Moi Hoon Yap

A novel algorithm for initial lesion detection in ultrasound breast images

This paper proposes a novel approach to initial lesion detection in ultrasound breast images. The objective is to automate the manual process of region of interest (ROI) labeling in computer‐aided diagnosis (CAD). We propose the use of hybrid filtering, multifractal processing, and thresholding segmentation in initial lesion detection and automated ROI labeling. We used 360 ultrasound breast images to evaluate the performance of the proposed approach. Images were preprocessed using histogram equalization before hybrid filtering and multifractal analysis were conducted. Subsequently, thresholding segmentation was applied on the image. Finally, the initial lesions are detected using a rule‐based approach. The accuracy of the automated ROI labeling was measured as an overlap of 0.4 with the lesion outline as compared with lesions labeled by an expert radiologist. We compared the performance of the proposed method with that of three state‐of‐the‐art methods, namely, the radial gradient index filtering technique, the local mean technique, and the fractal dimension technique. We conclude that the proposed method is more accurate and performs more effectively than do the benchmark algorithms considered. PACS numbers: 87.57.Nk

Automatic Wrinkle Detection Using Hybrid Hessian Filter

Aging as a natural phenomenon affects different parts of the human body under the influence of various biological and environmental factors. The most pronounced changes that occur on the face is the appearance of wrinkles, which are the focus of this research. Accurate wrinkle detection is an important task in face analysis. Some have been proposed in the literature, but the poor localization limits the performance of wrinkle detection. It will lead to false wrinkle detection and consequently affect the processes such as age estimation and clinician score assessment. Therefore, we propose a hybrid Hessian filter (HHF) to cope with the identified problem. HHF is composed of the directional gradient and Hessian matrix. The proposed filter is conceptually simple, however, it significantly increases the true wrinkle localization when compared with the conventional methods. In the experimental setup, three coders have been instructed to annotate the wrinkle of 2D forehead image manually. The inter-reliability among three coders is 93 % of Jaccard similarity index (JSI). In comparison to the state-of-the-art Cula method (CLM) and Frangi filter, HHF yielded the best result with a mean JSI of 75.67 %. We noticed that the proposed method is capable of detecting the medium to coarse wrinkle but not the fine wrinkle. Although there is a gap between human annotation and automated detection, this work demonstrates that HHF is a remarkably strong filter for wrinkle detection. From the experimental results, we believe that our findings are notable in terms of the JSI.

Atlas-registration based image segmentation of MRI human thigh muscles in 3D space

Automatic segmentation of anatomic structures of magnetic resonance thigh scans can be a challenging task due to the potential lack of precisely defined muscle boundaries and issues related to intensity inhomogeneity or bias field across an image. In this paper, we demonstrate a combination framework of atlas construction and image registration methods to propagate the desired region of interest (ROI) between atlas image and the targeted MRI thigh scans for quadriceps muscles, femur cortical layer and bone marrow segmentations. The proposed system employs a semi-automatic segmentation method on an initial image in one dataset (from a series of images). The segmented initial image is then used as an atlas image to automate the segmentation of other images in the MRI scans (3-D space). The processes include: ROI labeling, atlas construction and registration, and morphological transform correspondence pixels (in terms of feature and intensity value) between the atlas (template) image and the targeted image based on the prior atlas information and non-rigid image registration methods.

Human Activity Recognition for Physical Rehabilitation

The recognition of human activity is a challenging topic for machine learning. We present an analysis of Support Vector Machines (SVM) and Random Forests (RF) in their ability to accurately classify Kinect kinematic activities. Twenty participants were captured using the Microsoft Kinect performing ten physical rehabilitation activities. We extracted the kinematic location, velocity and energy of the skeletal joints at each frame of the activity to form a feature vector. Principle Component Analysis (PCA) was applied as a pre-processing step to reduce dimensionality and identify significant features amongst activity classes. SVM and RF are then trained on the PCA feature space to assess classification performance, we undertook an incremental increase in the dataset size. We analyse the classification accuracy, model training and classification time quantitatively at each incremental increase. The experimental results demonstrate that RF outperformed SVM in classification rate for six out of the ten activities. Although SVM has performance advantages in training time, RF would be more suited to real-time activity classification due to its low classification time and high classification accuracy when using eight to ten participants in the training set.

Micro-Facial Movements: An Investigation on Spatio-Temporal Descriptors

This paper aims to investigate whether micro-facial movement sequences can be distinguished from neutral face sequences. As a micro-facial movement tends to be very quick and subtle, classifying when a movement occurs compared to the face without movement can be a challenging computer vision problem. Using local binary patterns on three orthogonal planes and Gaussian derivatives, local features, when interpreted by machine learning algorithms, can accurately describe when a movement and non-movement occurs. This method can then be applied to help aid humans in detecting when the small movements occur. This also differs from current literature as most only concentrate in emotional expression recognition. Using the CASME II dataset, the results from the investigation of different descriptors have shown a higher accuracy compared to state-of-the-art methods.

Processed images in human perception: A case study in ultrasound breast imaging

Two main research efforts in early detection of breast cancer include the development of software tools to assist radiologists in identifying abnormalities and the development of training tools to enhance their skills. Medical image analysis systems, widely known as Computer-Aided Diagnosis (CADx) systems, play an important role in this respect. Often it is important to determine whether there is a benefit in including computer-processed images in the development of such software tools. In this paper, we investigate the effects of computer-processed images in improving human performance in ultrasound breast cancer detection (a perceptual task) and classification (a cognitive task). A survey was conducted on a group of expert radiologists and a group of non-radiologists. In our experiments, random test images from a large database of ultrasound images were presented to subjects. In order to gather appropriate formal feedback, questionnaires were prepared to comment on random selections of original images only, and on image pairs consisting of original images displayed alongside computer-processed images. We critically compare and contrast the performance of the two groups according to perceptual and cognitive tasks. From a Receiver Operating Curve (ROC) analysis, we conclude that the provision of computer-processed images alongside the original ultrasound images, significantly improve the perceptual tasks of non-radiologists but only marginal improvements are shown in the perceptual and cognitive tasks of the group of expert radiologists.

A Short Review of Methods for Face Detection and Multifractal Analysis

The purpose of this paper is to present short reviews the face detection techniques and to study the effect of Multifractal analysis in detecting facial features. In reviewing the existing techniques, we have compared the performance of Nilsson et al’s algorithm, Haar Training implemented within Open Source Computer Vision Library (OpenCV) and Keinzle’s algorithm in face detection. Then, we have produced some experimental results suggesting that the Multifractal approach would help to extract key facial features of a human face. We conclude the paper with some discussions on how the work can be taken further.

Fully Automatic Lesion Boundary Detection in Ultrasound Breast Images

We propose a novel approach to fully automatic lesion boundary detection in ultrasound breast images. The novelty of the proposed work lies in the complete automation of the manual process of initial Region-of-Interest (ROI) labeling and in the procedure adopted for the subsequent lesion boundary detection. Histogram equalization is initially used to pre-process the images followed by hybrid filtering and multifractal analysis stages. Subsequently, a single valued thresholding segmentation stage and a rule-based approach is used for the identification of the lesion ROI and the point of interest that is used as the seed-point. Next, starting from this point an Isotropic Gaussian function is applied on the inverted, original ultrasound image. The lesion area is then separated from the background by a thresholding segmentation stage and the initial boundary is detected via edge detection. Finally to further improve and refine the initial boundary, we make use of a state-of-the-art active contour method (i.e. gradient vector flow (GVF) snake model). We provide results that include judgments from expert radiologists on 360 ultrasound images proving that the final boundary detected by the proposed method is highly accurate. We compare the proposed method with two existing state-of- the-art methods, namely the radial gradient index filtering (RGI) technique of Drukker et. al. and the local mean technique proposed by Yap et. al., in proving the proposed methods robustness and accuracy.

Benchmarking human motion analysis using kinect one: An open source dataset

There is a clear advantage to developing automated systems to detect human motion in the field of computer vision for applications associated with healthcare. We have compiled a diverse dataset of clinically-relevant motions using the Microsoft Kinect One sensor and release the dataset to the community as an open source solution for benchmarking detection, quantification and recognition algorithms. The dataset, namely Kinect 3D Active (K3Da), includes motions collected from young and older men and women ranging in age from 18-81 years. Participants performed standardised tests, including the Short Physical Performance Battery, Timed-Up-And-Go, vertical jump and other balance assessments which were recorded using depth sensor technology and extracted to generate motion capture data, sampled at 30 frames-per-second. Preliminary evaluations using Support Vector Machines, Random Forests, Artificial Neural Networks and Boltzmann Machines show age-related differences in many of the movements. These results demonstrate the relevance of the dataset to support benchmarking of algorithms associated and/or intended for use in a healthcare setting.

Exemplar-Based Human Action Recognition with Template Matching from a Stream of Motion Capture

Recent works on human action recognition have focused on representing and classifying articulated body motion. These methods require a detailed knowledge of the action composition both in the spatial and temporal domains, which is a difficult task, most notably under real-time conditions. As such, there has been a recent shift towards the exemplar paradigm as an efficient low-level and invariant modelling approach. Motivated by recent success, we believe a real-time solution to the problem of human action recognition can be achieved. In this work, we present an exemplar-based approach where only a single action sequence is used to model an action class. Notably, rotations for each pose are parameterised in Exponential Map form. Delegate exemplars are selected using k-means clustering, where the cluster criteria is selected automatically. For each cluster, a delegate is identified and denoted as the exemplar by means of a similarity function. The number of exemplars is adaptive based on the complexity of the action sequence. For recognition, Dynamic Time Warping and template matching is employed to compare the similarity between a streamed observation and the action model. Experimental results using motion capture demonstrate our approach is superior to current state-of-the-art, with the additional ability to handle large and varied action sequences.