123

Smart Innovation, Systems and Technologies 45

Yen-Wei ChenCarlos ToroSatoshi TanakaRobert J. HowlettLakhmi C. Jain Editors

Innovation in Medicine and Healthcare 2015

Smart Innovation, Systems and Technologies

Volume 45

Series editors

Robert J. Howlett, KES International, Shoreham-by-Sea, UKe-mail: rjhowlett@kesinternational.org

Lakhmi C. Jain, University of Canberra, Canberra, Australia, andUniversity of South Australia, Adelaide, Australiae-mail: Lakhmi.jain@unisa.edu.au

About this Series

The Smart Innovation, Systems and Technologies book series encompasses thetopics of knowledge, intelligence, innovation and sustainability. The aim of theseries is to make available a platform for the publication of books on all aspects ofsingle and multi-disciplinary research on these themes in order to make the latestresults available in a readily-accessible form. Volumes on interdisciplinaryresearch combining two or more of these areas is particularly sought.

The series covers systems and paradigms that employ knowledge andintelligence in a broad sense. Its scope is systems having embedded knowledgeand intelligence, which may be applied to the solution of world problems inindustry, the environment and the community. It also focusses on the knowledge-transfer methodologies and innovation strategies employed to make this happeneffectively. The combination of intelligent systems tools and a broad range ofapplications introduces a need for a synergy of disciplines from science,technology, business and the humanities. The series will include conferenceproceedings, edited collections, monographs, handbooks, reference books, andother relevant types of book in areas of science and technology where smartsystems and technologies can offer innovative solutions.

High quality content is an essential feature for all book proposals accepted forthe series. It is expected that editors of all accepted volumes will ensure thatcontributions are subjected to an appropriate level of reviewing process and adhereto KES quality principles.

More information about this series at http://www.springer.com/series/8767

Yen-Wei Chen ⋅ Carlos ToroSatoshi Tanaka ⋅ Robert J. HowlettLakhmi C. JainEditors

Innovation in Medicineand Healthcare 2015

123

EditorsYen-Wei ChenCollege of Information Scienceand Engineering

Ritsumeikan UniversityKusatsu, ShigaJapan

Carlos ToroVicomtech-IK4Donostia San SebastianSpain

Satoshi TanakaCollege of Information Scienceand Engineering

Ritsumeikan UniversityKusatsu, ShigaJapan

Robert J. HowlettKES InternationalShoreham-by-SeaUK

Lakhmi C. JainFaculty of Education, Science, Technologyand Mathematics

University of CanberraCanberraAustralia

ISSN 2190-3018 ISSN 2190-3026 (electronic)Smart Innovation, Systems and TechnologiesISBN 978-3-319-23023-8 ISBN 978-3-319-23024-5 (eBook)DOI 10.1007/978-3-319-23024-5

Library of Congress Control Number: 2015946760

Springer Cham Heidelberg New York Dordrecht London© Springer International Publishing Switzerland 2016This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or partof the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmissionor information storage and retrieval, electronic adaptation, computer software, or by similar ordissimilar methodology now known or hereafter developed.The use of general descriptive names, registered names, trademarks, service marks, etc. in thispublication does not imply, even in the absence of a specific statement, that such names are exemptfrom the relevant protective laws and regulations and therefore free for general use.The publisher, the authors and the editors are safe to assume that the advice and information in thisbook are believed to be true and accurate at the date of publication. Neither the publisher nor theauthors or the editors give a warranty, express or implied, with respect to the material containedherein or for any errors or omissions that may have been made.

Printed on acid-free paper

Springer International Publishing AG Switzerland is part of Springer Science+Business Media(www.springer.com)

Preface

The third KES International Conference on Innovation in Medicine and Healthcare(InMed-15) was held during 11–12 September 2015 in Kyoto, Japan, organized byKES International and co-organized by Research Center of Advanced ICT forMedical and Healthcare, Ritsumeikan University, Japan.

InMed-15 is the third edition of the InMed series of conferences. The first andthe second InMed conferences were held in Greece and Spain, respectively.InMed-15 is the first conference held outside of European countries. The conferencefocuses on the major trends and innovations in modern intelligent systems applyingto medicine, surgery, healthcare and the issues of an ageing population. The pur-pose of the conference is to exchange new ideas, new technologies and currentresearch results in these research fields.

We received 78 submissions from 16 countries. All submissions were carefullyreviewed by at least two reviewers of the International Programme Committee.Finally 53 papers (including one short paper) were accepted to be presented at theproceedings. The major areas covered at the conference and presented at theproceedings include: (1) Medical Informatics; (2) Biomedical Engineering;(3) Management for Mealthcares; (4) Advanced ICT for Medical and Healthcare;(5) Simulation and Visualization/VR for Medicine; (6) Statistical Signal Processingand Artificial Intelligence; (7) Smart Medical and Healthcare System and(8) Healthcare Support System. In addition to the accepted research papers, threekeynote speeches by leading researchers were presented at the conference.

We thank Dr. Xian-hua Han and Dr. Rui Xu of Ritsumeikan University for theirvaluable editing assistance for this book. We are also grateful to the authors andreviewers for their contributions.

September 2015 Yen-Wei ChenCarlos Toro

Satoshi TanakaRobert J. HowlettLakhmi C. Jain

v

InMed 2015 Organization

General Chair

Yen-Wei Chen, Ritsumeikan University, Japan

General Co-chairs

Shigehiro Morikawa, Shiga University of Medical University, JapanLakhmi C. Jain, University of South Australia, Australia

Executive Chair

Robert J. Howlett, Bournemouth University, UK

Programme Chairs

Satoshi Tanaka, Ritsumeikan University, JapanEdward J. Ciaccio, Columbia University, USA

Publicity Chair

Lanfen Lin, Zhejiang University, China

vii

Co-Chairs of Workshop on Smart Medicaland Healthcare Systems

Carlos Toro, Vicomtech-IK4, SpainIvan Macia Oliver, Vicomtech-IK4, Spain

Organization and Management

KES International (www.kesinternational.org) in partnership with RitsumeikanUniversity, Japan (www.ritsumei.jp) and the Institute of Knowledge Transfer(www.ikt.org.uk)

International Programme Committee Members and Reviewers

Sergio Albiol-Pérez, Universidad de Zaragoza, SpainArnulfo Alanis, Instituto Tecnológico de Tijuana, MexicoDanni Ai, Beijing Institute of Technology, ChinaTsukasa Aso, National Institute of Technology, Toyama College, JapanShinichiro Ataka, Osaka International University, JapanAhmad Taher Azar, Faculty of Computers and Information, Benha University,EgyptVitoantonio Bevilacqua, Dipartimento di Ingegneria Elettrica edell’Informazione –Politecnico di Bari, ItalyGiosue Lo Bosco, University of Palermo, ItalyChristopher Buckingham, Computer Science, Aston University, Birmingham, UKM. Emre Celebi, Louisiana State University in Shreveport, USAYen-Wei Chen, Ritsumeikan University, JapanD. Chyzhyk, Ritsumeikan University, JapanLuis Enrique Sánchez Crespo, Universidad de las Fuerzas Armadas, EcuadorGuifang Duan, Zhejiang University, ChinaNashwa Mamdouh El-Bendary, Arab Academy for Science, Technology, andMaritime TransportMassimo Esposito, ICAR-CNR, ItalyCecilia Dias Flores, Federal University of Health Sciences of Porto AlegreJosé Manuel Fonseca, Faculty of Sciences and Technology of Universidade Novade Lisboa, LisbonAmir H. Foruzan, Shahed University, IranArfan Ghani, University of Bolton, Greater Manchester, UKManuel Graña, University of the Basque Country, SpainHiroshi Hagiwara, Ritsumeikan University, Japan

viii InMed 2015 Organization

Xian-hua Han, Ritsumeikan University, JapanKyoko Hasegawa, Ritsumeikan University, JapanAboul Ella Hassanien, Cairo University, EgyptIoannis Hatzilygeroudis, University of Patras, GreeceElena Hernández-Pereira, University of A Coruña, SpainYasushi Hirano, Yamaguchi University, JapanRobert Howlett, Bournemouth University, UKMonica Huerta, Universidad Simón Bolívar-Venezuela and Universidad PolitécnicaSalesiana—EcuadorHongjie Hu, Zhejiang University, ChinaAjita Ichalkaranje, Seven Steps Physiotherapy Clinic, Adelaide, South Australia.Nikhil Ichalkaranje, Government of South Australia, Adelaide, South Australia.Soichiro Ikuno, Tokyo University of Technology, JapanIgnacio Illan, Universitat Rovira i Virgili (URV), Catalonia, SpainDavid Isern, Universitat Rovira i Virgili (URV), Catalonia, SpainSandhya Jain, Medical Practitioner, Adelaide, South Australia.Huiyan Jiang, Northeastern University, ChinaKyoji Kawagoe, Ritsumeikan University, JapanHiroharu Kawanaka, Graduate School of Engineering, Mie University, JapanTakayuki Kawaura, Kansai Medical University, JapanAkinori Kimura, Ashikaga Institute of Technology, JapanZiad Kobti, University of Windsor, CanadaTomohiro Kuroda, Kyoto University Hospital, JapanJoo-ho Lee, Ritsumeikan University, JapanLenin G. Lemus-Zúñiga, Universitat Politècnica de València, EspañaJingbing Li, Hainan University, ChinaSergio Magdaleno, Instituto Tecnologico de TijuanaPaco Martinez, Instituto Tecnologico de TijuanaEsperanza Manrique, Universidad Autonoma de Baja CaliforniaTakafumi Marutani, Ritsumeikan University, JapanYasushi Matsumura, Osaka University, JapanNaoki Matsushiro, Osaka Police Hospital, JapanKazuyuki Matsumoto, University of Tokushima, JapanYoshiyuki Matsumoto, Shimonoseki City University, JapanTadashi Matsuo, Ritsumeikan University, JapanRashid Mehmood, College of Computer Science, King Khalid UniversityNora del Carmen Osuna Millan, Universidad Autónoma de Baja Californiaand Osuna ConsultorHongying Meng, Brunel University London, UKTakashi Mitsuda, Ritsumeikan University, JapanJose Montanana, Universidad Complutense de Madrid, SpainAntonio Moreno, Universitat Rovira i Virgili (URV), SpainLouise Moody, Coventry School of Art and Design, Coventry University, UKKeisuke Nagase, Kanazawa University, JapanKazuo Nakazawa, National Cerebral and Cardiovascular Center, Japan

InMed 2015 Organization ix

Marek R. Ogiela, AGH University of Science and Technology, PolandManuel G. Penedo, University of CoruñaDorin Popescu, University of Craiova, Faculty of Automation, Computers andElectronicsJose-Luis Poza-Lujan, Universitat Politècnica de ValènciaMargarita Ramirez, UABC/Instituto Tecnologico de TijuanaAna Respício, Universidade de Lisboa, PortugalJoel Rodrigues, Instituto de Telecomunicações, University of Beira Interior,PortugalJohn Ronczka, Australian Society of Rheology, AustraliaBaltazar Rosario, Leon Institute of TechnologyJuan Manuel Gorriz Saez, University of Granada, SpainNaohisa Sakamoto, Kyoto UniversityEider Sanchez, VicomtechMaricela Sevilla, Parque Industrial Internacional TijuanaSyed Shafiq, The University of NewcastleAkihiko Shiino, Shiga University of Medical Science, JapanNobutaka Shimada, Ritsumeikan University, JapanNaruhiro Shiozawa, Ritsumeikan University, JapanKenji Suzuki, Illinois Institute of Technology, USAPawel Swiatek, Wroclaw University of Technology, PolandKazuyoshi Tagawa, Ritsumeikan University, JapanTadamasa Takemura, University of Hyogo, JapanSatoshi Tanaka, Ritsumeikan University, JapanTomoko Tateyama, Ritsumeikan University, JapanGancho Vachkov, University of the South Pacific (USP), FijiEloisa Vargiu, Barcelona Digital Technology Center, SpainAthanasios V. Vasilakos, Kuwait UniversityJunzo Watada, Waseda University, JapanXiong Wei, Institute for Infocomm Research, SingaporeAndree Woodcock, Coventry University, UKRui Xu, Ritsumeikan University, JapanYoshiyuki Yabuuchi, Shimonoseki City University, JapanBogart Yail, Instituto Tecnologico de TijuanaJian Yang, Beijing Institute of Technology, ChinaHiro Yoshida, Massachusetts General Hospital and Radiology, Harvard MedicalSchoolHaoxi Zhang, University of Newcastle, Australia

x InMed 2015 Organization

Contents

Part I Emerging Research Leaders of Medical Informaticsin KANSAI Area

Prediction of Clinical Practices by Clinical Dataof the Previous Day Using Linear Support Vector Machine . . . . . . . . . 3Takashi Nakai, Tadamasa Takemura, Risa Sakurai, Kenichiro Fujita,Kazuya Okamoto and Tomohiro Kuroda

Method for Detecting Drug-Induced Interstitial Pneumoniafrom Accumulated Medical Record Data at a Hospital . . . . . . . . . . . . 15Yoshie Shimai, Toshihiro Takeda, Shirou Manabe, Kei Teramoto,Naoki Mihara and Yasushi Matsumura

Visualization and Quantitative Analysis of Nursing StaffTrajectories Based on a Location System . . . . . . . . . . . . . . . . . . . . . . 25Kikue Sato, Tomohiro Kuroda and Akitoshi Seiyama

A Web-Based Stroke Education Application for OlderElementary Schoolchildren Focusing on the FAST Message . . . . . . . . . 37Shoko Tani, Hiroshi Narazaki, Yuta Ueda, Yuji Nakamura,Tenyu Hino, Satoshi Ohyama, Shinya Tomari, Chiaki Yokota,Naoki Ohboshi, Kazuo Minematsu and Kazuo Nakazawa

Part II Biomedical Engineering, Trends, Researchand Technologies

A Review of Mobile Apps for Improving Quality of Lifeof Asthmatic and People with Allergies . . . . . . . . . . . . . . . . . . . . . . . . 51Miguel A. Mateo Pla, Lenin G. Lemus-Zúñiga, José-Miguel Montañana,Julio Pons and Arnulfo Alanis Garza

xi

Physical Implementation of a Customisable System to Assista User with Mobility Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65Sandra López, Rosario Baltazar, Miguel Ángel Casillas,Víctor Zamudio, Juan Francisco Mosiño, Arnulfo Alanisand Guillermo Méndez

Mobile Applications in Health, Competitive Advantagein the State of Baja California Mexico . . . . . . . . . . . . . . . . . . . . . . . . 75Nora del Carmen Osuna Millán, Margarita Ramírez Ramírez,María del Consuelo Salgado Soto, Bogart Yali Márquez Lobatoand Arnulfo Alanís Garza

Mental Activation of Seniors Incorporating ICT in TheirDaily Lives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85Maricela Sevilla Caro, María del Consuelo Salgado Soto,Margarita Ramírez Ramírez, Esperanza Manrique Rojas,Hilda Beatriz Ramírez Moreno and Lenin G. Lemus Zúñiga

Quality of Life and Active Aging Through EducationalGerontology in Information Technology and Communication . . . . . . . 93Esperanza Manrique Rojas, Hilda Beatriz Ramírez Moreno,Margarita Ramírez Ramirez, Nora del Carmen Osuna Millan,Arnulfo Alanís Garza and José Sergio Magdaleno Palencia

Simulation of Cervical Cord Compression Using FiniteElement Method (FEM) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103Nur Fadhlina Binti Shaari, Junji Ohgi, Norihiro Nishida,Xian Chen, Itsuo Sakuramoto and Toshihiko Taguchi

Part III Management for Healthcare

A Study of Older People’s Socializing Form with Others:Comparative Analysis of “Living Alone” and “Livingwith a Spouse” Using Quantification Theory Type II . . . . . . . . . . . . . 115Takayuki Kawaura and Yasuyuki Sugatani

Analysis of Japanese Health using Fuzzy PrincipalComponent Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127Yoshiyuki Yabuuchi and Takayuki Kawaura

Analysis of Time-Series Data Using the Rough Set . . . . . . . . . . . . . . . 139Yoshiyuki Matsumoto and Junzo Watada

xii Contents

Study on Multi-Period Transportation Problem ConsideringOpportunity Loss and Inventory . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149Shinichiro Ataka and Hisayoshi Horioka

Diagnosis of ECG Data for Detecting Cardiac DisorderUsing DP-Matching and Artificial Neural Network . . . . . . . . . . . . . . . 161Mohamad Sabri bin Sinal and Eiji Kamioka

Smart Technology for a Smarter Patient: Sketchingthe Patient 2.0 Profile . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173Luca Buccoliero, Elena Bellio, Maria Mazzola and Elisa Solinas

Construction and Evaluation of Bayesian Networks Relatedto the Specific Health Checkup and Guidanceon Metabolic Syndrome . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183Yoshiaki Miyauchi and Haruhiko Nishimura

Medical Care Delivery at the XXVII World SummerUniversiade Kazan 2013 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195Timur Mishakin, Elena Razumovskaya, Michael Popovand Olga Berdnikova

Part IV Advanced ICT for medical and Healthcare

Coccurrence Statistics of Local Ternary Patterns for HEp-2Cell Classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205Xian-Hua Han, Yen-Wei Chen and Gang Xu

Combined Density, Texture and Shape Features of Multi-phaseContrast-Enhanced CT Images for CBIR of Focal LiverLesions: A Preliminary Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215Yingying Xu, Lanfen Lin, Hongjie Hu, Huajun Yu, Chongwu Jin,Jian Wang, Xianhua Han and Yen-Wei Chen

Supporting Nurses’ Work and Improving Medical SafetyUsing a Sensor Network System in Hospitals . . . . . . . . . . . . . . . . . . . 225Misa Esashi, Haruo Noma and Tomohiro Kuroda

Eye-Hand Coordination Analysis According to SurgicalProcess in Laparoscopic Surgery Training . . . . . . . . . . . . . . . . . . . . . 237Takafumi Marutani, Hiromi T. Tanaka, Nobutaka Shimada,Masaru Komori, Yoshimasa Kurumi and Shigehiro Morikawa

Contents xiii

An Improvement of Surgical Phase Detection Using LatentDirichlet Allocation and Hidden Markov Model . . . . . . . . . . . . . . . . . 249Dinh Tuan Tran, Ryuhei Sakurai and Joo-Ho Lee

Implementing a Human-Behavior-Process Archive and SearchDatabase System Using Simulated Surgery Processes . . . . . . . . . . . . . 263Zhang Zuo, Kenta Oku and Kyoji Kawagoe

Unobtrusive Sensing of Human Vital Functions by a SensoryUnit in the Refrigerator Door Handle . . . . . . . . . . . . . . . . . . . . . . . . . 275D. Zazula, S. Srkoč and B. Cigale

Measurement of 3-D Workspace of Thumb Tip with RGB-DSensor for Quantitative Rehabilitation . . . . . . . . . . . . . . . . . . . . . . . . 287Tadashi Matsuo and Nobutaka Shimada

Automatic Segmentation Method for Kidney Using DualDirection Adaptive Diffusion Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . 299Xu Qiao, Wujing Lu, Xuantao Su and Yen-Wei Chen

Automatic Registration of Deformable Organs in MedicalVolume Data by Exhaustive Search . . . . . . . . . . . . . . . . . . . . . . . . . . 309Masahiro Isobe, Shota Niga, Kei Ito, Xian-Hua Han, Yen-Wei Chenand Gang Xu

Part V Simulation and Visualisation/VR for Medicine

GPU Acceleration of Monte Carlo Simulation at the Cellularand DNA Levels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323Shogo Okada, Koichi Murakami, Katsuya Amako, Takashi Sasaki,Sébastien Incerti, Mathieu Karamitros, Nick Henderson,Margot Gerritsen, Makoto Asai and Andrea Dotti

A Study on Corotated Nonlinear Deformation Modelfor Simulating Soft Tissue Under Large Deformation . . . . . . . . . . . . . 333Kazuyoshi Tagawa, Takahiro Yamada and Hiromi T. Tanaka

Remote Transparent Visualization of Surface-Volume FusedData to Support Network-Based Laparoscopic SurgerySimulation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 345Rui Xu, Asuka Sugiyama, Kyoko Hasegawa, Kazuyoshi Tagawa,Satoshi Tanaka and Hiromi T. Tanaka

xiv Contents

Study of Surgical Simulation of Flatfoot Using a FiniteElement Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353Zhongkui Wang, Kan Imai, Masamitsu Kido, Kazuya Ikomaand Shinichi Hirai

A Study of Meditation Effectiveness for Virtual Reality BasedStress Therapy Using EEG Measurementand Questionnaire Approaches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 365Gamini Perhakaran, Azmi Mohd Yusof, Mohd Ezanee Rusli,Mohd Zaliman Mohd Yusoff, Imran Mahaliland Ahmad Redza Razieff Zainuddin

Part VI Statistical Signal Processing and Artificial Intelligence

Late Onset Bipolar Disorder Versus Alzheimer Disease . . . . . . . . . . . . 377Darya Chyzhyk, Marina Graña-Lecuona and Manuel Graña

Short-term Prediction of MCI to AD Conversion Basedon Longitudinal MRI Analysis and Neuropsychological Tests . . . . . . . 385Juan Eloy Arco, Javier Ramírez, Juan Manuel Górriz, Carlos G. Puntonetand María Ruz

Ensemble Tree Learning Techniques for Magnetic ResonanceImage Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 395Javier Ramírez, Juan M. Górriz, Andrés Ortiz, Pablo Padillaand Francisco J. Martínez-Murcia for the Alzheimer DiseaseNeuroimaging Initiative

Part VII Smart Medical and Healthcare System 2015 Workshop

Integrating Electronic Health Records in Clinical DecisionSupport Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 407Eider Sanchez, Carlos Toro and Manuel Graña

An Ad-Hoc Image Segmentation of Subcutaneous and VisceralAdipose Tissue from Abdomino-Pelvic MagneticResonance Images . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 417Oier Echaniz, Borja Ayerdi, Alexandre Savio and Manuel Graña

Automated Segmentation of Subcutaneous and Visceral AdiposeTissues from MRI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 427Borja Ayerdi, Oier Echaniz, Alexandre Savio and Manuel Graña

Contents xv

Enabling a Smart and Distributed CommunicationInfrastructure in Healthcare . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 435Chrystinne Oliveira Fernandes, Carlos José Pereira de Lucena,Carlos Alberto Pereira de Lucena and Bruno Alvares de Azevedo

Ultrasound Image Dataset for Image Analysis AlgorithmsEvaluation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 447Camilo Cortes, Luis Kabongo, Ivan Macia, Oscar E. Ruizand Julian Florez

Approaches of Phase Lag Index to EEG Signalsin Alzheimer’s Disease from Complex Network Analysis . . . . . . . . . . . 459Shinya Kasakawa, Teruya Yamanishi, Tetsuya Takahashi,Kanji Ueno, Mitsuru Kikuchi and Haruhiko Nishimura

Part VIII Healthcare Support System

A Benchmark on Artificial Intelligence Techniques for AutomaticChronic Respiratory Diseases Risk Classification . . . . . . . . . . . . . . . . 471Sebastián A. Ríos, Fabián García Tenorioand Angel Jimenez-Molina

Toward Non-invasive Polysomnograms Through the Studyof Electroencephalographic Signal Correlation for RemotelyMonitored Cloud-Based Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . 483Claudio Estevez, Diego Vallejos, Sebastián A. Ríosand Pablo Brockmann

Work with Iodine-125: 8 Years Experience in Brachy-TherapySources Production Lab . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 493C.D. Souza, F.S. Peleias Jr, M.E.C.M. Rostelato, C.A. Zeituni,R. Tiezzi, B.T. Rodrigues, A. Feher, J.A. Moura and O.L. Costa

Comprehensible Video Acquisition for CaregivingScenes—How Multimedia Can Support Caregiver Training. . . . . . . . . 503Yuichi Nakamura, Kazuaki Kondo, Taiki Mashimo,Yoshiaki Matsuoka and Tomotake Ohtsuka

Care at the End of Life: Design Priorities for Peoplewith Dementia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 517Alastair S. Macdonald and Louise Robinson

xvi Contents

A Wireless and Autonomous Sensing System for Monitoringof Chronic Wounds in Healthcare . . . . . . . . . . . . . . . . . . . . . . . . . . . 527Alex Hariz and Nasir Mehmood

Maturity Models for Hospital Information Systems Management:Are They Mature?. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 541João Vidal de Carvalho, Álvaro Rochaand José Braga de Vasconcelos

Why Cannot Control Your Smartphones by Thinking?Hands-Free Information Control System Based on EEG . . . . . . . . . . . 553Yuanyuan Wang, Tomoki Hidaka, Yukiko Kawai and Jiro Okuda

Comparative Analysis of Retinal Fundus Images with the DistantPast Images Using a Vessel Extraction Technique . . . . . . . . . . . . . . . . 565Toshio Modegi, Yoichi Takahashi, Tae Yokoi, Muka Moriyama,Noriaki Shimada, Ikuo Morita and Kyoko Ohno-Matsui

Interactive Segmentation of Pancreases from AbdominalCT Images by Use of the Graph Cut Techniquewith Probabilistic Atlases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 575Takenobu Suzuki, Hotaka Takizawa, Hiroyuki Kudoand Toshiyuki Okada

Validation of Knot-Tying Motion by Temporal-SpatialMatching with RGB-D Sensor for Surgical Training . . . . . . . . . . . . . . 585Yoko Ogawa, Nobutaka Shimada, Yoshiaki Shirai,Yoshimasa Kurumi and Masaru Komori

Erratum to: Prediction of Clinical Practices by Clinical Dataof the Previous Day Using Linear Support Vector Machine . . . . . . . . . E1Takashi Nakai, Tadamasa Takemura, Risa Sakurai,Kenichiro Fujita, Kazuya Okamoto and Tomohiro Kuroda

Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 591

Contents xvii

Part IEmerging Research Leaders of Medical

Informatics in KANSAI Area

Prediction of Clinical Practices by ClinicalData of the Previous Day Using LinearSupport Vector Machine

Takashi Nakai, Tadamasa Takemura, Risa Sakurai, Kenichiro Fujita,Kazuya Okamoto and Tomohiro Kuroda

Abstract For preventing mistakes and guaranteeing quality, Clinical DecisionSupport System (CDSS) is expected to support decision-making of healthcareprofessionals. Until now, there had been many studies of CDSS based on rule-basedproduction system. However practical use of CDSS had been limited because of thecomplexity of medical care. On the other hand, the machine learning that requiredhuge amounts of data is used to make data-driven predictions or decisions incomputer science recently. We considered that this method was effective in prac-tical care. Firstly, we constructed a model based on the hypothesis that clinicalpractices of a day might determine them of next day for the same patients. Next, wecreated predictors using linear support vector machines and clinical actions onDiagnosis Procedure Combination/Per-Per Diem Payment System introduced formedical billing in Japan. Finally we evaluated predictive performance by crossvalidation. As a result, the clinical actions whose frequency of appearance were

The erratum of this chapter can be found under DOI 10.1007/978-3-319-23024-5_54

T. Nakai (✉) ⋅ T. Takemura ⋅ R. Sakurai ⋅ K. FujitaGraduate School of Applied Informatics University of Hyogo, Kobe, Japane-mail: ab14h405@ai.u-hyogo.ac.jp

T. Takemurae-mail: takemura@ai.u-hyogo.ac.jp

R. Sakuraie-mail: ab14v403@ai.u-hyogo.ac.jp

K. Fujitae-mail: kfujita@kuhp.kyoto-u.ac.jp

K. Fujita ⋅ K. Okamoto ⋅ T. KurodaDivision of Medical Information Technology and Administration Planning,Kyoto University Hospital, Kyoto, Japane-mail: kazuya@kuhp.kyoto-u.ac.jp

T. Kurodae-mail: tomo@kuhp.kyoto-u.ac.jp

© Springer International Publishing Switzerland 2016Y.-W. Chen et al. (eds.), Innovation in Medicine and Healthcare 2015,Smart Innovation, Systems and Technologies 45,DOI 10.1007/978-3-319-23024-5_1

3

higher trend to have higher predictive performance. Thus, it was suggested thatpredictive performance was improved by preparing sufficient amount of data andensuring the frequency of appearance.

1 Introduction

For preventing medical mistakes and guaranteeing medical quality, Clinical Deci-sion Support System (CDSS) is expected to provide reference information fordecision-making of healthcare professionals [1, 2]. A lot of studies have been madeon CDSS based on rule-based production system that began in MYCIN in the1970s. However, it is difficult to make rules for all clinical decision because medicalcare is very complicated and complex. Only a few systems are in practical use withlimited condition such as irregular pulse determination system. On the other hand,in computer science, the machine learning that required huge amounts of data isused to make data-driven predictions or decisions. Recently, several these studiesgive high predictive performance because of development of methods of machinelearning and computers. For example, several systems by applying machinelearning that required huge amounts of data have beat professional players in shogi.We consider these methods would be applicable to medical care.

In this study, we examined the predictability of clinical practices to a patient of aday by using huge clinical actions and machine learning. Actual data of clinicalactions was entered and accumulated into hospital information system, however weused clinical actions data of Diagnosis Procedure Combination/Per-Diem PaymentSystem (DPC/PDPS) in this study. DPC/PDPS is a system for medical billing inJapan and the data of clinical actions based on DPC/PDPS are represented as theCombination of Diagnosis and Procedure. These data are being accumulated basedon domestic standards in Japan. We constructed a model based on the hypothesisthat clinical practices can be predicted from the data of clinical practices done onthe day before for the same patient. We predicted all clinical actions appeared in thefuture by machine learning and evaluated the results. Concretely we created pre-dictors by using linear support vector machines and 6 months of the DPC data asclinical actions on all admission at a large-scale university hospital. We evaluatedpredictive performance by cross validation.

2 Background and Related Work

2.1 Proposed Model for Prediction of Clinical Actions

In this paper, we propose the model based on the hypothesis that clinical practicescan be predicted from the data of clinical practices done on the day before for the

4 T. Nakai et al.

same patient. Figure 1 shows the concept of this model. We examined that clinicalpractice set Xd(t′1,t′2,…t′n) on a day d can be predicted from clinical practice setX(d-1)(t1, t2,…,tm) done on the day before for the same patient. Concretely, wecreated predictors by applying machine learning to clinical practice sets of all d andd-1 in clinical action data. We examined predictability of appearance of the clinicalpractice using this predictor.

In this paper, we used Diagnosis Procedure Combination (DPC) data as clinicalactions data. DPC data is accumulated by Diagnosis Procedure Combination/Per-Diem Payment System (DPC/PDPS) in Japan, and used linear support vectormachine as the machine learning method.

2.2 Diagnosis Procedure Combination/Per-Diem PaymentSystem

Diagnosis Procedure Combination/Per-Diem Payment System (DPC/PDPS) is theflat-rate payment system based on Diagnosis Procedure Combination (DPC) [3].DPC was made considering clinical circumstance in Japan based on DiagnosisRelated Group (DRG) of USA. The purposes of DPC and DPC/PDPS are theimprovement of medical costs and services. DPC/PDPS was first introduced to 82advanced treatment hospitals that had 66,497 beds totally in April 2003, andintroduced to 1585 hospitals that had 492,206 bets totally in April 2014 [4].

The files of DPC are easy handle format for computer such as “style 1”, “D file”,“E file” and “F file”. For medical service comparison, clinical actions for patientsare accumulated. E and F file include details of clinical actions. On these file,clinical actions to patients are represented as the data category code and the unifiedelectronic receipt code. Particularly the data category code shows the rough cate-gory of the clinical actions and indicates if the clinical action is practiced in anysituation. Table 1 shows examples of data category code.

2.3 Linear Support Vector Machines

Linear support vector machine (Linear SVM) is the methods of machine learningand one kind of supervised learning [5]. Linear SVMs are mainly used as 2 class

Fig. 1 Concept of prediction model

Prediction of Clinical Practices by Clinical Data … 5

classifier for data whose class is not known. In linear SVM, data is handled asfeature vectors. If the feature vectors are m-dimensional vectors, linear SVMdetermines a class of data using separating hyperplane expressed as the following.

f xð Þ=wTx + b

w is a m-dimensional vector and b is a bias term. There are innumerable sep-arating hyperplane. Linear SVM uses the maximum margin separating hyperplanewhich has the maximum distance (margin) to teacher data of each class because themaximum margin separating hyperplane is expected to have good classificationaccuracy for unknown data.

3 Experiment of Prediction of Clinical Practices

This chapter describes the experiment of prediction of clinical practices as previ-ously explained.

Table 1 Examples of data categories

Data category code Data category name

11 First consultation fee13 Education14 Home care21 Medicines for internal use22 Potion23 Medicine for external use24 Drug compounding = hospitalization26 Narcotic drug and so on27 Basic technical fee for dispensing31 Intramuscular injection32 Intravenous infusion33 others (injection)40 Procedures50 Operation54 Anesthesia60 Examination70 Radiology and image diagnosis80 Others (rehabilitation, radiation therapy and so on)90 Basic fee for hospitalization = hospitalization92 Specific hospital charges = hospitalization97 Hospital meals standard obligation fees = hospitalization

6 T. Nakai et al.

3.1 Target Data

We used the E and F files about the patients who had admitted and discharged in thelarge-scale university hospital within the period from 2006/4 to 2006/6 for theexperiment. These files included the data of 6,445 hospitalizations and118,150 days as shown in Table 2.

In these files, a clinical practice is represented as receipt computerized systemcode. However, if two practices have same unified electronic receipt code anddifferent data category code, it is possible to consider different clinical practice. Forexample, we can separate that “medicine (name)” is used “operation” and “medi-cine for external use” using the date category code. So, if two clinical practices haddifferent data category code, we considered these clinical practices were different inthis study. Therefore we regarded 5,279 combinations of “data category code” and“unified electronic receipt code” as targets of prediction.

3.2 Experiment

The purpose of the proposed model is the clinical practices a day could determinethe clinical practice next day. Linear SVM handles data as vector as described inSect. 2.3, therefore we handle data as 5279-dimensional vectors in this experiment.Each element of the vector is mapped to 5279 clinical practices as shown inTable 3.

We make the experimental data set for each clinical practice. We evaluate thepredictive performance.

1. Make lists of clinical practices for each patient by the day from DPC data

Make vectors from lists of step 1 as shown in Fig. 2. The value of vectorelements is 1 if the clinical action is done. The value of vector elements is 0 if theclinical action is not done.

2. Classify vectors of step 0 into class 1 if the target clinical practice is done on theday after. Classify vectors into class 0 if not. For example, Table 4 shows the

Table 2 Summary of E files and F files

Item Number of data

Number of hospitalizations 6,445Total number of days of hospitalizations 118,150“data category code” + “receipt computerized system code” Combinations 5,279Number of records of F files 2,543,205

Prediction of Clinical Practices by Clinical Data … 7

example of experimental data whose target clinical practice is ‘Arterial bloodsampling’. The 4543-th element is mapped to ‘Arterial blood sampling’ asshown in Table 3. Therefore we classify the vector into class 1 if the 4543-thelement of vector of the day after is 1. We classify the vector into class 0 if not.We remove the vectors for discharge date from the experimental data becausenone of clinical practices on the next day. As a result we got 111,705 vectors asexperimental data of each clinical practice of prediction target.

3. Divide experimental data into 10 parts. Use one of parts as test data in rotationand use other parts as teacher data. Create the classifier by using teacher data,and evaluated prediction performance for teacher data by 10-fold cross vali-dation [5]. We used Liblinear [6] to create classifiers and to predict for test data.

Table 3 Elements of vector for linear svm

Elementnumberof featurevector

Datacategorycode

Receipt computerized systemcode

Medical specification name

1 11 111000110 first consultation fee(hospital)

2 11 111000370 first consultation fee(infants)

3 11 111000570 first consultation fee(overtime)

4 11 111000670 first consultation fee(holiday)

:4543 60 160101210 Arterial blood sampling:5278 97 197000470 special diet fee (per one

meal)5279 97 197000570 diet fee in dining room

Fig. 2 Conversion of list ofclinical practices to vectors

8 T. Nakai et al.

3.3 Result

Table 5 shows the result of experiment about ‘Arterial blood sampling’ by themethod described in Sect. 3.2. ‘True Positive’ is the number of test data of class 1predicted correctly. ‘False Positive’ is the number of test data of class 0 predicted tobe class 1 by mistake. ‘False Negative’ is the number of test data of class 1predicted to be class 0 by mistake. ‘True Negative’ is the number of test data ofclass 0 predicted correctly. We calculated precision and recall rate for each 10 testdata. Based on cross validation, the average values of these rates are final precision

Table 4 Image of experimental data for ‘Arterial blood sampling’

Table 5 Results of prediction for “Arterial blood sampling”

Targettestdata

Truepositive(TP)

Falsepositive(FP)

Falsenegative(FN)

Truenegative(TN)

Precision =TP/(TP + FP)

Recall =TP/(TP + FN)

1 66 26 186 10893 0.717 0.2622 99 43 140 10889 0.697 0.4143 55 25 149 10942 0.688 0.2704 92 57 161 10861 0.618 0.3645 68 29 147 10927 0.701 0.3166 73 37 167 10893 0.664 0.3047 124 81 207 10758 0.605 0.3758 97 37 168 10868 0.729 0.3669 47 19 163 10941 0.712 0.224

10 49 30 134 10957 0.620 0.268The average precision and recall (result of 10-folds crossvalidation)

0.675 0.316

Prediction of Clinical Practices by Clinical Data … 9

and recall rate of ‘Arterial blood sampling’ in this paper. As a result, precision ratewas 0.675 and recall rate was 0.316.

We predicted 5279 kinds of clinical practices by the same steps. Figure 3 showsnumber of clinical actions by clinical data category in experimental data. Thenumber of appearances of clinical actions can be restated as the number of vectorsclassified into class 1 by the step 4 of 0, and was divided by data category code andby unified electronic receipt code.

The 235 clinical practices only appeared on the first day of admission in total5279 practices, so this system did not predict these admission date practices. Inaddition, few clinical practices appeared in this data could not be fitted SVMsbecause this machine learning method needs a certain amount data to learn andjudge the correct answers. Therefore, on 10-fold cross validation, we omittedclinical practices that did not appear on all validation from the total practices. As aresult, there were 657 clinical practices could be calculated on 10-fold validationthis time. Figure 4 shows a scatter diagram of precision rate versus recall rate. Thesizes of circle are proportional to the frequency of appearance.

We shows examples of clinical practices and their precision rate and recall rateof ‘33 others (injection)’ as Table 6 and ‘21 medicines for internal use’ as Table 7.In addition, Fig. 5 shows a scatter diagram of precision rate versus recall rate of ‘33others (injection)’ as well as Fig. 4. Figure 6 is the same diagram of ‘21 medicinesfor internal use’.

Fig. 3 Number of clinical actions by clinical data category

10 T. Nakai et al.

Fig. 4 Scatter diagram ofprecision rate versus recallrate

Table 6 Examples of precision and recall of others (injection)

Receipt computerizedsystem code

Medical specification name Number ofappearance

Precision Recall

130004410 Central venous injection 6249 0.896 0.885640451009 Omepral 20 mg 2862 0.836 0.758

643910067 Stronger Neo-MinophagenC 20 mL

1595 0.806 0.680

620002258 Twinpal 500 mL 1204 0.805 0.721130000210 Precise continuous

intravenous infusion5797 0.801 0.758

620002191 Glyceol 200 mL 744 0.796 0.624620001934 Neoparen No. 2 1000 mL 1207 0.790 0.722620001933 Neoparen No. 1 1000 mL 870 0.789 0.706

Table 7 Examples of precision and recall of medicines for internal use

Receipt computerizedsystem code

Medical specificationname

Number ofappearance

Precision Recall

616130532 Cefzon 100 mg 891 0.519 0.167612320346 Selbex 50 mg 2162 0.435 0.077611140694 Loxonin 60 mg 1215 0.383 0.043610411058 Flomox 100 mg 1209 0.377 0.053612520024 Methergin 0.125 mg 109 0.318 0.108620003469 Plavix 75 mg 242 0.305 0.093613330003 Warfarin 1 mg 1494 0.290 0.047610451009 Prograf 0.2 mg 425 0.227 0.082

Prediction of Clinical Practices by Clinical Data … 11

3.4 Discussion

In this study, we used machine learning and all clinical practice data of the daybefore as teacher data, and evaluated predictability of clinical practices for a patientof a day. Although there are some differences in each data category, some clinicalpractices have high precision rate and recall rate and others have low precision rateand recall rate. Generally, recall rate seems to be positively correlated with preci-sion rate especially when precision rate is more than 0.4 as shown in Fig. 4. On theother hand, when precision rate is less than 0.4, recall rate trends to be low.Furthermore, from the aspect of the number of data included in each data category(i.e. the number of times that is appeared as a clinical action), the precision rate ofthe practice with higher frequency tends to be higher. It is assumed that largeramount of data improves the prediction performance. Additionally, the basic codesfor medical calculation appear almost every day. It is assumed that it is easy topredict these codes. Concretely, data categories such as ‘injection’ and ‘procedures’have relatively high precision rate and recall rate as shown in Table 6 and Fig. 5. Itis seems that this is the result of prediction by learning quantitatively patterns ofpractices as training data. On the other hand, data categories such as ‘medicines for

Fig. 5 Scatter diagram ofprecision rate versus recallrate of others (injection)

Fig. 6 Scatter diagram ofprecision rate vs. recall rate ofmedicines for internal use

12 T. Nakai et al.

internal use’, ‘potion’, ‘medicine for external use’ and ‘examination’ have relativelylow precision rate and recall rate as shown in Table 7 and Fig. 6. Especially, a lot ofmedicines are included in categories such as ‘medicines for ‘internal use’, ‘potion’and ‘medicine for external use’. Therefore, the frequency of appearances of indi-vidual code is not relatively high. As a result, it seems that linear svm was not beenable to learn in the 10-fold cross-validation sufficiently. In fact, codes for medicinewas 1163 types, whereas, codes that were able to be calculated precision rate wereonly 65 types. The linear svm could predict more types of medicine if we could getmore data. On the other hand, the precision rate and recall rate of the codes of‘examination’ with high frequency are high relatively but the precision rate andrecall rate of the practice which appears rarely is low relatively.

As a result, we consider that the precision rate is high when the frequency ofappearance is increased. We intend to improve the prediction performance for moreclinical practices by using larger data.

References

1. Bates, D., et al.: Reducing the frequency of errors in medicine using information technology.J. Am. Med. Inform. Assoc. 8(4), 299–308 (2001)

2. Jensen, P.B., Jensen, L.J., Brunak, S.: Mining electronic health records: towards better researchapplications and clinical care. Nat. Rev. Genet. 13(6), 395–405 (2012)

3. DPC System (Summary and Basic concept DPC/PDPS). http://www.mhlw.go.jp/stf/shingi/2r985200000105vx-att/2r98520000010612.pdf (in Japanese)

4. Summary of a revision of payment for medical services in Heisei 26. http://www.mhlw.go.jp/file/06-Seisakujouhou-12400000-Hokenkyoku/0000039616.pdf (in Japanese)

5. Abe, S.: Approach support vector machines for pattern recognition. Morikita Publishing Co.,Ltd. 55, 16–26 (2013) (in Japanese)

6. LIBLINEAR—A Library for Large Linear Classification from http://www.csie.ntu.edu.tw/∼cjlin/liblinear/

Prediction of Clinical Practices by Clinical Data … 13

Method for Detecting Drug-InducedInterstitial Pneumonia from AccumulatedMedical Record Data at a Hospital

Yoshie Shimai, Toshihiro Takeda, Shirou Manabe, Kei Teramoto,Naoki Mihara and Yasushi Matsumura

Abstract Drug-induced interstitial pneumonia (DIP) is a serious adverse drugreaction. The occurrence rete of DIP was evaluated by clinical trial before availablein the market. However, due to limited number of cases in clinical trials, it may beinapplicable to the real market. We aimed to seek a method to evaluate theoccurrence rate of DIP using clinical data warehouse at a hospital. Initially wedeveloped a method that assesses whether presence of IP was written in reports bynatural language processing. Next we detected DIP by estimating IP before, duringand after the drug administration. Presence of IP was determined according to thereports of CT if CT was performed, otherwise it was determined based on thechanges in the results of chest X-ray, level of KL-6 or SP-D. DIP was determinedaccording to the pattern of presence of IP in each phase. In this study we choseamiodarone as a target drug. The number of patients who suffered from IP causedby amiodarone was 16 (3.9 %), including one definitively diagnosed and 15 strongdoubt cases. Most of them could be validated by medical record chart. Using thismethod, we were able to successfully detect occurrence of DIP from accumulateddata in a hospital information system.

1 Introduction

Various adverse events occur related to medication use. Information regarding therisk of adverse events for each medicine is important for clinical practice. The safetyof medicines is evaluated in clinical trials before the drugs are introduced into themarket. However, because the number of subjects in clinical trials is limited, infor-mation regarding adverse events generated in clinical trials may be inadequate [1].

Y. Shimai (✉) ⋅ T. Takeda ⋅ S. Manabe ⋅ K. Teramoto ⋅ N. Mihara ⋅ Y. MatsumuraMedical Informatics, Osaka University Graduate School of Medicine, Osaka, Japane-mail: shimai@hp-info.med.osaka-u.ac.jp

© Springer International Publishing Switzerland 2016Y.-W. Chen et al. (eds.), Innovation in Medicine and Healthcare 2015,Smart Innovation, Systems and Technologies 45,DOI 10.1007/978-3-319-23024-5_2

15