Yutaka Hata

This paper presents a feature extraction model 'MAGNET' to find the deepest point of branched sulcus. Our model demonstrates magnetic principle and consists of four types of ideal magnetic poles: an N-pole and three S-poIes. According to attractive or repulsive Coulomb forces between their poles, one of the S-poles is pushed to the deepest point of the sulcus. First, we explain our model on the simple sulcus model. Second, we apply it to the sulcus with implicit branches. Our model can detect the target points in every branch. Then an example to realize the model on a synthetic image is introduced. We apply our model to human brain MR images and human foot CT images. Experimental results on human brain MR images show that our method enable us to successfully detect the points.

PURPOSE: To determine the features of cortical atrophy in frontotemporal dementia (FTD) and Alzheimer disease by using a hemispheric surface display generated with magnetic resonance (MR) images. MATERIALS AND METHODS: The extent of cortical atrophy was evaluated with automated MR hemispheric surface display and volumetry in 18 patients with FTD and in 18 matched patients with Alzheimer disease. Results were compared with those in 18 healthy, matched control subjects. RESULTS: Most cortical regions, were significantly atrophic in FTD and Alzheimer disease. The frontal and anterior temporal lobes were significantly more atrophic in FTD than in Alzheimer disease. The mean hemispheric-to-intracranial volume ratio in patients with FTD (56.2%) and those with Alzheimer disease (58.4%) was significantly smaller than the ratio in the control subjects (66.0%). Asymmetry of hemispheric volume was significantly larger in the FTD group than in Alzheimer disease and control groups. CONCLUSION: Cortical atrophy in FTD is more widespread than was previously thought. Asymmetric frontal and anterior temporal atrophy is a distinctive feature of FTD and distinguishes it from Alzheimer disease. Hemispheric surface display is a useful complement to tomographic images and is useful for the evaluation of focal cortical atrophy in degenerative dementia, especially FTD.

This paper proposes a segmentation method based on Kleene Algebra. for an input image including some regions of interests (ROIs for short), consider three segmented states: Shortage, correct, Excess for the target region on applying segmentation method based on standard intensity thresholding. For the target image, we do thresholding to each of ROIs, then to derive all "Correct" for ROIs, unate function (one model of Kleene Algebra) based approach proposes to find all "Correct" states. However, the method is not complete for some cases, that is, correctly segmented ratio is about 70% for three and four ROI segmentation. For the failed cases, it is proved that Brzozowski operations are provided to completely find all "Correct" states. The experimental results on a humanitarian MR image and a foot CT image show that our method can correctly segment the ROI.

One of the most serious problems in medical science is an increment of Alzheimer's disease. It is known that the patient's brain atrophy is a result of neural cell loss. It is useful for the diagnosis of Alzheimer's disease to measure the volumes of the brain portions and to display them. We can obtain anatomical information from 2D slice images produced by MRI. We propose a computer-aided system for the diagnosis of Alzheimer's disease. The system consists of: (1) extraction of the portions from MRI data, (2) measurement of the volumes of the portions and then displaying them, and (3) user interface for a medical doctor. In this paper we describe procedures for the above. For extraction of the brain portions we propose the method based on standard region growing algorithm and the method of figure decomposition using the distance value. Comparison of the volumes of our extracted portions with volumes manually measured by a physician shows that the error rate, on the average, is 1.74% for 48 MRI data. We also discuss the 3D display, the measuring range, and the construction of the user interface for a physician. © 1997 Scripta Technica, Inc.

MR脳画像同士の位置あわせは, 同じ脳疾患を持つ複数の患者の脳画像を重ねあわせることによって, 濃度値などの情報から共通する部位の異常を視覚的に捉えることが目的である. しかし, MR脳画像は大局的に見ると一様に見える部分であっても, 局所的には個々により異なる部分が多く, このことがMR脳画像同士の位置あわせを困難としている. 従来, 画像同士の位置あわせの方法がいくつか紹介されている. それらは, 画像の濃度値による相関関係による方法, 輪郭の曲率や変極点の位置情報による方法などである. 相関関係による方法のような対象画像の濃度値をもとに行われる方法は, 濃度値情報を取得するというMR脳画像を位置あわせする目的から使用することができない. また, これらの方法は対象画像の局所的な一致度に基づいて全体の位置あわせを行うものであり, 局所的に多くの部分が異なる脳画像の位置あわせに適用するのは困難である. 本文では, ラバー (ゴム) をモデル化した物理モデルを提案し, このラバーモデルを用いてMR脳画像の位置あわせを行う. ラバーモデルとは, 平面状のゴムを想定し, このゴムに歪みが加わったときのゴム全体の力学的な振る舞いをモデル化したものである. ラバーモデルでは局所的な変位が画像全体に影響を及ぼす構造をしており, 全体として協調しながら画像全体が移動するため, 大局的な位置あわせが可能である.

近年、聴覚障害者と健聴者とのコミュニケーションの支援を目的とし、手話を日本語へ翻訳する研究が注目されている。我々は、日本語の五十音に相当する手話の指文字のうち、表現に動きを用いない 41 文字を対象として画像処理を用いた聴覚障害者のためのインタフェースについて検討を行ってきた[1]。今回は手形状の伸展指特徴による指文字の分類手法を示し、手と照明との位置関係による画像不良に対処するために用いたグローブの有用性について検討した結果を報告する。

現在, ニューラルネットワークは画像処理やシステム制御などのさまざまな分野に応用され, その有効性が示されている. しかしこれらの研究のほとんどが, 与えた入力に対する出力値や学習に対する反応のみに注目しており, ネットワークの内の各ユニットがどのような機能を実現しているかには, あまり注意が払われていない. 与えられた入出力パターンを学習したネットワークは, 設計者の望む機能を実現しており, 入出力パターンから何らかの知識を獲得しているはずである. しかし, ネットワークが獲得した知識は複雑で, 人間に理解できる形で導出することは非常に困難である. 本研究では, 加重和のかわりにmin演算やmax演算等を用いることで, ニューロンモデルと知識の対応を明確にし, ネットワークから知識を導出する. 同時に, ネットワークの枝刈りを行い中間層ユニットを削減することで, 知識の最適化を行う. この概念を論理回路の設計に適用する.

FPGA (Field Programable Gate Array) を設計する際, 与えられた多変数関数を数個の部分関数に分解する手続きである関数分解は, 回路コスト削除のための必要不可欠な手法である. これまでは, 真理値表を用いた関数分解法が多く用いられてきたが, これは, 複数段の回路構成には適用が困難だった. 本文では, 多段回路設計効率化のために, Multiple-valued Decision Diagram (MDD) 使用による関数分解法を提案する. MDDは実用的な多値論理関数を計算機内部で効率よく表現できるグラフとして知られており, 関数分解にも有効である期待できる. 本方法の評価は, 多値算術関数に対する適用結果を基に行う.

近年, MRやCT装置などの医療機械の発展によって, 人体内部が簡単に撮影できるようになった. また, 最近のコンピュータの高速化, メモリの大容量化によって, それらの情報を解析・可視化し, 評価することが容易になってきた. 一般的には, 興味ある部位をその医療画像から抽出し, それを評価・可視化・解析することが必要である. その際, 抽出目的部位の濃度幅が異なるため単純なしきい値処理では抽出できないのが現状である. そこで, 本稿では, 複数存在する抽出目的部位毎にしきい値処理を行って得られた画像群を, クリーネ代数を用いて統合する方法を示す. 更に, クリーネ代数を用いても解が得られない場合に対して, Brzozowskiの演算を用いることで解が得られることを示す.

Recently, with rapid advances in VLSI(Very Large Scale Integration)technology, WSI(Wafer Scale Integration)that is composed of many multiprocessors arranged massively on the warfer has received considerable attention. In this paper, we propose the network system that has features of mesh-connected multiprocessor system and those of tree-connected multiprocessor system. In addition, we show the reconfiguring method for our system that has several faulty PEs(Processing Elements). We define the equation that computes the reliability of our system. The equation has a variable whose value is equal to the reliability of each processing element and it is used in order to evaluate our system. Finally, we compare our system with the conventional mesh-connected multiprocessor system and discuss the advantages of our system.

This paper investigates a novel technique for constructing and evaluating neuro-models of pattern generation. The modeling approach consists of two steps : first a neural network is trained to learn a core concept expressed on fuzzy logic, and then the trained network is augmented with additional processing nodes and connections.The augmented netwoek is then tested on its ability to solve problems related to the core concept for which its was trained. We present results from applying our model to monotone-function generation, decision making and image generation.

This paper proposes an approach of fuzzy logic to 3D foot CT image segmentation. We show a fuzzy knowledge representation method to represent the knowledge needed to segment the foot bone into their ossa. In it we consider boundary surface knowledge, intensity knowledge and adjacency knowledge. They are expressed by fuzzy if-then rules and compiled to a total degree as the measure of segmentation. The degree is evaluated in region growing technique and which segments the foot bone. The experimental result shows that our method extract the ossa precisely.

In this paper, we show a medical image segmentation for MR angiography images. In it, we introduce a cluster analysis technique aided by fuzzy logic. Our method classifies the clusters into the blood vessels and the others by comparing the intensity histogram with predefined models. In the experimental results, we show the target MIP and volume rendering images of the segmented region. They are useful for evaluating the lesions such as the areury.

This paper proposes an approach of fuzzy logic to 3D MR image segmentation. We show a fuzzy knowledge representation method to represent the knowledge needed to segment the target portions, and apply our method to 3D MR human brain image segmentation. In it we consider position knowledge, boundary surface knowledge and intensity knowledge. They are expressed by fuzzy if-then rules and compiled to a total degree as the measure of segmentation. The degree is evaluated in region growing technique and which segments the whole brain region into the left cerebral hemisphere, the right cerebral hemisphere, the cerebellum and the brain stem. The experimental result on 36 MR voxel data shows that our method extracted the portions precisely.

This paper presents a robust automatic threshold finding method for the human brain MR image segmentation. The method is based on fuzzy information granulation shown by Zadeh. The human brain MR image consists of several parts; the gray matter, white matter, cerebrospinal fluid and so on. By treating their parts as the fuzzy granules in the gray level histogram of the image and developing fuzzy matching technique, we can find required thresholds and can segment the brain region from the MR image. An experiment is done on 50 gray level histograms of the human brain MR volumes. To evaluate our method, we extract the brain region using the obtained thresholds. A comparison of the obtained region with canonical atlas images shows that our method find the thresholds of the gray matter and white matter correctly.

in this paper, we discuss problems in design and fault masking of multiple-valued cellular arrays where basic cells having simple switch functions are arranged iteratively. The stuck-at faults of switch cells are assumed to be fault models. First, we introduce a universal single-level array and derive the ratio of the number of single faults whose influence can be masked to the total number of single faults. Next, we propose a universal two-level array that outputs correct values even if single Faults occur in it and derive the ratio of the number of double faults whose influence can be masked compared to the total number of double faults. By evaluating the universal single-level array and the universal two-level array from the viewpoints of design and fault masking, we show that the latter is superior to the former. Finally, we compare our universal two-level array with formerly presented arrays in order to demonstrate the advantages of our universal two-level array.

脳外科分野において手術計画を立てるとき,腫瘍等対象部位から脳表面のしわの奥までの距離が重要な情報となる.MRIは脳内軟部組織を高いコントラストで撮影できるので,それにより脳形状等の複雑な情報が正確に得られる.本文では,MR断層画像を基に構成した3次元画像を平面で切断し,切断面の断層像から脳内部位-脳表面しわ間の距離を測定する方法を提案する.

フェイルセイフ論理系の一つとして,1/2をとる入力が一つでもあれば出力値が1/2となるC型フェイルセイフが提案されている.本文ではこの概念を基本とし,故障が生じれば,正常値または安全状態1/2を出力する多値セルラアレーを提案する.

This paper describes Kleenean coefficients that are a subset of Kleenean functions for use in representing multiple-valued logic functions. A conventional multiple-valued sum-of-products expression uses product terms that are the MIN of literals and constants. In this paper, a new sum-of-products expression is allowed to sum product terms that also include variables and complements of variables. Since the conventional sum-of-products expression is complete, so also is the augmented one. A minimization method of the new expression is described based on the binary Quine-McCluskey algorithm. The result of computer simulation shows that a saving of the number of implicants used in minimal expressions by approximately 9% on the average can be obtained for some random functions. A result for some arithmetic functions shows that the minimal solutions of MOD radix SUM, MAX and MIN functions require much fev er implicants than those of the standard sum-of-products expressions. Thus, this paper clarifies that the new expression has an advantage to reduce the number of implicants in minimal sum-of-products expressions.

One of the most serious problems in medical science is an increment of the Alzheimer's disease. It is known that the patients have the brain atrophy because of the neural cell loss. It is useful for the diagnosis of the Alzheimer's disease to measure the volumes of and to display the brain portions. We can obtain anatomical information from 2D slice images produced by MRI system. We construct the 3D voxel data from these images. We propose a 3D computer aided system for the diagnosis of the Alzheimer's disease. The system consist of three parts: (1) extraction of the portions from MRI data based on region growing method, (2) measure the volumes of and display the portions, and (3) user interface for a easy manipulation of a medical doctor. In this paper we describe procedures to carry out them. The comparison of the obtained volumes with manually obtained volumes by a doctor shows that the error ratio is about 2% for 48 MRI data.

This paper investigates a novel technique for constructing and evaluating neurofuzzy models of human pattern generation. The modeling approach consists of two steps: first a neural network is trained to learn a core concept and then the trained network is augmented with additional processing nodes and connections. The augmented network is then tested on its ability to solve problems related to the core concept for which it was trained. We present results from applying our model to image generation and decision making.

In this paper, we propose a method to design ternary cellular arrays by using Ternary Decision Diagrams (TDD's). Our cellular array has a rectangular structure composed of ternary switch cells. The ternary functions represented by TDD's are realized by mapping the TDD's to the arrays directly. That is, both the nodes and the edges in the TDD are realized by some sets of the cells. Since TDD's can represent easily multiple-output functions without large memory requirements, our arrays are suitable for the realization of multiple-output functions. To evaluate our method, we apply our method to some benchmark circuits, and compare our arrays with the ternary PLA's. The experimental results show that our arrays have the advantage for their sizes, especially in the realization of symmetric functions. The results also clarify that the size of our arrays depends on the size of TDD's.

In this letter we propose an interpolation technique for low-quality fingerprint images for highly reliable feature extraction. To improve the feature extraction rate, we extract fingerprint features by referring to both the interpolated image obtained by using a directional Laplacian filter and the high-contrast image obtained by using histogram equalization. Experimental results show the applicability of our method.

This paper proposes a repairable and diagnosable k-valued cellular array. We assume a single fault, i.e., either stuck-at-0 fault or stuck-at-(k - 1) fault of switches occurs in the array. By building in a duplicate column iteratively, when a stuck-at-(k - 1) fault occurs in the array, the fault never influences the output of the array. That is, we can construct a fault-tolerant array for the stuck-at-(k - 1) fault. While, for the stuck-at-0 fault, the diagnosing method is simple and easy because we don't have to diagnose the stuck-at-(k - 1) fault. Moreover, our array can be repaired easily for the fault. The comparison with other rectangular arrays shows that our array has advantages for the number of cells and the cost of the fault diagnosis.

This paper proposes a multi-font Japanese character recognition method using direction frequency distribution among pixels that construct character-lines(Directional Feature)and spectra around a character(Peripheral Feature).These two features seem to be a quality that they are stable against the distortion of character forms,changes of shift and so on.Besides,we discuss a scheme for placing small areas used in directional feature extraction and that for placing partitions used in peripheral feature extraction. On the basis of the above discussion,we propose a feature extraction method that would be useful for more precision recognition.Finally,some experiments were done for 10215 Japanese characters(Hiragana,Katakana and Chinese character)of five fonts(Mincho,Socho,Gothic,Kyokasho and Square style).As a result, the recognition rate overheads 99.22%.

A devlopment of CT and MR scanners enables us to obtain the human internal images.However,it is difficult to extract some volumes of various brain parts from the images obtained from them, because the images are 2D slice images.Therefore it is necessary to do the medical image processing as a technique that makes us understand the obtained images.This paper proposes an automatic extraction method for cerebral region based on region growing.This method enables us to obtain the volumes of various brain parts. Moreover,it shows a method to do 3D display of the extracted shape.

This paper proposes a multi-font Japanese character recognition method using direction frequency distribution among pixels that construct character-lines (Directional Feature) and spectra around a character (Peripheral Feature). These two features seem to be a quality that they are stable against the distortion of character forms, changes of shift and so on. Besides, we discuss a scheme for placing small areas used in directional feature extraction and that for placing partitions used in peripheral feature extraction. On the basis of the above discussion, we propose a feature extraction method that would be useful for more precision recognition. Finally, some experiments were done for 10215 Japanese characters (Hiragana. Katakana and Chinese character) of five fonts(Mincho, Socho, Gothic, Kyokasho and Square style). As a result, the recognition rate overheads 99.22%.

A devlopment of CT and MR scanners enables us to obtain the human internal images. However, it is difficult to extract some volumes of various brain parts from the images obtained from them, because the images are 2D slice images. Therefore it is necessary to do the medical image processing as a technique that makes us understand the obtained images. This paper proposes an automatic extraction method for cerebral region based on region growing. This method enables us to obtain the volumes of various bra,in parts. Moreover, it shows a method to do 3D display of the extracted shape.