Joint Age Estimation and Gender Classification of Asian Faces Using Wide ResNet.

Two key facial features, age and gender, have been widely explored. Companies and organizations have investigated in related applications in several fields including insurance, retails, marketing, etc. It would bring tremendous benefit, which allow companies to easily identify their customer demographics. Several approaches have been proposed with remarkable results. However, because of the lack of open and multi-ethnic datasets, most modern age and gender estimating models were trained solely based on white people with Western facial features, and thus fall short with non-Caucasian people. In this paper, we developed an applicable Wide ResNet model to estimate the age and the gender of Asian faces. The model was trained with a newly improved Asian face database. The experiments have shown promising results, as it can match the performance of Microsoft's how-old API estimator in a specific dataset.

Fully automated MR liver volumetry using watershed segmentation coupled with active contouring.

PURPOSE: Our purpose is to develop a fully automated scheme for liver volume measurement in abdominal MR images, without requiring any user input or interaction. METHODS: The proposed scheme is fully automatic for liver volumetry from 3D abdominal MR images, and it consists of three main stages: preprocessing, rough liver shape generation, and liver extraction. The preprocessing stage reduced noise and enhanced the liver boundaries in 3D abdominal MR images. The rough liver shape was revealed fully automatically by using the watershed segmentation, thresholding transform, morphological operations, and statistical properties of the liver. An active contour model was applied to refine the rough liver shape to precisely obtain the liver boundaries. The liver volumes calculated by the proposed scheme were compared to the "gold standard" references which were estimated by an expert abdominal radiologist. RESULTS: The liver volumes computed by using our developed scheme excellently agreed (Intra-class correlation coefficient was 0.94) with the "gold standard" manual volumes by the radiologist in the evaluation with 27 cases from multiple medical centers. The running time was 8.4 min per case on average. CONCLUSIONS: We developed a fully automated liver volumetry scheme in MR, which does not require any interaction by users. It was evaluated with cases from multiple medical centers. The liver volumetry performance of our developed system was comparable to that of the gold standard manual volumetry, and it saved radiologists' time for manual liver volumetry of 24.7 min per case.

Liver Tumor Segmentation from MR Images Using 3D Fast Marching Algorithm and Single Hidden Layer Feedforward Neural Network.

Objective. Our objective is to develop a computerized scheme for liver tumor segmentation in MR images. Materials and Methods. Our proposed scheme consists of four main stages. Firstly, the region of interest (ROI) image which contains the liver tumor region in the T1-weighted MR image series was extracted by using seed points. The noise in this ROI image was reduced and the boundaries were enhanced. A 3D fast marching algorithm was applied to generate the initial labeled regions which are considered as teacher regions. A single hidden layer feedforward neural network (SLFN), which was trained by a noniterative algorithm, was employed to classify the unlabeled voxels. Finally, the postprocessing stage was applied to extract and refine the liver tumor boundaries. The liver tumors determined by our scheme were compared with those manually traced by a radiologist, used as the "ground truth." Results. The study was evaluated on two datasets of 25 tumors from 16 patients. The proposed scheme obtained the mean volumetric overlap error of 27.43% and the mean percentage volume error of 15.73%. The mean of the average surface distance, the root mean square surface distance, and the maximal surface distance were 0.58 mm, 1.20 mm, and 6.29 mm, respectively.

Fully automatic scheme for measuring liver volume in 3D MR images.

In this paper, a fully automatic scheme for measuring liver volume in 3D MR images was developed. The proposed MRI liver volumetry scheme consisted of four main stages. First, the preprocessing stage was applied to T1-weighted MR images of the liver in the portal-venous phase to reduce noise. The histogram of the 3D image was determined, and the second-to-last peak of the histogram was calculated using a neural network. Thresholds, which are determined based upon the second-to-last peak, were used to generate a thresholding image. This thresholding image was refined using a gradient magnitude image. The morphological and connected component operations were applied to the refined image to generate the rough shape of the liver. A 3D geodesic-active-contour segmentation algorithm refined the rough shape in order to more precisely determine the liver boundaries. The liver volumes determined by the proposed automatic volumetry were compared to those manually traced by radiologists; these manual volumes were used as a "gold standard." The two volumetric methods reached an excellent agreement. The Dice overlap coefficient and the average accuracy were 91.0 ±2.8% and 99.0 ±0.4%, respectively. The mean processing time for the proposed automatic scheme was 1.02 ±0.08 min (CPU: Intel, core i7, 2.8GHz), whereas that of the manual volumetry was 24.3 ±3.7 min (p < 0.001).

Hematocrit estimation using online sequential extreme learning machine.

Hematocrit is a blood test that is defined as the volume percentage of red blood cells in the whole blood. It is one of the important indicators for clinical decision making and the most effective factor in glucose measurement using handheld devices. In this paper, a method for hematocrit estimation that is based upon the transduced current curve and the neural network is presented. The salient points of this method are that (1) the neural network is trained by the online sequential extreme learning machine (OS-ELM) in which the devices can be still trained with new samples during the using process and (2) the extended features are used to reduce the number of current points which can save the battery power of devices and speed up the measurement process.

Computerized liver volumetry on MRI by using 3D geodesic active contour segmentation.

OBJECTIVE: Our purpose was to develop an accurate automated 3D liver segmentation scheme for measuring liver volumes on MRI. SUBJECTS AND METHODS: Our scheme for MRI liver volumetry consisted of three main stages. First, the preprocessing stage was applied to T1-weighted MRI of the liver in the portal venous phase to reduce noise and produce the boundary-enhanced image. This boundary-enhanced image was used as a speed function for a 3D fast-marching algorithm to generate an initial surface that roughly approximated the shape of the liver. A 3D geodesic-active-contour segmentation algorithm refined the initial surface to precisely determine the liver boundaries. The liver volumes determined by our scheme were compared with those manually traced by a radiologist, used as the reference standard. RESULTS: The two volumetric methods reached excellent agreement (intraclass correlation coefficient, 0.98) without statistical significance (p = 0.42). The average (± SD) accuracy was 99.4% ± 0.14%, and the average Dice overlap coefficient was 93.6% ± 1.7%. The mean processing time for our automated scheme was 1.03 ± 0.13 minutes, whereas that for manual volumetry was 24.0 ± 4.4 minutes (p < 0.001). CONCLUSION: The MRI liver volumetry based on our automated scheme agreed excellently with reference-standard volumetry, and it required substantially less completion time.

Computerized segmentation of liver in hepatic CT and MRI by means of level-set geodesic active contouring.

Computerized liver volumetry has been studied, because the current "gold-standard" manual volumetry is subjective and very time-consuming. Liver volumetry is done in either CT or MRI. A number of researchers have developed computerized liver segmentation in CT, but there are fewer studies on ones for MRI. Our purpose in this study was to develop a general framework for liver segmentation in both CT and MRI. Our scheme consisted of 1) an anisotropic diffusion filter to reduce noise while preserving liver structures, 2) a scale-specific gradient magnitude filter to enhance liver boundaries, 3) a fast-marching algorithm to roughly determine liver boundaries, and 4) a geodesic-active-contour model coupled with a level-set algorithm to refine the initial boundaries. Our CT database contained hepatic CT scans of 18 liver donors obtained under a liver transplant protocol. Our MRI database contains 23 patients with 1.5T MRI scanners. To establish "gold-standard" liver volumes, radiologists manually traced the contour of the liver on each CT or MR slice. We compared our computer volumetry with "gold-standard" manual volumetry. Computer volumetry in CT and MRI reached excellent agreement with manual volumetry (intra-class correlation coefficient = 0.94 and 0.98, respectively). Average user time for computer volumetry in CT and MRI was 0.57 ± 0.06 and 1.0 ± 0.13 min. per case, respectively, whereas those for manual volumetry were 39.4 ± 5.5 and 24.0 ± 4.4 min. per case, respectively, with statistically significant difference (p < .05). Our computerized liver segmentation framework provides an efficient and accurate way of measuring liver volumes in both CT and MRI.

Performance comparison of SLFN training algorithms for DNA microarray classification.

The classification of biological samples measured by DNA microarrays has been a major topic of interest in the last decade, and several approaches to this topic have been investigated. However, till now, classifying the high-dimensional data of microarrays still presents a challenge to researchers. In this chapter, we focus on evaluating the performance of the training algorithms of the single hidden layer feedforward neural networks (SLFNs) to classify DNA microarrays. The training algorithms consist of backpropagation (BP), extreme learning machine (ELM) and regularized least squares ELM (RLS-ELM), and an effective algorithm called neural-SVD has recently been proposed. We also compare the performance of the neural network approaches with popular classifiers such as support vector machine (SVM), principle component analysis (PCA) and fisher discriminant analysis (FDA).

An improvement of extreme learning machine for compact single-hidden-layer feedforward neural networks.

Recently, a novel learning algorithm called extreme learning machine (ELM) was proposed for efficiently training single-hidden-layer feedforward neural networks (SLFNs). It was much faster than the traditional gradient-descent-based learning algorithms due to the analytical determination of output weights with the random choice of input weights and hidden layer biases. However, this algorithm often requires a large number of hidden units and thus slowly responds to new observations. Evolutionary extreme learning machine (E-ELM) was proposed to overcome this problem; it used the differential evolution algorithm to select the input weights and hidden layer biases. However, this algorithm required much time for searching optimal parameters with iterative processes and was not suitable for data sets with a large number of input features. In this paper, a new approach for training SLFNs is proposed, in which the input weights and biases of hidden units are determined based on a fast regularized least-squares scheme. Experimental results for many real applications with both small and large number of input features show that our proposed approach can achieve good generalization performance with much more compact networks and extremely high speed for both learning and testing.