Yuki Wakata

Automated fuzzy logic based skull stripping in neonatal and infantile MR images

Automated morphometric analysis using human brain magnetic resonance (MR) images is an effective approach to investigate the morphological changes of the brain. However, even though many methods for adult brain have been studied, there are few studies for infantile brain. Same as the adult brain, it is effective to measure cerebral surface and for quantitative diagnosis of neonatal and infantile brain diseases. This article proposes a skull stripping method that can be applied to the neonatal and infantile brain. The proposed method can be applied to both of T1 weighted and T2 weighted MR images. First, the proposed method estimates intensity distribution of white matter, gray matter, cerebrospinal fluid, fat, and others using a priori knowledge based Bayesian classification with Gaussian mixture model. The priori knowledge is embedded by representing them with fuzzy membership functions. Second, the proposed method optimizes the whole brain by using fuzzy active surface model, which evaluates the deforming model with fuzzy rules. The proposed method was applied to 26 neonatal and infantile subjects between −4 weeks and 4 years 1 month old. The results showed that the proposed method stripped skull well from any neonatal and infantile MR images.

Neonatal Brain Age Estimation Using Manifold Learning Regression Analysis

The neonatal cerebral disorders severly languish the quality of life (QOL) of patients and also their families. It is required to detect and cure in their early stage for the sake of decreasing the degree of symptoms. However, it is difficult to evaluate neonatal brain disorders based on morphological analysis because the neonatal brain grows quickly and the brain development progress is different from person to person. Previously, we proposed a method of calculating growth index using Manifold learning. The growth index is effective to evaluate the brain morphological development progress, although, it does not directly correspond to the brain development delay. To evaluate brain development delay, this paper proposes an estimation method of neonatal brain age using Manifold learning, principal component analysis, and multiple regression model. The regression model is trained using a 4-D standard brain, which is constructed using training subjects with growth index. To evaluate the proposed method, we constructed a multiple regression model using 11 normal subjects (revised age: 0-4 month old), and estimated brain age of 4 normal subjects. And, we estimated brain age of 4 abnormal subjects to evaluate the detection accuracy of brain development abnormality. The results showed that the method found the differences of brain development for abnormal subjects.

Fuzzy object growth model for newborn brain using Manifold learning

To develop a computer-aided diagnosis system for neonatal cerebral disorders, some literatures have shown atlas-based methods for segmenting parenchymal region in MR images. Because neonatal cerebrum deforms quickly by natural growth, we desire an atlas growth model to improve the accuracy of segmenting parenchymal region. This paper proposes a method for generating fuzzy object growth model (FOGM), which is an extension of fuzzy object model (FOM). FOGM is composed of some growth index weighted FOMs. To define the growth index, this paper introduces two methods. The first method calculates the growth index from revised age. Because the growth index will be different from person to person even through the same age, the second method estimates the growth index from cerebral shape using Manifold learning. To evaluate the proposed methods, we segment the parenchymal region of 16 subjects (revised age; 0-2 years old) using the synthesized FOGM. The results showed that FOGM was superior to FOM, and the Manifold learning based method gave the best accuracy. And, the growth index estimated with Manifold learning was significantly correlated with both of revised age and cerebral volume (p<;0.001).

Neonatal brain MRI normalization with 3-D cerebral sulci registration

MR image registration (IR) has been used in brain function analysis, voxel-based-morphometry, etc. The conventional IR methods mainly use MR signal based likelihood. However, they cannot prevent miss registration of different gyri because they do not evaluate correspondence of sulci. Also, we cannot directly apply methods for adult brain to neonatal brain because there are large differences in MR signal and sulcal width. This paper focuses on neonatal brain MR images, and introduces a new feature called sulcal-distribution index (SDI), which is calculated from MR signal around the cerebral surface. Next, this paper proposes a non-rigid 3-D IR method based on a flattening with SDI. The likelihood used is mutual information of SDI. The new method evaluates the correspondence of cerebral sulci. And, the method will be effective for neonatal brain in which the accurate delineation of cerebral surface is difficult because the method evaluates the MR signal around the cerebral surface. Results in 3 neonates (modified age; 3-5 weeks) showed that the method registered one brain with the other brain successfully.

Neonatal brain MR image segmentation based on system-of-systems in engineering technology

Measurement of cerebral volume and surface area using magnetic resonance (MR) image is effective for quantitative diagnosis of cerebral diseases. The measurement should require a brain segmentation process. Although many approaches for adult brain have been studied, there are few studies for neonatal brain. This study proposes a brain segmentation method for neonatal brain. Based on system of systems engineering technology, the proposed approach is composed from two systems; automated fuzzy logic based skull striping (AFSS) system and contour shape based modeling (CSM) system. AFSS segments the cerebral region based on Bayesian classification with Gaussian mixture model. CSM evaluates the skull stripping result of AFSS, and updates AFSS system parameters. Experimental results in 34 neonates (revised age between −2 weeks 1 day and 2 years 5 months) showed that the proposed approach segmented the brain region with sensitivity of 98.1% and false-positive rate of 27.9%.

Neonatal brain segmentation using 4-D fuzzy object model

Brain region segmentation in neonatal magnetic resonance (MR) images is an essential task for computer-aided diagnosis of neonatal brain disorders using MR images. We have proposed a neonatal brain segmentation method using a fuzzy object model (FOM), which represents a prior knowledge of brain shape and location. The FOM is constructed from multiple neonatal brain MR images whose revised age was between 0 and 4 weeks. The method segmented the brain region with a good accuracy for subjects whose age matches of the training data set. To enhance the method, we need multiple FOMs for each age. The other solution is to develop a growable model. This paper introduces 4-D FOM and applies it to neonatal brain segmentation. This paper introduces a neonatal brain segmentation method using 4-D FOM. The proposed method consists of three components. The first part proposes a method for estimating the brain development progress, called growth index in this study, from MR images based on Manifold learning. The second part shows a procedure for generating 4-D FOM using the estimated growth index. The third part is to segment brain region based on fuzzy-connectedness image segmentation using 4-D FOM. The proposed method was applied to 16 neonatal subjects. The results show that 4-D FOM is superior to stable 3-D FOM for segmenting neonatal brain region from MR images.

A priori knowledge based deformable surface model for newborn brain MR image segmentation

Newborn brain MR image segmentation is a crucial procedure for computer-aided diagnosis of brain disorders using MR images. We have previously proposed an automated method for segmenting parenchymal region. The method is based on a fuzzy rule based deformable surface model. In order to improve the segmentation accuracy, this paper introduces a priori knowledge represented by fuzzy object radial model called FORM. The FORM is generated from learning data set, and represents knowledge on shape and MR signal of parenchymal region in MR images. The performance of the proposed method has been validated by using 12 newborn volunteers whose revised age was between -1 month and 1 month. In comparison with the previous method, the proposed method showed the best performance, and the sensitivity was 87.6 % and false-positive-rate (FPR) was 5.68 %. And, leave-one-out cross validation (LOOCV) was conducted to evaluate the robustness. Mean sensitivity and FPR in LOOCV was 86.7 % and 12.1 %.

ICP based neonatal brain MRI normalization method

Magnetic resonance (MR) images are widely used to diagnose cerebral diseases. The diseases may deform the brain shape, and the deformed region differs among types of diseases. To evaluate the brain shape deformation, MR image registration (IR) has been used. There are some IR methods for brain MR images but they mainly use MR signal based likelihood. We cannot directly apply methods for adult brain to neonatal brain because there are large differences in MR signal distribution and brain shape. This paper focuses on neonatal brain MR images, and introduces a sulcus extraction method using Hessian matrix based on a feature called sulcal-distribution index (SDI). SDI is calculated from MR signal on the cerebral surface. Next, this paper proposes an iterative closest point (ICP) based brain shape registration method using the extracted sulci. The proposed method will be effective for neonatal brain in which the accurate delineation of cerebral surface is difficult because the method evaluates the correspondence of cerebral sulci distribution. Results in seven neonates (modified age was between 3 weeks and 2 years) showed that the method registered one brain with the other brain successfully.

Cerebral surface extraction based on particle method in neonatal MR images

It is effective to evaluate magnetic resonance (MR) images for diagnosing neonatal cerebral disorders because they often accompany the deformation of the brain shape. However, there are many difficulties when radiologists manually extract cerebral surface from the MR images. Therefore, it requires to extract the cerebral surface from neonatal MR images automatically. There are many methods to extract cerebral surface from adult MR images, but there are few methods for neonatal MR images. This paper proposes a new extraction method based on particle method. The proposed method introduces three kinds of particles corresponding to cerebrospinal fluid, gray matter and white matter. First, particles are assigned according to the cerebral shape. Second, particles are moved to form the homogeneous particles, and are transited to the other particles with respect to MR signal. The proposed method was applied to neonatal MR images. The results showed that the proposed method extracted cerebral surface with high accuracy.