A knowledge centric methodology for dental implant technology assessment using ontology based patent analysis and clinical meta-analysis

A knowledge centric methodology for dental implant technology

assessment using ontology based patent analysis and clinical

meta-analysis

Charles V. Trappey

, Amy J.C. Trappey

, Hsin-Yi Peng

, Li-Deh Lin

, Tong-Mei Wang

Department of Management Science, National Chiao Tung University, Hsinchu, Taiwan b

Department of Industrial Engineering and Engineering Management, National Tsing Hua University, Hsinchu, Taiwan c

School of Dentistry, National Taiwan University, Taipei, Taiwan

a r t i c l e

i n f o

Article history:

Received 19 September 2013

Received in revised form 23 February 2014 Accepted 5 March 2014

Available online 13 April 2014 Keywords: Patent analysis Ontology schema Meta-analysis Clinical trials

a b s t r a c t

The medical equipment industry has been one of the fastest growing sectors of the decade with predicted global sales reaching US$ 430 billion in 2017[22]. During the period from 1995 to 2008, the patent appli-cations in medical technology increased rapidly worldwide (World Intellectual Property Organization, 2012). Patent analysis, although useful in forecasting technology development trends, has posed a chal-lenging analysis task since the volume and diversity of new patent applications has surpassed the ability of regular firms and research teams to process and identify relevant information. Further, medical related technologies rely on clinical trials to validate and gain regulatory approval for patient treatment even though patents, protecting the intellectual property rights of inventors, have been granted. This research focuses on developing a knowledge centric methodology and system to analyze and assess viable medical technology innovations and trends considering both patents and clinical reports. Specifically, the design innovations of dental implant connections are used as a case study. A novel and generic methodology combining ontology based patent analysis and clinical meta-analysis is developed to analyze and identify the most effective patented techniques in the dental implant field. The research establishes and verifies a computer supported analytical approach and system for the strategic prediction of medical technology development trends.

Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction

The World Health Organization[59]reported that about 30% of

the population whose age is between 65 and 74 are likely to lose some of their natural teeth. Dental implants are a medical treat-ment with a range of products used to restore oral functions when losing teeth to caries, periodontitis, or accident. The global dental implant and prosthetics market was valued at US $ 6.8 billion dol-lars in 2011 and is expected to reach US $10.5 billion doldol-lars in

2016 [5]. The surgical success and consumer acceptance have

increased the global demand for implants and the prosthesis mar-ket. The demand for dental implants continues to attract compa-nies and researchers to improve the design and development of dental implant components, devices, and techniques.

Modern dental implants have been used since the 1960s[1].

Since then, many improvements in dental implants have been introduced resulting in a variety of patents filed and granted.

Fig. 1depicts the number of patents related to dental implants in the United States Patent and Trademark Office (USPTO) from 1990 to 2012. Most dental implants consist of implant bodies (screws embedded in the jawbone), abutments (the platform for connection between the implant and crown), and crowns (the aes-thetic and functional artificial replacement to the tooth). Many forms of dental implant connections have been developed as a crit-ical part of dental implant R&D to improve torque transfer, gain stability between the implant body and the abutment, and subse-quently minimize implant connection failure. Thus, this research focuses on the case study of dental implant connections to demon-strate the knowledge centric methodology of DS technology assessment and trend prediction.

The FDA[17]establishes regulations for dental implants

abut-ments and enforces rigorous procedures of mechanical tests and clinical studies. However, there are still some implant designs in

http://dx.doi.org/10.1016/j.aei.2014.03.001 1474-0346/Ó 2014 Elsevier Ltd. All rights reserved.

⇑ Corresponding author. Tel.: +886 35742651; fax: +886 35722204.

E-mail addresses: [email protected] (C.V. Trappey), trappey @ie.nthu.edu.tw (A.J.C. Trappey), [email protected] (H.-Y. Peng), [email protected](L.-D. Lin),[email protected](T.-M. Wang).

Contents lists available atScienceDirect

Advanced Engineering Informatics

use with relatively high failure rates that pass the FDA regulations

[46,2]. This research focuses on efficiently and accurately predict-ing medical technology trends with computer supported analyses of published patents and collective clinical trial literatures. Patent documents contain technical details of the innovations and inven-tions. In order to better understand the performance of new med-ical technologies and gain approval from regulatory agencies, reports of updated clinical trials on human subjects are also collec-tively analyzed. The objective of this research is to combine text mining, data mining, and meta-analysis within a specific domain, i.e., dental implant connections, including related patents and the corresponding clinical trials to better understand successful trends in medical technology innovation and adaptation.

2. Literature review

The literature related to dental implants, ontology, knowledge discovery, patent analysis and meta-analysis are discussed in this section. We first provide a brief background review of dental im-plants to provide a better understanding of the domain knowledge. The entire analytical procedure is based on domain specific (DS) ontology. The definition of ontology for knowledge representation and the ontology-based knowledge discovery applying text and data mining techniques are depicted in Sections2.2 and 2.3. In

Sec-tions2.4 and 2.5, macro- and micro-patent analyses and

system-atic meta-analysis of clinical literatures are described. 2.1. Dental implants

A dental implant is an artificial tooth root which is placed into a patient’s jaw to hold a prosthesis replacing a missing tooth. Most dental implant systems consist of a crown, an abutment, and an implant body. Dental implant procedures are divided into two stages. First, the implant body is implanted into the jaw. Second, once the implant is stable inside the jaw bone, the abutment is connected to the implant body and the crown is attached to the abutment. The abutment is the component for connecting the im-plant body and the final outer crown or artificial tooth. The abut-ment usually connects to the implant body via a screw. One of the features which vary among dental implant systems is the type of connection that allows the abutments and prosthesis to be at-tached to the implant body. These connections include external connections, internal connections, or Morse taper connections

[39]as shown inFig. 2. Implant systems with external connections

have a polygonal protrusion at the upper part of the implant body. For internal connections, the implant body has a notched polygonal cavity at the upper part which matches the polygonal protrusion at the abutment end. The internal hex connection combined with a Morse taper implant body is considered an alternative to the

exter-nal hex implant[39]. Marginal bone loss (around the connection) is

used as one of the most critical indicators of dental implant quality and long term stability (Papaspyridakos et al., 2012). New designs

of the implant-abutment interface, such as one-piece implants and platform switching (an implant body connected with a narrower abutment), have increased the success rate of implant technology

[3]. However, the comparison of design effects has not been

stud-ied and there is no research demonstrating which design has a greater impact on implant quality and long term success.

2.2. Ontology

An ontology is an explicit specification of a knowledge domain and consists of a set of concepts, relations, objects and functions

[24]. Another definition given by Grüninger and Fox[25] is the

ontology is a formal description of a set of entities and their prop-erties, behaviors and relations. Therefore, the ontology is consid-ered to be a representational model of some portion of a real

world knowledge domain[28]and is a set of objects and the

rela-tionships among these objects, which may be represented in the form of graphs and figures. The ontology types include terminology based ontologies, information ontologies, and knowledge modeling

ontologies[26]. Domain ontologies focus on a specific field and

de-scribe the concepts of the domain entities as well as the attribute values and characteristics of the domain. Researchers focus on a specific domain through the visualization of knowledge, but there are no standard procedures to build specific ontology. Ding and Foo

[15]define the methods used to construct an ontology as

bottom-up, top-down and middle-off depending on how the ontology schema is initiated (e.g., bottom-up patching and synthesizing, top-down detailing and propagating, or working from both ends). The ontology reasoning technique is widely applied in the applica-tion of expert systems, artificial intelligence, and knowledge

man-agement in industry. Liou et al. [34] proposed a development

procedure that includes planning, design, testing and modification, deployment, and integration to build an ontology based database.

Trappey et al. [49] proposed a method for automatic patent

document summarization which used ontology trees to

Fig. 1. US patents (1990 through 2012) filed and issued that are related to dental implants.

automatically shorten patents into abstracts containing the key

concepts of lengthy patent documents. Lin[32]used domain

ontol-ogies to classify and analyze the technology coverage of patents

within companies. For specific dental implant domains,[53]define

a domain ontology schema for patents search and analysis. 2.3. Knowledge discovery

Knowledge discovery is a process to extract implicit, previously unknown, and potentially useful knowledge from known data

which are relevant and useful[18]. Depending on the data type,

knowledge discovery is divided into two categories, i.e., discover-ing knowledge in a database (KDD or data mindiscover-ing) and in a textual document base (KDT or text mining). KDD is the process used to automatically discover previously unknown patterns, rules, and

other types of content in large volumes of data[19,11]. The steps

of KDD consists of identifying the analytical objectives, creating the target data set, cleaning and preprocessing data, data reduction or projection, using data mining techniques or algorithms to search the patterns of data, and interpreting the patterns[19].

Text mining is commonly known as knowledge discovery when

analyzing documents or text[20]. The framework of text mining

consists of text refinement and knowledge distillation[48]. Term

frequency-inverse document frequency (TF-IDF) is a statistical method which uses the frequency of word occurrence in text to

re-flect the importance of a word in a given document set[45]. Salton

and Buckley[44]reported that the length of a document can affect

the term weight, therefore, TF-IDF was modified and called nor-malized term frequency-inverse document frequency (NTF-IDF). The number of words in a set of documents is used to normalize the value of term frequency. For patent analysis, text mining tech-niques, including text segmentation, summary extraction, feature selection, term association, cluster generation, topic identification

and information mapping, are commonly applied [54]. Trappey

et al.[49]combined the techniques of ontology based text mining

and data mining to develop new methodologies for patent analysis and case study documentation.

2.4. Patent analysis

A patent is a form of intellectual property. Patent documents contain extensive information about the technology research and development results. Therefore, patents as knowledge documents are invaluable sources for investigating and analyzing new

tech-nologies and trends [31]. A typical patent analysis scenario

includes tasks such as searching, segmenting, abstracting,

clustering, visualizing, and interpreting technical texts[54]. The

approaches used for patent analysis include patent maps, patent abstraction, patent clustering and classification, and rating patent quality. A patent map is a visual display of patent documents using

graphic software[47]. The patent map helps companies

differenti-ate designs and identify new design opportunities[6]. Since the

number of patents continues to increase rapidly in world patent corpuses, automatic methods of extracting and synthesizing patent information and intelligence have become important strategic tools. The principle steps of document summarization are selecting key representative words via text mining and then generating

sen-tences as the final document summary[4]. Document

categoriza-tion (also called classificacategoriza-tion) is a method which assigns documents to pre-defined classes, whereas document clustering is a method of generating homogeneous sets according to a pre-specified properties or indices without pre-defined categories.

Trappey et al. [52] applied ontology-based artificial neural

networks to systematically and automatically categorize or classify

patent documents. Further, Trappey et al. [50] proposed a

non-exhaustive clustering methodology for automatically

forecasting and analyzing technical trends to develop sustainable corporate R&D strategies. Patent quality represents the market va-lue of patents. The quality of patents can be measured using com-bined indices, such as forward and backward citations, numbers of International Patent Classifications, numbers of US Patent

Classifi-cations, and numbers of claims[51,40]and[16].

2.5. Systematic meta-analysis of clinical reports

Meta-analysis is a research approach where previously pub-lished studies in a given domain are collected and analyzed to

inte-grate findings using statistical analyses [10]. Meta-analysis

provides summarized analytical results combining independently conducted studies. Meta analysis is commonly used to appraise, summarize, and communicate the statistical results of several sim-ilar studies[23]. Meta-analysis is useful and frequently applied to help researchers identify effective trends in a specific domain[30]. Meta-analysis is widely used in medical research because of the large volume of related clinical trials. If the data are related, a sum-mary of effects can be quantitatively assessed using meta-analysis. But meta-analysis should only be conducted when the studies are similar between research questions, populations, and outcomes. The major challenge with meta-analysis is to summarize and syn-thesize the results across a diverse range of related studies that use different methodologies, samples, and experimental designs. 3. Methodology

The methodology for this research study includes a domain spe-cific ontology based patent analysis of dental implant connections and a meta-analysis of the related clinical trials. An overview of the

methodology flow is shown inFig. 3. The individual steps are

de-scribed briefly in the sub-sections and include the cited literature. 3.1. Domain specific patent analysis

The processes of extracting knowledge from domain specific pat-ent documpat-ents are described in the following steps, i.e., collecting data, extracting key phrases, creating the domain ontology, and ana-lyzing clusters of sub-technologies. This stream of analysis follows the detailed procedure and algorithms developed in one previous

research[53]. Thus, the methodology will only be briefly described.

First, the patent data collection procedure adopted in this study

Domain patent analysis Clinical reports analysis

Patent collection

Key phrase extraction from patent documents Ontology engineering Ontology verification by expert Patent clustering Clinical report collection

Key phrase extraction from clinical reports

Ontology modification Meta-analysis of clinical reports Integrated interpretation of technology trends

follows three steps. The first step is to search the patents from the USPTO database over a thirty-six year timeframe from 1976 to Au-gust 2012. The second step is to study the patent context and select the patents related to the domain of dental implant connections. Fi-nally, a dental implant expert verifies the selected patents.

After collecting patents, the key phrases of the selected patents

are extracted automatically using text mining techniques[56]. The

key phrase extraction process produces a ranked list of the most frequently appearing phrases in the patents. Key phrases are ranked based on normalized term frequency (NTF) values in order to extract the common key phrases among the set of dental im-plant connection patents. NTF is the method used to calculate the frequency of a phrase among documents while the number of words in a document is used to normalize the term frequency.

The NTF value of key phrases is computed usingEq. (1)where ntfjk

is the NTF value of term j in document k, tfjkis the occurrence

num-ber of term j in document k, and dnkis the total number of words in

document k[44]. ntfjk¼ tfjk PN s¼1dns N  dnk ð1Þ

The formula used to calculate NTF-IDF value is shown in Eq. (2)

where idfjis the inverse document frequency of key phrase j. Given

that dfjis the number of documents where the term j appears, the

formula for inverse document frequency is computed usingEq. (3).

NTF  IDF ¼ ntfjk idfj ð2Þ idfj¼ log2 N dfj   ð3Þ The domain-specific ontology, thereafter, can be built using the key phrases extracted automatically from the domain patents and ver-ified by domain experts. The key phrases with strong associations are defined as concept relations which are automatically linked be-tween key phrases. The concept relation links are also verified and modified by dental implant experts. Meantime, additional key phrases, found in the next stage clinical literature or identified by domain experts, can be added to enrich the ontology schema. Based on the built ontology, patents related to dental implant connections are collected and are further clustered using the K-means algorithm. The clusters are generated based on the patents’ key phrase similar-ities. The key phrases representing each patent group are used to better describe the sub-technological attributes of the dental im-plant connections. The procedure for applying the K-means algo-rithm is summarized as follows:

1. Select K clusters as an initial partition;

2. Assign each patent to its closest cluster center to produce a new partition;

3. Compute new cluster centers of new partitions;

4. Repeat step 2 and step 3 until the cluster membership is stable.

3.2. Systematic clinical trial literature analysis

The second part of the research is to analyze the collection of clinical trial reports. The performance results of the patented tech-nologies and the treatment effect sizes are compared across the clinical trial literature sample. The process consists of collecting as large and as comprehensive a sample of peer reviewed clinical trials as possible, extracting the key phrases of these reports, improving the domain-specific ontology by adding new key phrases derived from this literature, and performing the meta-analysis on the abstracted statistics.

The study first searches the literature of clinical trials for differ-ent technologies of ddiffer-ental implant connections archived in the

PubMed database from January 1980 to November 2012. PubMed’s primary data source is MEDLINE which covers the fields of den-tistry, medicine, veterinary medicine, nursing, health care systems, and preclinical sciences. The key search phrases used are ‘‘implant connection’’ and ‘‘implant connection interface’’ and the type of publication is clinical trials. The papers related to dental implant connections are selected and further verified independently by the dental implant specialist who also reviewed the cited refer-ences within each paper for possible inclusion.

The list of the key phrases is generated using text mining tech-niques and the key phrases are ranked based on their normalized

term frequencies (NTF) values [29]. After extracting the key

phrases of the article, the common criteria for implant survival and the clinical trial evaluation parameters are incorporated into the dental implant connection ontology. The ontology is then cor-rected to include branches and nodes which represent the most effective clinical trials. The meta-analysis steps used in this study include defining the problem, collecting data, recording the charac-teristics of the clinical trials, and summarizing the results. The re-search questions are specified in a clear, unambiguous and structured form with a statement of the intervention, the patient demographics, and clinical outcomes.

The literature search strategy follows two steps which are the systematic electronic search and the independent manual search and verification. After the electronic database search of articles re-lated to dental implants, an independent dental researcher elimi-nates articles which may be biased or poorly suited to the domain. A data extraction tablet is used to record information from the articles including authors, year of publication, follow-up peri-od, implant system, site and number of implants, type of dental im-plant connection, and the survival rate of dental imim-plants.

The findings from separate studies are aggregated using a three-stage quantitative assessment process. The first three-stage is to calcu-late the effect size and weight of each study. There are several types of effect size measures including risk difference and mean difference. Risk difference is a measure obtained by subtracting the risk of an event happening in one group from the risk in an-other group. The mean difference is obtained by subtracting the mean of an event happened in one group from the mean in another group. The formula used to calculate the risk difference and the standard error are shown inEq. (4) and Eq. (5)where Ptiis the

pro-portion of events occurring in the treatment group, Pci is the

pro-portion of events occurring in the control group, SERD,i is the

standard error of i-th study, Nt

i is the sample size of treatment

group in the i-th study, and the Nci is the sample size of control

group in the i-th study. RDi¼ Pti P c i ð4Þ SERD;i¼ ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Pt ið1  P t iÞ Nti þP c ið1  P c iÞ Nci s ð5Þ When the outcomes of the studies are continuous, the effect size is

calculated as the mean difference. The formula is shown inEq. (6)

where Mti represents the treatment group and M

c

i represents the

i-th study control group. SDi is the standard deviation of either

groups and SEMD,iis the standard error of i-th study. The values

are computed using the formula shown inEq. (7)andEq. (8).

MDi¼ Mti M c i ð6Þ SDi¼ ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ðNti 1Þ  SD t2 i þ ðN c i 1Þ  SD c2 i ðNtiþ N c i  2Þ v u u t _ð7Þ SEMD;i¼ ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi SDti2 Nt i þSD c2 i Nc i v u u t _ð8Þ

The pooled effect sizes are aggregated using either the fixed-effect model or the random-effect model[27]. If the effect sizes are fixed, the effect sizes are pooled using the fixed-effect model. If the effect sizes vary from study to study, the pooled effect size should be cal-culated using the random effect model. The weight of the study is given using the inverse variance method. Therefore, a smaller vari-ance yields a larger weight. The optimal weights are calculated

using Eq. (9). The pooled effect size is a linear combination of

weights and effect sizes as shown inEq. (10). The standard error

SE_ESof ES is computed usingEq. (11). Wi ¼ ðSE2iÞ 1 ð9Þ ES ¼X k i¼1 Wi Pk i¼1Wi  ESi ð10Þ SE_ES¼ ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Xk i¼1Wi  1 r ð11Þ The heterogeneity of the effect size is tested with the hypothesis H0:

ES1= ES2= . . . = ESkversus the alternative hypothesis that one ESi

differs from the remainder. The test of H0is based on the statistic

Q which is shown inEq. (12). If all k studies have equivalent effect sizes, then the test statistic Q has a chi-square distribution with k  1 degrees of freedom. On the other hand, if Q exceeds the 100(1 

a

Q ¼X k i Wi ESi Pk i¼1WiESi Pk i¼1Wi !2 ð12Þ If the hypothesis is rejected, then the random-effects model is used to pool the effect sizes of the studies. In both the fixed and random effect models, the pooled effect size is computed using the weighted

mean. For the random-effects model, the ESiare not fixed and are

considered not to have a chi-square distribution. Therefore, the var-iance

s

is SE02 i ¼ SE 2 i þ

s

s

s

and Wiare given inEq. (15). The weights of studies in random-effect

model are given by W0i¼ ðSE

02 i Þ

1

ð15Þ The pooled effect size is given by

ES0_¼X k i¼1 W0 i Pk i¼1W 0 i  ESi: ð16Þ Table 1

Patents related to dental implant connections. No. US Patent No. Title 1 US5100323 Dental implant

2 US5106300 Dental implant attachment structure and method 3 US5415545 Dental implant system

4 US5433606 Interlocking, multi-part endosseous dental implant systems 5 US5449291 Dental implant assembly having tactile feedback 6 US5704788 Dental implant abutment screw lock

7 US5733122 Dental implant attachment assembly including device and method for resisting loosening of attachment 8 US5759034 Anatomical restoration dental implant system for posterior and anterior teeth

9 US5904483 Dental implant systems and methods

10 US6419492 Dental implant system incorporating an external hex and Morse tapered walls 11 US6431867 Dental implant system

12 US6464500 Dental implant and abutment system 13 US6857874 Dental implant structure

14 US7014464 Multi-part abutment and transfer cap for use with an endosseous dental implant with non-circular, beveled implant/abutment interface 15 US7090495 Dental implant screw and post system

16 US7300282 Bio-functional dental implant 17 US7338286 Dental implant system

18 US7682152 Force distributing dental implant assembly

19 US8142193 Compound angular joint for connecting an abutment to a dental implant in a predefined angle 20 US8162663 Dental implant

Table 2

The matrix of normalized key phrase frequencies derived from the patent collection.

Key phrases Patents NTF Sum

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Implant 72 105 113 197 136 87 48 146 85 210 175 152 78 95 95 78 150 23 93 222 2360 Screw 67 27 0 17 25 105 151 20 37 30 43 7 109 11 142 69 20 48 8 34 971 Dental implant 38 3 22 24 19 39 3 53 4 55 18 67 74 13 34 37 19 22 55 25 623 Cavity 0 2 0 78 0 0 0 3 8 74 0 0 0 0 0 4 47 11 0 91 317 Bone 50 3 0 6 18 4 0 13 7 11 63 20 0 0 0 14 7 27 11 20 274 Hexagonal 107 2 0 2 4 12 2 9 4 20 0 0 5 0 22 0 12 6 0 6 213 Fixture 0 0 0 0 0 23 0 89 0 0 0 0 64 0 4 0 0 0 0 0 179 Prosthesis 0 0 16 0 10 14 4 0 3 28 8 13 0 0 22 0 0 0 31 3 151 Splines 0 0 0 0 91 0 0 0 0 0 0 0 0 0 0 45 0 0 0 0 136 Crown 0 0 0 0 0 6 0 0 0 0 23 0 9 0 0 81 0 0 0 0 120 Protrusion 52 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 39 91

The standard error SE0

ESof mean effect size estimate ES0is the square

root of its variance and is given by SE0 ES¼ ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Xk i¼1W 0 i  1 r : ð17Þ

Finally, the results are pooled to calculate the estimated mean effect size. A forest plot[35]is used to illustrate and compare the relative strength of the treatment effects between studies.

4. Case study

The proposed methodology is applied to predict the technical trend for dental implant connections. The case study includes an ontology-based patent analysis and a clinical trial meta-analysis for the specific dental device domain.

4.1. Patent analysis of implant connections

The patents related to dental implant connections were col-lected and the key phrase descriptions of these patents were auto-matically extracted using text mining techniques to create the ontology diagram.

4.1.1. Patent data collection and the key phrases extraction

First, the patent documents related to dental implants were col-lected from the USPTO using the keyword ‘‘dental implant’’ for the search. As a result, a total of 564 patents were collected. The patent documents not related to dental implant connections were excluded. Next, the dental implant specialist verified the patents which were suitable for building the ontology of dental implant connections. As a result, 20 patents were selected as dental implant

connection related patents.Table 1summarizes the sample patent

data.Table 2presents the list of key phrases extracted automatically

from related patents using the IPDSS software[57]with text mining

technique described in Section3.1. 4.1.2. Domain-specific ontology creation

Fig. 4depicts the ontology schema of the dental implant con-nection technologies, consisting of key phrases that describe the key concepts and critical linkages between the implant body and the abutment. The ontology defines the structure of the dental im-plant and the types of connections. The ontology tree is divided into several parts. The sub ontology schema describes implant body, abutment screws, and the level of connections in the onto-logical hierarchy. The ontology is also used to identify the design variations as shown inFig. 2.

4.1.3. Sub-technology analysis

The training patents and the testing patents were collected and categorized into several groups. As a result, the patents were sep-arated into 4 groups and the clusters are listed inTable 3.

The key phrases extracted from patents related to dental

im-plants are shown inTable 4. There are several common key words

in each cluster; e.g., dental implant, bone, and cavity, that are inef-fective to identify the characteristics of a specific cluster. Therefore, key words appearing in two or more clusters are excluded. Cluster 1 is best described as containing patents related to one-piece abut-ments. The patents of Cluster 2 are related to the designs of inter-nal implant connections. Cluster 3 contains patents related to connections, dental implant systems, and washers and therefore includes multi-feature implants. Cluster 4 is related to the charac-teristics of the connections between the abutments screw and im-plants. The results of the patent cluster analysis and the ontology schema reveal several key phrase groups including the characteris-tics of an implant head, connections, and one-piece abutments.

4.2. Analysis of clinical papers related to implant connections The clinical trial papers are reviewed in order to compare the performance of designs related to the patent ontology. The meta-analysis results are used to summarize the effectiveness of the pat-ented designs.

4.2.1. Clinical paper collection

The electronic search for clinical trials used the PubMed data-base. The key phrases used for the search included ‘‘implant connection,’’ ‘‘implant abutment interface,’’ and the type of publi-cations included were limited to human clinical trials. The articles retrieved were published between 1992 and 2012. After the initial electronic search, the abstracts and the content of the retrieved articles were independently analyzed by the dental implant spe-cialist. The articles related to dental implant connections were ar-chived and the key phrases were reviewed. The meta-analysis statistics were used to measure the effectiveness of the different patented designs. Finally, the clinical trials selected for the

com-parison of patented designs are shown inTable 5.

4.2.2. Key phrases extraction and ontology modification

The key phrases of the clinical trials related to dental implant connections were extracted. The full text of selected papers was analyzed and the key phrases were ranked based on their NTF

va-lue shown inTable 6. The key phrases with the greatest frequency

of occurrence describe the common criteria for implant survival and the clinical trial evaluation parameters. These newly extracted key phrases were incorporated into the dental implant connection ontology with additional branches and nodes added to the schema indicating common criteria for clinical trials (Fig. 5).

4.2.3. Results of the clinical trial meta-analysis

The purpose of meta-analysis is to examine whether there are differences in survival rates and marginal bone loss between differ-ent types of patdiffer-ented ddiffer-ental implant connections. The following inclusion criteria were applied to select the clinical reports for the comparative meta-analysis between the treatment and the control groups.

1. A human study population.

2. The clinical trial includes the key words from the dental implant connection techniques ontology.

3. The type of publication is a clinical trial. 4. The minimal follow-up period was 1 year.

5. Only studies that provide sufficient data for coding and anal-ysis are included.

A preliminary search found 48 related articles, but only ten studies were electronically chosen as valid reports that satisfy the above four criteria. A manual search was further conducted and an additional 5 studies were included in the meta-analysis yielding a total of 15 studies published between 2001 and 2012. Among the 15 articles, one study applied external connections, one study focused on internal (non-Morse taper) connections, five studies used Morse taper connections, and eight studies specifi-cally compared dental implant connections with platform switch or platform match connections. The average follow-up period of

Table 4

Key phrases of dental implant connection patents clusters.

Cluster 1: One-piece abutments Cluster 2: Internal connections Cluster 3: Muti-features Cluster 4: Connections

Converter Tiltable Plasma Feedback

Inserting Thermoplastic Implant-abutment Anti-rotation cavity

Slit Trunco-conical Longitudinally Integrating

Supragingival Joint Cones Conduit

Coronally Adjustable Tab Non-curved

Triangle Unthreaded Non-circular Sidewall

Frusto-conical Heat-removable Pocket Morse

Bone-embedded fixture Stability Geometry Interproximal

Alveolus Coaxial Dental implant-abutment Tactile

One-piece Frustoconical Prosthetic device Bone integrating

Aluminum oxide Hex-shaped cavity Washers Extent

Table 3

Clustering results of implant connection patents.

Implant clusters Training patents Testing patents Cluster 1

One piece abutments

US5040982 US6358050 US6857874 US5125839 US8057229

Cluster 2 Internal

connections

US4758161 US7108510 US5759034 US5071350 US7665990 US6464500 US5399090 US7780447 US8142193 US6843653 US8123524

US7059855 Cluster 3

Multiple features US4826434 US6168436 US5100323 US5904483 US5071351 US7090493 US5106300 US7014464 US5188800 US7104797 US5415545 US7090495 US5435723 US7207800 US5704788 US7300282 US5482463 US7708559 US5733122 US7682152 US5636989 US8038442

US5810589 US8202088 US6068479

Cluster 4

Connections US5195892 US7112063 US8162663 US6419492 US5458488 US7484959 US5433606 US6431867 US5449291 US7338286

each study was between 1 and 5 years. Among the final fifteen clin-ical reports, there were seven observational studies conducted

without control groups. Drago [14] used dental implants with

external connections to study the effectiveness of Gold-Tite square abutment screws and reported that the survival rate reached 99%

and the rate of abutment screw loosening was 0.96%. Norton[41]

used dental implants with internal connections and the study re-vealed that the internal connections prevent mechanical complica-tions such as abutment screw loosening or breakage. Döring et al.

[13], Mangano et al.[36], Degidi et al.[9], Mangano et al.[37], and

Mangano et al.[38]measured the performance of Morse taper

con-nections for implants. Mangano et al.[37]reported that the rate of

abutment screw loosening was 0.66%, which showed that Morse taper implants have higher mechanical stability and significantly

reduced prosthetic complications [36]. Nonetheless, the clinical

studies listed inTable 7were not used to compare effect sizes since control groups were not used. Therefore, only studies with control groups and treatment groups were chosen for the meta-analysis.

Table 8lists the data extracted from the final sample of studies, which compared both implant failure rates and marginal bone loss as the performance indices. These studies applied platform switch connectors as the treatment groups and platform match connec-tors as the control group for the clinical trial.

For binary data, the risk difference was used to calculate the effect size and the event was defined as implant failure. There were 2 implant failures from platform switched implants and 3 implant

failures from the platform match group. Table 9 shows the risk

difference (RD) of individual studies. The analysis was based on the fixed-effect model. The pooled risk difference is 0.000 and the 95% confidence interval was between 0.012 and 0.012, and the p-value is 0.957 indicating that there was no signif-icant difference between trials regarding implant failures. The rel-ative strength of treatment effects in multiple quantitrel-ative studies and the pooled risk difference are shown inFig. 6.

For marginal bone loss (MBL), the mean difference was used to compute the effect size. The mean difference was obtained by sub-tracting the mean MBL value from the platform switch group from

the mean MBL value of the platform match group.Table 10

pre-sents the mean differences from the individual studies. The ran-dom-effects model was used to aggregate the effect size. The variance

s

stud-ies are shown in Fig. 7 and the pooled mean difference is

0.33 mm and the 95% confidence interval was between 0.49 mm and 0.17 mm, and the p-value was less than 0.001. The result of the meta-analysis shows that the platform switch im-plant has significantly less marginal bone loss across the trials. This is a key finding since there are few studies reporting no significant differences in marginal bone losses between platform switch and platform match implant techniques.

The meta-analysis results revealed that the rates of implant failure were not significantly different between platform switch and platform match connections, while the platform switch connections consistently and significantly outperform platform match connections in preventing marginal bone loss.

Table 6

Key phrases derived from clinical papers of dental implant connections and their normalized term frequencies (NTF).

Key phrases Clinical papers NTF Sum

1 2 3 4 5 6 7 8 9 10 11 12

Marginal bone loss 0.0 0.0 22.4 0.0 0.0 0.0 0.0 22.0 10.1 5.1 0.0 0.0 59.5

Implant placement 5.7 0.0 4.7 0.0 6.6 5.7 0.0 8.7 0.0 0.0 4.1 19.3 54.7

Platform switching 0.0 0.0 0.0 0.0 0.0 0.0 0.0 8.7 0.0 0.0 0.0 32.1 40.8

Implant abutment connection 0.0 0.0 0.0 0.0 0.0 0.0 11.9 0.0 8.6 5.1 0.0 9.0 34.6

Abutment loosening 0.0 0.0 0.0 5.3 0.0 0.0 19.9 0.0 0.0 0.0 0.0 0.0 25.1

Marginal bone resorption 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 18.7 0.0 0.0 5.1 23.9

Morse taper connection 0.0 0.0 0.0 7.0 0.0 0.0 0.0 0.0 0.0 7.2 8.2 0.0 22.4

Abutment 0.0 0.0 14.0 0.0 0.0 19.5 0.0 0.0 0.0 0.0 0.0 0.0 33.5

Taper connection implants 0.0 0.0 0.0 6.2 0.0 0.0 0.0 0.0 0.0 0.0 9.0 0.0 15.2

Screw loosening 0.0 9.0 0.0 6.2 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 15.2

Edentulous mandible 3.4 0.0 0.0 0.0 0.0 11.5 0.0 0.0 0.0 0.0 0.0 0.0 14.9

Microgap 0.0 0.0 0.0 7.0 0.0 0.0 7.0 0.0 0.0 0.0 0.0 0.0 14.0

Mandible 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 13.1 0.0 13.1

Platform switched abutment 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 13.0 0.0 0.0 0.0 13.0

Internal connection 0.0 0.0 0.0 0.0 0.0 0.0 0.0 6.3 0.0 5.8 0.0 0.0 12.1

Table 5

Clinical trials related to dental implant connections.

No. Studies Title

1 Cooper et al.[7] Treatment of edentulism using Astra Tech implants and ball abutments to retain mandibular overdentures 2 Drago[14] A clinical study of the efficacy of gold-tite square abutment screws in cement-retained implant restorations 3 Donati et al.[12] Immediate functional loading of implants in single tooth replacement: A prospective clinical multicenter study 4 Mangano et al.[36] Prospective clinical evaluation of 1920 Morse taper connection implants: Results after 4 years of functional loading

5 Crespi et al.[8] Radiographic evaluation of marginal bone levels around platform-switched and non–platform-switched implants used in an immediate loading protocol

6 Degidi et al.[9] Prospective study with a 2-year follow up on immediate implant loading in the edentulous mandible with a definitive restoration using intra-oral welding

7 Mangano et al.[37] Prospective clinical evaluation of 307 single-tooth Morse taper-connection implants: A multicenter study

8 Pieri et al.[43] Influence of implant-abutment interface design on bone and soft tissue levels around immediately placed and restored single tooth implants: A randomized controlled clinical trial

9 Veis et al.[55] Evaluation of peri-implant marginal bone loss using modified abutment connections at various crestal level placements 10 Linkevicius et al.[33] Influence of thin mucosal tissues on crestal bone stability around implants with platform switching: A 1-year pilot study 11 Mangano et al.[38] Morse taper connection implants supporting ‘‘planned’’ maxillary and mandibular bar-retained overdentures: A 5-year prospective

multicenter study 12 Peñarrocha-Diago

et al.[42]

Table 7

Prospective studies related to dental implant connections (without control groups).

Study Follow-up period Implant system Position No. of implants No. of implants in the end Survival rate External connections

Drago[14] 1 year Implant innovations Maxilla; Mandible 110 104 99%

Internal connections

Norton[41] 5 years Astra Tech Maxilla 23 14 can be reviewed n.r.a

Morse Taper connections

Döring et al.[13] 38 months Ankylos Anterior and posterior jaw regions 275 270 98.2%

Mangano et al.[36] 4 years Leone Maxilla; Mandible 1920 1884 97.6%

Degidi et al.[9] 2 years Ankylos Mandible 80 80 100%

Mangano et al.[37] 4 years Leone Maxilla; Mandible 307 302 98.4%

Mangano et al.[38] 5 years Leone Maxilla; Mandible 288 282 98%

a

n.r.: Not reported.

4.3. R&D strategy and market opportunities

According to the results of the patent analysis, the types of dental implant connections were identified. Using the patent clustering results, the main sub-technologies of dental implant connections were identified. The patent groups included one-piece abutments, internal connections, multi-features and connection techniques. The results of the clinical literature text mining

revealed that the dental implant evaluation parameters were based on implant failure rates and marginal bone loss. The marginal bone loss appears to be related to the implant-abutment interface. Few clinical studies reported significant advantages for platform switch

implants in preventing marginal bone loss[8,55,33]. However, the

results of the meta-analysis reveal that platform switch techniques do significantly reduce marginal bone loss levels when used for implant connections. Dental implants overall have a high survival

Table 9

Risk difference data of implant failure.

Study Treatment group Control group RD SE p-Value Weight

No. of failures No. of implants No. of failure No. of implants

Peñarrocha-Diago et al.[42] 1 72 1 69 0.001 0.020 0.976 1093.8

Pieri et al.[43] 1 19 2 20 0.050 0.083 0.546 8402.9

Fickl et al.[21] 0 75 0 14 0.000 0.047 1.000 150.3

Veis et al.[55] 0 89 0 193 0.000 0.009 1.000 52.8

Vigolo and Givani[58] 0 97 0 85 0.000 0.011 1.000 145.4

Crespi et al.[8] 0 30 0 34 0.000 0.030 1.000 2517.5

Total 0.000 0.006 12362.7

Q = 0.362, Pooled RD = 0.000 (0.012, 0.012), p-value = 0.957

Fig. 6. Forest plot of implant failure risk differences. Table 8

Data extraction from the six clinical trials included in the meta-analysis. Author(s) Follow-up

period

Implant system Position No. of patients

Connection type (No. of Implants)

No. of implants in the end

Marginal bone loss (Mean ± SD)

Survival rate (%) Peñarrocha-Diago

et al.[42]

1 year Osseous; Maxilla

Mandible

18 Ext conna

(69) 68; 0.38 ± 0.51 98.6

Inhex Int conn PSb

(72) 71 0.12 ± 0.17 98.6

Pieri et al.[43] 1 year Samo Smiler System, BioS Park

Maxilla 40 Int connc

(20) 19; 0.49 ± 0.25 100;

Int conn PS (20) 18 0.19 ± 0.17 94.7

Fickl et al.[21] 1 year Osseotite Certain Biomet 3i

Maxilla Mandible

36 Ext conn (14) 14; 1.00 ± 0.22 100;

Ext conn PS (75) 75 0.39 ± 0.07 100

Veis et al.[55] 2 years Osseotite, Biomet 3i Maxilla Mandible

n.r. Ext conn (193) 193; 0.88 ± 0.85 100;

Ext conn PS (89) 89 0.75 ± 0.55 100

Vigolo and Givani [58]

5 years 3i/ Implant Innovations

Maxilla Mandible

144 Ext conn (85) 85; 1.1 ± 0.3 100;

Ext conn PS (97) 97 0.6 ± 0.2 100

Crespi et al.[8] 2 years Seven Sweden & Martina;

Maxilla Mandible

45 Ext conn (34) 34; 0.78 ± 0.45 100

Ankylos Plus Int conn PS (30) 30 0.73 ± 0.52 100

a

Ext conn: External connection. b

PS: Platform switch. c

rate regardless of design. However, preventing bone loss enhances the long term survival of the implant and improves the appearance of the replaced tooth. For dental implant professionals, the results demonstrate the significance of relating the performance of de-signs as an aid in the selection of improved dental implant prod-ucts and techniques.

Given the newly discovered outcomes of clinical trial meta-analysis, the patent data set were further analyzed and separated into several groups depending on the types of implant connections. There were 14 external connection patents, 7 internal connection patents, and 3 platform switch patents included in the multi-features cluster. There were 3 external connection patents and 9 internal connections in the connections cluster. The average patent age of external connection patents is about 19 years, the average age of internal connection patents is about 13 years, and the average age of platform switch patents is the youngest (11 years). The number of external connection patent applications increased between 1989 and 1995 and decreased after 1995. The applications for internal connection patents have increased after the year 2000 with fewer platform switch patents being registered. The patent statistics indicate that the external and internal connec-tion techniques are relatively mature technologies when compared to platform switch techniques.

By mapping the matrix of assignees and the numbers patent applications in sub-technologies (platform switch, internal connec-tion platform match, and external connecconnec-tion platform match), the

research identifies companies that dominate the market due to their patents and innovative dental implant connections. The dom-inant assignees are Nobel Biocare, Biomet 3i, Zimmer, Astra Tech, and Straumann. The patents of these five assignees cover the majority of internal connectors used for patient treatment. How-ever, the internal connection patents of Zimmer have expired. The other four companies continue to file patents for internal con-nections. The external connection patents of Zimmer and Biomet 3i have expired indicating that there are few companies developing external connection implants. Nobel Biocare owns the patents for platform switch techniques. Companies interested in developing platform switch implant related products must fully understand the patents and products of Nobel Biocare to avoid patent infringe-ment and intellectual property concerns.

5. Conclusions

This study proposes a methodology combining patent analysis and clinical meta-analysis. The critical text of patent documents is extracted using knowledge discovery to build the ontology. The ontology is validated by a dental expert and the construction of a visual map of key terms expresses and inter-relates the diverse terminologies from patent documents. The results help researchers utilize the knowledge of dental implant connections for further research and development. Meta-analysis provides a summary of

Fig. 7. Forest plot of mean differences in marginal bone loss. Table 10

Mean difference data for marginal bone loss.

Study Treatment group Control group SD MDd

SE p-Value We W’f Na Mb SDc N M SD Peñarrocha-Diago et al.[42] 71 0.12 0.17 68 0.38 0.51 0.37 0.26 0.17 0.000 244.7 26.8 Pieri et al.[43] 19 0.2 0.17 18 0.51 0.24 0.20 0.31 0.37 0.000 215.8 26.4 Fickl et al.[21] 75 0.39 0.07 14 1 0.22 0.11 0.61 0.52 0.000 1034.9 29.3 Veis et al.[55] 89 0.75 0.55 193 0.88 0.85 0.77 0.13 0.13 0.187 103.2 23.3

Vigolo and Givani[58] 97 0.6 0.2 85 1.1 0.3 0.25 0.5 0.18 0.000 715.3 28.9

Crespi et al.[8] 30 0.73 0.52 34 0.78 0.49 0.50 0.05 0.25 0.692 62.7 20.4

Total 381 412 2376.6 155.1

Q = 60.07,s2

= 0.033 , Pooled MD = 0.33 (0.49, 0.17), p-value = 1  1011 a _{N: number of implants.}

b _{M: marginal bone loss value.} c _{SD: standard deviation.} d

MD: mean difference. e

W: weight in fixed-effect model. f

the results from clinical trials which in turn is used to study the effectiveness of the related patents and implant designs. The re-search uses dental implants as a case study. The results help researchers and market analysts understand the implant connector market trends and determine the future design directions. In addi-tion, the methodology developed in this research is broad enough to be applied to medical devices in general and can easily be ex-tended to pharmaceutical products. The ability to link successful clinical trials and their effectiveness to patent designs and innova-tions provides a new means to access patent value.

Acknowledgement

This research is partially supported by the National Science Council research grants to the authors’ affiliating universities in Taiwan.

References

[1]R. Adell, U. Lekholm, B. Rockler, P.I. Brånemark, A 15-year study of osseointegrated implants in the treatment of the edentulous jaw, Int. J. Oral Surg. 10 (6) (1981) 387–416.

[2]T. Albrektsson, J. Gottlow, L. Meirelles, P.O. Ostman, A. Rocci, L. Sennerby, Survival of nobel direct implants: an analysis of 550 consecutively placed implants at 18 different clinical centers, Clin. Implant. Dent. Relat. Res. 9 (2) (2007) 65–70 (June).

[3]M. Bateli, W. Att, J.R. Strub, Implant neck configurations for preservation of marginal bone level: a systematic review, Int. J. Oral Max. Impl. 26 (2) (2011) 290–303.

[4]R. Brandow, K. Mitze, L.F. Rau, Automatic condensation of electronic publications by sentence selection, Inf. Process. Manage. 31 (5) (1995) 675– 685.

[5] Ceramic Industry. (2012) ‘‘Dental Implants and Prosthetics Market Continues Growth.’’ Available at: <http://www.ceramicindustry.com/articles/92515-dental-implants-and-prosthetics-market-continues-growth> (15.09.12). [6]R. Chen, Design patent map visualization display, Expert Syst. Appl. 36 (10)

(2009) 12362–12374.

[7]L.F. Cooper, M.S. Scurria, L.A. Lang, A.D. Guckes, J.D. Moriarty, D.A. Felton, Treatment of edentulism using Astra Tech implants and ball abutments to retain mandibular overdentures, Int. J. Oral Max. Impl. 14 (5) (1999) 646–653. [8]R. Crespi, P. Capparè, E. Gherlone, Radiographic evaluation of marginal bone levels around platform-switched and non-platform-switched implants used in an immediate loading protocol, Int. J. Oral Max. Impl. 24 (5) (2009) 920–926. [9]M. Degidi, D. Nardi, A. Piattelli, Prospective study with a 2-year follow-up on immediate implant loading in the edentulous mandible with a definitive restoration using intra-oral welding, Clin. Oral Implant Res. 21 (4) (2010) 379– 385.

[10]R. DerSimonian, N. Laird, Meta-analysis in clinical trials, Control. Clin. Trials 7 (3) (1986) 177–188.

[11]V. Devedzic, Knowledge discovery and data mining in databases, Handbook Softw. Eng. Knowl. Eng. 1 (2001) 615–637.

[12]M. Donati, V. La Scala, M. Billi, B. Di Dino, P. Torrisi, T. Berglundh, Immediate functional loading of implants in single tooth replacement: a prospective clinical multicenter study, Clin. Oral Implant Res. 19 (8) (2008) 740–748. [13]K. Döring, E. Eisenmann, M. Stiller, Functional and esthetic considerations for

single-tooth Ankylos implant-crowns: 8 years of clinical performance, J. Oral Implantol. 30 (3) (2004) 198–209.

[14]C.J. Drago, A clinical study of the efficacy of gold-tite square abutment screws in cement-retained implant restorations, Int. J. Oral Max. Impl. 18 (2) (2003) 273–278.

[15]Y. Ding, S. Foo, Ontology research and development. Part 1 – a review of ontology generation, J. Inform. Sci. 28 (2) (2002) 123–136.

[16]H. Ernst, N. Omland, The Patent Asset Index – A new approach to benchmark patent portfolios, World Patent Inf. 33 (1) (2011) 34–41.

[17] FDA (2013). Guidance for Industry and FDA Staff – Class II Special Controls Guidance Document: Root-form Endosseous Dental Implants and Endosseous Dental Abutments – issued on May 12, 2004. Available at: <http:// www.fda.gov/MedicalDevices/DeviceRegulationandGuidance/

GuidanceDocuments/ucm072424.htm> (10.12.13).

[18]W.J. Frawley, G. Piatetsky-Shapiro, C.J. Matheus, Knowledge discovery in databases: an overview, AI Magazine 13 (3) (1992) 57–70.

[19]U. Fayyad, G. Piatetsky-Shapiro, P. Smyth, From data mining to knowledge discovery in databases, AI Magazine 17 (3) (1996) 37–54.

[20]R. Feldman, I. Dagan, Knowledge discovery in textual databases (KDT), in: Proceedings of the First International Conference on Knowledge Discovery and Data Mining (KDD-95), Montreal, Canada, August 20–21, AAAI Press, 1995, pp. 112–117.

[21]S. Fickl, O. Zuhr, J.M. Stein, M.B. Hürzeler, Peri-implant bone level around implants with platform-switched abutments, Int. J. Oral Max. Impl. 25 (3) (2010) 577–581.

[22] Global Information, Inc., (2013) Medical Equipment Annual Deals Analysis: M&A and Investments Trends – 2012. Available at: <http:// www.giiresearch.com/report/gd233652-medical-equipment-annual-deals-analysis-2012.html> (16.06.13).

[23]S. Green, Systematic reviews and meta-analysis, Singapore Med. J. 46 (6) (2005) 270–273.

[24]T. Gruber, Toward principles for the design of ontologies used for knowledge sharing, Int. J. Hum. Comput. Stud. 43 (5–6) (1993) 907–928.

[25] M. Grüninger, M.S. Fox, Methodology for the Design and Evaluation of Ontologies, in: International Joint Conference on Artificial Intelligence (IJCAI95), Workshop on Basic Ontological Issues in Knowledge Sharing, 1995. [26]N. Guarino, Understanding, building and using ontologies, Int. J. Hum. Comput.

Stud. Arch. 46 (2–3) (1997) 293–310.

[27]L.V. Hedges, J.L. Vevea, Fixed- and random-effects models in meta-analysis, Psychol. Methods 3 (4) (1998) 486–504.

[28]M.N. Huhns, L.M. Stephens, Personal ontologies, IEEE Internet Comput. 3 (5) (1999) 85–87.

[29]F.-C. Hsu, A.J.C. Trappey, C.V. Trappey, J.-L. Hou, S.-J. Liu, Technology and knowledge document cluster analysis for enterprise R&D strategic planning, Int. J. Technol. Manage. 36 (4) (2006) 336–353.

[30]K.S. Khan, R. Kunz, J. Kleijnen, G. Antes, Five steps to conducting a systematic review, J. R. Soc. Med. 96 (3) (2003) 118–121.

[31]S. Lee, B. Yoon, Y. Park, An approach to discovering new technology opportunities: keyword-based patent map approach, Technovation 29 (6–7) (2009) 481–497.

[32] T.-H. Lin, Using Data Mining Technique to Analyze LED Patent Litigation and Licensing Strategies, M.S. thesis, Dept. Industrial Management, National Taipei University of Technology, Taiwan, 2012.

[33]T. Linkevicius, P. Apse, S. Grybauskas, A. Puisys, Influence of thin mucosal tissues on crestal bone stability around implants with platform switching: a 1-year pilot study, J. Oral Maxillofac. Surg. 68 (9) (2010) 2272–2277. [34]A.-H.A. Liou, Y.-T. Wang, K.-T. Lu, Collaborated construction and sharing of

domain knowledge using ontology, J. Inform. Manage. Concepts Syst. Appl. 9 (2) (2007) 109–126.

[35]A.G. Lalkhen, A. McCluskey, Statistics V: introduction to clinical trials and systematic reviews, Cont. Educ. Anesth. Crit. Care Pain 8 (4) (2008) 143–146. [36]C. Mangano, F. Mangano, A. Piattelli, G. Iezzi, A. Mangano, L. La Colla, Prospective clinical evaluation of 1920 Morse taper connection implants: results after 4 years of functional loading, Clin. Oral Implant Res. 20 (3) (2009) 254–261.

[37]C. Mangano, F. Mangano, A. Piattelli, G. Iezzi, A. Mangano, L. La Colla, Prospective clinical evaluation of 307 single-tooth Morse taper-connection implants: a multicenter study, Int. J. Oral Max. Impl. 25 (2) (2010) 394–400. [38]C. Mangano, F. Mangano, J.A. Shibli, M. Ricci, R.L. Sammons, M. Figliuzzi, Morse

taper connection implants supporting ‘‘planned’’ maxillary and mandibular bar-retained overdentures: a 5-year prospective multicenter study, Clin. Oral Implant Res. 22 (10) (2011) 1117–1124.

[39]P. Narang, H. Gupta, A. Arora, A. Bhandari, Biomechanics of implant abutment connection: a review, Indian J. Stomatol. 2 (2) (2011) 113–116.

[40]F. Narin, Patent bibliometrics, Scientometrics 30 (1) (1994) 147–155. [41]M.R. Norton, Biologic and mechanical stability of single-tooth implants: 4- to

7-year follow-up, Clin. Impl. Dent. Related Res. 3 (4) (2001) 214–220; P. Papaspyridakos, C.-J. Chen, M. Singh, H.-P. Weber, G.O. Gallucci, Success criteria in implant dentistry: a systematic review, J. Dent. Res. 91 (3) (2012) 242–248.

[42] M.A. Peñarrocha-Diago, A.J. Flichy-Fernández, R. Alonso-González, D. Peñarrocha-Oltra, J. Balaguer-Martínez, M. Peñarrocha-Diago, Influence of implant neck design and implant-abutment connection type on peri-implant health. Radiological study, Clin. Oral Implant Res. (2012),http://dx.doi.org/ 10.1111/j.1600-0501.2012.02562.x.

[43]F. Pieri, N.N. Aldini, C. Marchetti, G. Corinaldesi, Influence of implant-abutment interface design on bone and soft tissue levels around immediately placed and restored single-tooth implants: a randomized controlled clinical trial, Int. J. Oral Max. Impl. 26 (1) (2011) 169–178.

[44]G. Salton, C. Buckley, Term-weighting approaches in automatic text retrieval, Inf. Process. Manage. 24 (5) (1988) 513–523.

[45]G. Salton, C.S. Yang, On the specification of term values in automatic indexing, J. Doc. 29 (4) (1973) 351–372.

[46]M.S. Schwarz, Mechanical complications of dental implants, Clin. Oral Implant Res. 11 (Suppl 1) (2000) 156–158.

[47]B. Shneiderman, The next generation of graphical user interfaces: information visualization and better window management, Displays 17 (34) (1997) 125– 129.

[48]A.H. Tan, Text mining: the state of the art and the challenges, Proc. PAKDD Workshop Knowl. Discov. Adv. Databases (1999) 71–76.

[49]A.J.C. Trappey, C.V. Trappey, C.Y. Wu, Automatic patent document summarization for collaborative knowledge systems and services, J. Syst. Sci. Syst. Eng. 18 (1) (2009) 71–94.

[50]C.V. Trappey, A.J.C. Trappey, C.-Y. Wu, Clustering patents using non-exhaustive overlaps, J. Syst. Sci. Syst. Eng. 19 (2) (2010) 162–181.

[51]A.J.C. Trappey, C.V. Trappey, C.-Y. Wu, C.-W. Lin, A patent quality analysis for innovative technology and product development, Adv. Eng. Inform. 26 (1) (2012) 26–34.

[52]A.J.C. Trappey, C.V. Trappey, T.-A. Chiang, Y.-H. Huang, Ontology-based neural network for patent knowledge management in design collaboration, Int. J. Prod. Res. 51 (7) (2013) 1992–2005.

[53]C.V. Trappey, J.-M. Wang, S. Hoang, A.J.C. Trappey, Constructing a dental implant ontology for domain specific clustering and life span analysis, Adv. Eng. Inform. 27 (2013) 346–387.

[54]Y. Tseng, Y. Wang, Y. Lin, C. Lin, D. Juang, Patent surrogate extraction and evaluation in the context of patent mapping, J. Inform. Sci. 33 (6) (2007) 718– 736.

[55]A. Veis, N. Parissis, A. Tsirlis, C. Papadeli, G. Marinis, A. Zogakis, Evaluation of peri-implant marginal bone loss using modified abutment connections at various crestal level placements, Int. J. Periodont. Rest. Dent. 30 (6) (2010) 609–617.

[56]S. Weiss, N. Indurkhya, T. Zhang, F. Damerau, Text Mining Predictive Methods for Analyzing Unstructured Information, Springer, Berlin, 2005.

[57] Wheeljet.com, IPDSS – Intellectual Property Defense Support System, 2013. Available at: <http://www.wheeljet.com.tw/edu/> (6.12.13).

[58]P. Vigolo, A. Givani, Platform-switched restorations on wide-diameter implants: a 5-year clinical prospective study, Int. J. Oral Max. Impl. 24 (1) (2009) 103–109.

[59] World Health Organization (2012) ‘‘Oral health.’’ Available at: <http:// www.who.int/mediacentre/factsheets/fs318/en/> (1.12.12).