry-based interpolation under the assumption of a q-Gaussian distribution is the best modeling method. The average error in CS estimation was 0.409. We proposed an algorithm to construct a spatiotemporal statistical model of the eyeballs of a human embryo and tested its performance using the Kyoto Collection.