Evaluate Enterprises’ Technological Innovation Levels Based on the Technical Knowledge Flow Network a Case Study of the Medical Equipment Industry

Abstract

Evaluating enterprises’ technological innovation levels is crucial for assessing their technological capabilities and formulating technology strategy plans. Grounded in the knowledge flow theory, this study integrates patent metrics, social network analysis, and knowledge characteristics to develop a comprehensive index system for measuring enterprises’ technological innovation levels and analyzing the competitive landscape. It focuses on evaluating the technological innovation levels among enterprises in the medical equipment sector from 2012 to 2024. The findings indicate that, in the medical equipment sector, KONINKLIJKE PHILIPS ELECTRONICS N.V and FUJIFILM CORPORATION are advanced in terms of technological innovation. Moreover, the United States, Japan, and European countries constitute the core competitive landscape in this field, with the United States leading in high-tech innovations. As a relative newcomer country, China demonstrates varying degrees of technological innovation across its enterprises. Enhancing cooperation and establishing platforms for technological innovation collaboration represent effective strategies to strengthen the technological innovation capabilities of these countries.

Keywords

technological innovation knowledge flow network medical equipment competitive situation evaluation of technological innovation

Introduction

The medical equipment industry is a critical technological domain for national healthcare advancement (X. Wu et al., 2023), and medical policies worldwide have fostered numerous device enterprises and stimulated technological innovation. Recent studies have examined core technologies in this sector (Cai et al., 2024), analyzed global research trends (Bai et al., 2024), and investigated innovation risks and development strategies (Z. Zhang & Rao, 2021). In the knowledge economy era, knowledge reserves and innovation resources are pivotal drivers of economic growth (Shukla, 2021), particularly in medical equipment technology. They enhance enterprises’ innovation capacity, economic strength, and sustainable development, while shaping future knowledge, technology, and market competitiveness (Haq & Davies, 2023).

Technological innovation level reflects an enterprise’s innovative output and its contribution to advancing other enterprises. Systematic evaluation of technological innovation level facilitates clarify technological standing, optimize resource allocation, and strengthen competitive advantage (Lv, 2023), enabling analysis of firm positioning, competitive landscape, and performance benchmarking against industry leaders. Research on evaluating innovation levels and competitive positioning focuses on indicator selection and methodological innovation. Indicators typically include innovation input and output (Yi et al., 2021). Methods range from traditional statistics to network analysis, improving evaluation accuracy (S. Xu et al., 2025; Q. Zhang et al., 2023). Data visualization further enables intuitive depiction of technological competition (J. Li et al., 2024; R. Z. Zhang et al., 2025).

However, the existing research on the medical equipment industry lacks sufficient analysis of the technological innovation level and landscape, which is essential for enterprises to identify strengths and weaknesses, enhance their innovation capabilities, and inform latecomer countries’ innovation strategies. Moreover, current evaluation methods overlook the knowledge elements embedded in technological innovation. Incorporating enterprise knowledge and inter-firm knowledge flows into the assessment framework can improve both innovation capability evaluation and competitive landscape analysis. This study proposes a novel framework for constructing a technological innovation measurement indicator system from a knowledge flow perspective, which provides a foundation for evaluating innovation levels and fostering high-quality development by Emphasizing the role of knowledge generated during innovation.

Literature Review

Existing research on enterprise technological innovation evaluation primarily focuses on input and output dimensions. Input indicators include R&D investment and the number of personnel (Ma et al., 2023; Cummings & Knott, 2018). However, the inherently unpredictable nature of innovation outcomes limits the validity of using R&D investment as a reliable metric. Output indicators cover patent applications and grants, citation counts (Bradley et al., 2017), new product development (K. Lee & Yoon, 2015), and economic benefits (de Melo et al., 2021). They represent tangible results of innovation activities and serve as a direct performance indicator. Furthermore, new product metrics such as categories, sales volume, and customer satisfaction, though relevant, are difficult to access due to their link to core corporate resources. Patent data, as an open and structured dataset from enterprises with relatively comprehensive and standardized information, offers a suitable alternative for analyzing technological innovation levels (Y. N. Lee, 2015; Ponta et al., 2021). Consequently, patent-based evaluation systems are more widely adopted (Yu et al., 2020).

Several methodologies are employed to measure technological innovation capability, including principal component analysis (Yin & Su, 2021), entropy-TOPSIS (Yuan & Song, 2022), IPSO-ATT-MSCNN (Q. Zhang et al., 2023), data envelopment analysis, meta-frontier, and Tobit regression (Q. Zhang et al., 2023). System complexity demands more sophisticated assessment approaches. Luo and Lin (2022) propose a hybrid evaluation model based on consistent fuzzy preference relations (CFPR) and triangular intuitionistic fuzzy numbers. Velazquez-Cazares et al. (2021) underscores the application of expert models, sufficiency coefficients, and theory of forgetting effects. Pei et al. (2024) develop a multi-attribute decision-making approach that utilizes interval type-2 trapezoidal fuzzy information measures and is propelled by trigonometric functions.

With advances in social network theory, patent and literature analyses reveal relational structures among innovation actors (S. Xu et al., 2025), including research collaboration (J. Li et al., 2024), citation linkages (H. Y. Xu et al., 2020), and technological similarity (Qiu & Wang, 2023). Network centrality captures an entity’s prominence across these relational contexts. Zhao et al. (2019) construct co-citation and co-project networks to demonstrate influence diffusion among research teams. An et al. (2023) integrate PageRank and enhanced HITS algorithms with patent citation data to assess technological and enterprise innovation impact. Synthesizing these approaches offers new pathways for measuring innovation capability. K. K. Lai et al. (2020) combine patent indicators, network centrality, and main path analysis to help enterprises identify their competitive positioning.

Integrating innovation capability metrics with competitive landscape analysis is critical for decision-makers. Chang (2012) uses patent counts and citations to model technology positioning, advancing quantitative research on technology distribution. Wang et al. (2022) extend this by incorporating market domains, R&D capability, and composite patent indicators. As technological fields become more specialized, scholars adopt advanced methods such as Recurrent Graph Neural Networks (RGNN) and Latent Dirichlet Allocation (LDA) topic models to analyze technology ownership patterns (J. Li et al., 2024; R. Z. Zhang et al., 2025). These approaches are applied to high-tech sectors including digital twin, negative carbon (J. Li et al., 2024), electronic design automation (Wang et al., 2022), and autonomous driving technologies (R. Z. Zhang et al., 2025).

Current evaluation of technological innovation capability primarily emphasizes output quantification, while landscape analyses focus predominantly on technological dominance. The developmental role of knowledge generated through innovation remains underappreciated. Yet knowledge creation underpins technological advancement and is indispensable to innovation. This study foregrounds knowledge flow and contribution, using enterprises’ core technological knowledge as the primary metric for evaluating innovation levels and mapping competitive landscapes.

Methods

Based on established indicators from the literature, this study develops a three-dimensional framework for evaluating enterprises’ technological innovation level: technological patent holdings, contribution of technological innovation, and characteri- stics of the knowledge base. Patent holdings capture quantified technological output, including authorized invention patents, patent families, and citation counts (Geisler, 2002; Guerrero-Bote et al., 2021). The knowledge base reflects the value generated by an enterprise’s knowledge stock, serving as the foundation for innovation. Innovation outcomes depend on the technical scope and research depth embedded in these achievements, signaling innovation capability (Zhou et al., 2024). Contribution to external innovation represents the translation of knowledge into outputs that are absorbed by other enterprises, shaping technological trajectories. Enterprises producing high-quality innovations occupy central positions, drive technological progress, and exert broad influence (Kamasak & Bulutlar, 2010; Whittington et al., 2009). From a systems perspective, an enterprise’s contribution integrates with and extends beyond its knowledge base, reflecting innovation level and development quality (Cricchio et al., 2025). Building on this framework, we propose a method for measuring innovation levels and analyzing competitive patterns. Figure 1 illustrates the research framework.

Figure 1.

Research framework.

Selection Measurement Indicators

Technology Patentometrics

Patents are direct outputs of technological innovation and the quantity and quality of which reflect an enterprise’s innovation capability. An enterprise’s patent portfolio includes granted patents, citation frequency, and patent family diversity. Granted invention patents indicate technological advancement. The frequency of patent citations indicates the extent to which technology is acknowledged and utilized by other enterprises for further technological innovations, manifesting the quality and impact of technological innovation. Patent families capture the extent to which technology serves as a barrier across markets, embodying strategic influence.

Technological Innovation Contributions

Technological innovation contribution reflects the extent to which an enterprise’s innovation output provides knowledge that advances other enterprises’ technological development. This contribution is evaluated using a knowledge flow network derived from patent citations. The network is constructed as follows:

Step 1: Patent-producing and patent-citing enterprises are treated as nodes, with citation relationships as directed edges weighted by citation counts. Edge direction indicates knowledge flow. For example, if Enterprise B cites A three times and A cites B once, the edge weights from A to B and B to A are 3 and 1, respectively (Figure 2); Step 2: Examiner citations are excluded, as they do not represent inter-enterprise knowledge flow; Step 3: Each patent family is treated as a single invention and all citations within the family are consolidated into one knowledge flow event, regardless of country; Step 4: Self-citations (where the citing and cited enterprises are the same) are removed.

Figure 2.

The directed knowledge flow and its weight between enterprise A and B.

From the knowledge flow network, we selected four indicators to evaluate enterprises’ technological innovation contributions: indegree centrality, outdegree centrality, betweenness centrality, and closeness centrality.

Indegree centrality measures the extent and diversity of knowledge inflow to a node. High indegree nodes form numerous knowledge associations, acquire greater knowledge benefits, and enhance their knowledge reserves and quality. In a patent citation network, enterprises that absorb technical knowledge tend to generate new innovations, reflecting strong knowledge integration and transformation capabilities.

Outdegree centrality reflects a node’s influence in the knowledge flow network, indicating how often its knowledge is cited or used by others. Greater absorption and integration of an enterprise’s innovation outcomes by other nodes signals higher knowledge quality and a more prominent role in technological development.

Betweenness centrality captures a node’s role as a bridge in the network. Nodes with high betweenness access diverse critical knowledge and facilitate knowledge transfer among other nodes, contributing to overall technological innovation capacity.

Closeness centrality represents the average distance from a node to all others. Knowledge loss increases with transmission distance; thus, shorter average distances enable greater knowledge integrity to be conveyed to other nodes.

Table 1 provides detailed definitions and mathematical formulas for each indicator.

Table 1.

Indicators and Connotation of Technological Innovation Contributions.

Centrality indicators	Mathematical formulas		Indicators connotation
Indegree centrality	$C T_{i, in} = \frac{\sum_{j = 1, j \neq i}^{n - 1} x_{ij}}{n - 1}$	(1)	The ratio of external knowledge cited by enterprise i to the aggregate number of enterprises within the technological domain
Outdegree centrality	$C T_{i, out} = \frac{\sum_{j = 1, j \neq i}^{n - 1} y_{ij}}{n - 1}$	(2)	The ratio of enterprises citing enterprise i’s knowledge to all enterprises in the field.
Betweenness centrality	$C B_{i} = \sum \frac{S_{i}}{S}$	(3)	The proportion of enterprise i situated on the shortest path between the other two enterprises within the knowledge flow network
Closeness centrality	$C G_{i} = \frac{1}{\sum_{j = 1, j \neq i}^{n - 1} d (i, j)}$	(4)	The proximity of the connections between enterprise i and other enterprises.

In formulas (1)–(4), $x_{ij}$ and $y_{ij}$ indicate whether a directed edge exists from node i to j and from j to i, respectively (1 if present, 0 otherwise). n is the total number of nodes. $S_{i}$ denotes the number of shortest paths passing through node i, and S is the total number of shortest paths between all node pairs. Let $d (i, j)$ represent the shortest path length between nodes i and j.

Additionally, a node’s knowledge absorption capacity depends not only on the volume of its inputs but also on their quality. Incoming connections from high-quality nodes enhance a node’s importance through iterative network effects. However, indegree centrality captures only the quantity of incoming links. To better reflect node significance, the quality of upstream nodes must be considered. The PageRank algorithm addresses this by incorporating source node quality in its iterative computation. A higher PageRank value indicates greater node importance. The algorithm is formalized as follows:

P R_{i} = \frac{1 - d}{N} + d \sum_{j \in M_{j}} \frac{PR (j)}{L (j)},

(5)

where $M_{i}$ is the set of nodes linking to node i, $L (j)$ is the number of outgoing links from node j, N is the total number of nodes, and d is the damping factor (typically 0.85).

Knowledge Base Characteristics

A strong knowledge base enables enterprises to identify and seize new opportunities, thereby enhancing innovation performance. Prior research typically characterizes knowledge bases along two dimensions: knowledge breadth and knowledge depth, reflecting horizontal and vertical development (Mannucci & Yong, 2018). We adopt this framework to assess enterprises’ knowledge base as part of their technological innovation capability.

Knowledge breadth refers to the diversity and heterogeneity of an enterprise’s expertise across technological domains, capturing the range of its knowledge content and technical proficiency (de Luca & Atuahene-Gima, 2007; J. Wu & Shanley, 2009). Following Z. B. Li et al. (2021), we measure knowledge breadth by counting the first four IPC symbols in the enterprise’s patent applications:

Bo K_{i} = \sum Coun t_{i},

(6)

Knowledge depth captures the extent of an enterprise’s expertise within a specific domain, reflecting its proficiency with complex and advanced knowledge content in particular technologies or applications (de Luca & Atuahene-Gima, 2007; Jin et al., 2015). Following Liu et al. (2014), we measure knowledge depth as follows:

First, compute the ratio of patents held by enterprise i in technology field j (IPC subclass) to its total patents:

pro p_{ij} = \frac{N_{ij}}{\sum_{j = 1}^{n_{i}} N_{ij}},

(7)

here, $N_{ij}$ is the number of patent applications by enterprise i in technology subclass j during the measurement year, and $n_{i}$ is the total number of distinct IPC subclasses among those applications. Knowledge depth is then calculated as:

Ko S_{i} = \frac{σ_{i}}{μ_{i}},

(8)

where $μ_{i}$ is the mean proportion of patents across all technical fields for enterprise i, and $σ_{i}$ is the corresponding standard deviation.

Indicators and Weights

Following the framework in Section 3.1, we develop an indicator system (Table 2) that synthesizes technological innovation output metrics to evaluate technological innovation level of enterprises.

Table 2.

Indicators of Enterprise’s Technological Innovation Level Evaluation.

Target level	Criterion layer	Guideline layer
Technological innovation level of enterprises (A)	Technology patentometrics (B₁)	Number of patents granted (C₁)
		Number of patent citations (C₂)
		Number of patents in the same family (C₃)
	Technological innovation contributions (B₂)	Indegree centrality (C₄)
		Outdegree centrality (C₅)
		Betweenness centrality (C₆)
		Closeness centrality (C₇)
	Knowledge base characteristics (B₃)	Width of knowledge (C₈)
	Knowledge base characteristics (B₃)	Depth of knowledge (C₉)

We determine indicator weights using the analytic hierarchy process (AHP) and entropy weight method. AHP integrates qualitative and quantitative methodologies, with initial weights derived from expert ratings to incorporate domain knowledge. The entropy weight method emphasizes differentiation among indicator values, enabling quantitative weighting based on patent data. Integrate the weights obtained from these two methods according to formula (9):

ω = α ω_{1} + (1 - α) ω_{2},

(9)

where ω₁ and ω₂ represents the indicator weight determined by AHP and entropy weight method, respectively. Thus, the enterprises’ technological innovation level is calculated as follows:

T R_{i} = \sum_{i = 1}^{n} ω_{i} x_{i},

(10)

where ω_i and x_i represent the composite weight and value of the ith enterprise item, respectively.

Evaluation of Enterprises’ Technological Innovation Levels in Medical Equipment Industry

Data Collection and Processing

Patent Data Collection

Driven by global medical policies, the medical equipment industry has grown rapidly, fostering numerous enterprises and intensifying technological innovation. Yet, systematic analysis of technological competition in this sector remains limited. The industry integrates critical technologies—including chips, AI (W. Lai et al., 2022), materials science (Su et al., 2023), image recognition (Ibragimov & Mello-Thoms, 2024), big data, and neural networks (Mahmood et al., 2024; Rehman et al., 2024)—all vital to national development and public health. Its growth not only propels upstream high-tech sectors but also stimulates innovation in cutting-edge technologies. Hence, examining enterprise technological innovation levels in this domain is essential.

We used the Lens database, a global platform covering approximately 95% of publicly accessible patent literature. IPC classifications for the medical device industry were identified using United States Patent Classification (USPC)—North American Industry Classification System (NAICS) Consistent Correspondence Table and the USPC—International Patent Classification (IPC) Reverse Consistent Correspondence Table. Patents under IPC codes A61B, A61C, A61F, A61M, G02C, and 11 others were retrieved for 2012 to 2024, yielding 63,404 records. Patent numbers, family identifiers, priority numbers, and application dates were extracted. After sorting by patent family, we retained the earliest application within each family using a Python script, removing duplicates. This reduced the dataset to 41,701 patents, stored as Database-1.

Additional Patents’ References

We queried the Lens database using all patent numbers in Database-1, covering the period 2012 to 2024. For each patent, we retrieved its references and retained only those also present in Database-1. This process established the citation relationships among all patents in the dataset.

Precise Enterprise’s Name

Since not all patents record assignee information and the applicant represents the knowledge-producing entity, we substitute the applicant for the assignee when assignee data is missing. Patents lacking both fields are excluded. Extracting owner names from patent records presents challenges including name translation variants, country-of-origin inconsistencies, and spelling errors, causing the same entity to appear under multiple names. To address this, we apply French et al.’s (2000) method, using a Jaccard coefficient-based string similarity algorithm with character trigrams. Name pairs with similarity exceeding 0.75 are manually verified and consolidated, producing a standardized enterprise name list. This process identifies 4,166 technological innovation enterprises (including universities).

Indicator Value

Value of Technology Patentometrics Indicators

This study discusses the number of patents held by technological innovation enterprises, with the results presented in Table 3.

Table 3.

Value of Technology Patentometrics (Part).

Enterprises	C₁	C₂	C₃
GENERAL ELECTRIC COMPANY	7,075	740,643	63,627
KONINKLIJKE PHILIPS ELECTRONICS N.V	4,976	330,463	78,681
FUJIFILM CORP	4,113	181,314	50,196
TOSHIBA MEDICAL SYS CORP	4,039	116,187	40,902
HITACHI LTD	2,521	378,319	44,325
RAKUTEN GROUP INC	1	7	16
SONOVOL INC	1	11	8
SHENZHEN KEYA MEDICAL TECH CO LTD	1	19	4
NOVASIGNAL CORP	1	37	9
MODUS MEDICAL DEVICES INC	1	3	7

Value of Technological Innovation Contribution Indicators

Using Python’s Network X library, we constructed a knowledge flow network following the method in Section 3.1.2 and computed indegree, outdegree, betweenness, and closeness centrality for each node, then normalized. Due to the large number of nodes, we only present the ego-network of the enterprise with the highest outdegree centrality—KONINKLIJKE PHILIPS ELECTRONICS N.V (Figure 3). The PageRank value was calculated using Python (Table 4) subsequently.

Figure 3.

KONINKLIJKE PHILIPS ELECTRONICS N.V’s ego-network.

Table 4.

PageRank Value (Part).

Enterprise name	PageRank value
GENERAL ELECTRIC COMPANY	0.0214
SIEMENS HEALTHCARE GMBH	0.0192
TOSHIBA MEDICAL SYSTEMS CORPORATION	0.0092
SAMSUNG ELECTRONICS CO. LTD	0.0059
KONINKLIJKE PHILIPS N.V	0.0044
FUJIFILM CORPORATION	0.0129
RADIMETRICS INC	0.0002
HYPERMED INC	0.0008
LEICA INSTR SINGAPORE PTE LTD	8.51E-05
BEMING	8.51E-05

To integrate indegree centrality with PageRank, we use the entropy weight method to determine their respective weights (Table 5). Table 6 presents each enterprise’s technological innovation contribution indicators.

Table 5.

Weight of Enterprises’ Indegree Centrality and PageRank Value.

Indicators	Indegree centrality	PageRank value
Weights	0.2705	0.7295

Table 6.

Indicators Value of Technological Innovation Contribution of Enterprises (Part).

Enterprises	C₄	C₅	C₆	C₇
KONINKLIJKE PHILIPS ELECTRONICS N.V	0.5623	1.0000	1.0000	1.0000
FUJIFILM CORPORATION	0.5969	0.9225	0.9088	0.9866
GE MEDICAL SYSTEMS GLOBAL TECHNOLOGY COMPANY LLC	0.5405	0.5262	0.3012	0.8804
MEDTRONIC INC	0.3143	0.4109	0.3500	0.8939
TOSHIBA CORP	0.3859	0.4858	0.2300	0.8366
EASTMAN KODAK COMPANY	0.3002	0.3413	0.3135	0.9376
SIEMENS HEALTHCARE GMBH	0.9115	0.1385	0.0394	0.7310
HITACHI LTD	0.4115	0.1743	0.0857	0.8250
NEC CORPORATION	0.1051	0.1875	0.1633	0.9049
SHIMADZU CORP	0.2804	0.2372	0.0877	0.8161
SHANGHAI UNITED IMAGING HEALTHCARE CO LTD	0.0510	0.1948	0.0979	0.6579
TSINGHUA UNIVERSITY	0.0553	0.0471	0.0122	0.7728
SHENZHEN XPECTVISION TECHNOLOGY CO. LTD	0.0124	0.0000	0.0000	0.7767
EMBARK ENTERPRISES INC	0.0009	0.0000	0.0000	0.0010

Value of Knowledge Base Characteristics Indicators

The characteristics of the enterprises’ knowledge base (Table 7) are calculated by formulas (6)–(8).

Table 7.

Characteristics of Enterprises’ Knowledge Base Characteristics (Part).

Enterprise	C₈	C₉
KONINKLIJKE PHILIPS ELECTRONICS N.V	0.0017	0.1728
FUJIFILM CORPORATION	0.0021	0.1674
GE MEDICAL SYSTEMS GLOBAL TECHNOLOGY COMPANY LLC	0.0018	0.1318
MEDTRONIC INC	0.0018	0.1193
TOSHIBA CORP	0.0013	0.1514
EASTMAN KODAK COMPANY	0.0042	0.0801
SIEMENS HEALTHCARE GMBH	0.0014	0.2208
HITACHI LTD	0.0021	0.1888
NEC CORPORATION	0.0058	0.0855
SHIMADZU CORP	0.0021	0.0855
SHANGHAI UNITED IMAGING HEALTHCARE CO LTD	0.0020	0.0677
TSINGHUA UNIVERSITY	0.0040	0.0481
SHENZHEN XPECTVISION TECHNOLOGY CO. LTD	0.0068	0.0142
EMBARK ENTERPRISES INC	0.0000	0.0000

Collinearity Diagnosis of Indicators

Variance inflation factors (VIF) were calculated to ensure each indicator contributes uniquely to the composite score (Table 8). All VIF values are below 5, with outdegree centrality approaching but still near this threshold, indicating no significant multicollinearity.

Table 8.

Collinearity Diagnosis of Indicators.

			Collinearity statistics
Model	t	Sig.	Tolerance	VIF
Constant	6.035	0.000
C₁	11,094.891	0.000	0.261	3.828
C₂	11,860.207	0.000	0.411	2.433
C₃	24,684.027	0.000	0.646	1.547
C₄	130,672.032	0.000	0.315	3.170
C₅	120,342.534	0.000	0.181	5.534
C₆	193,382.538	0.000	0.268	3.737
C₇	1,644,250.254	0.000	0.988	1.012
C₈	46,036.199	0.000	0.990	1.010
C₉	128,370.298	0.000	0.394	2.536

Evaluation Results of Enterprises’ Technological Innovation Levels

Indicator Weights

We distributed electronic questionnaires to 20 experts and senior practitioners in medical device innovation. Twelve valid responses were received, from which we constructed an importance evaluation matrix. AHP weights were derived after passing consistency checks (see Appendix A for judgment matrix and test results). Entropy weights were then calculated from measurement data. Following formula (9), we assigned 0.4 weight to AHP and 0.6 to the entropy method based on expert input. Table 9 presents the final weights.

Table 9.

Weights of Evaluation Indicators.

AHP weights					Entropy weights	Fusion weights
A	B₁	0.1634	C₁	0.0485	0.1926	0.1350
			C₂	0.0882	0.1915	0.1502
			C₃	0.0267	0.1947	0.1275
	B₂	0.5396	C₄	0.0762	0.0763	0.0762
			C₅	0.0762	0.1234	0.1045
			C₆	0.2455	0.1774	0.2046
			C₇	0.1417	0.0074	0.0611
	B₃	0.2970	C₈	0.2376	0.0087	0.0290
	B₃	0.2970	C₉	0.0594	0.0281	0.1119

Evaluation Results

Following formula (10), enterprises’ technological innovation level was calculated. Table 10 presents the partial results.

Table 10.

Evaluation Results of Enterprises’ Technological Innovation Levels (Part).

Rankings	Enterprises’ name	Score of technological innovation	Country
1	KONINKLIJKE PHILIPS ELECTRONICS N.V	0.4370	The Netherlands
2	FUJIFILM CORPORATION	0.4124	Japan
3	GE MEDICAL SYSTEMS GLOBAL TECHNOLOGY COMPANY LLC	0.2320	The United States
4	MEDTRONIC INC	0.2121	The United States
5	TOSHIBA CORP	0.1988	Japan
6	EASTMAN KODAK COMPANY	0.1903	The United States
7	SIEMENS HEALTHCARE GMBH	0.1726	Germany
8	HITACHI LTD	0.1454	Japan
9	NEC CORPORATION	0.1262	Japan
10	SHIMADZU CORP	0.1257	Japan
29	SHANGHAI UNITED IMAGING HEALTHCARE CO LTD	0.0927	China
110	TSINGHUA UNIVERSITY	0.0615	China
290	SHENZHEN XPECTVISION TECHNOLOGY CO. LTD	0.0497	China
		0.0004
4,116	EMBARK ENTERPRISES INC	0.0001	The United States

Based on the technological innovation evaluation results (Table 10), KONINKLIJKE PHILIPS ELECTRONICS N.V and FUJIFILM CORPORATION rank first and second, significantly outperforming others and playing a pivotal role in advancing medical equipment technology through extensive knowledge contributions. Among the top 20 firms, 10 are based in the United States, 5 in Japan, 2 in Germany, and one each in the Netherlands, Canada, and Ireland, highlighting the competitiveness of U.S. innovators. Only two universities (University of Washington and Stanford University) appear in the top 20, consistent with enterprises’ greater strength in applied innovation and patenting. The Gini coefficient of the evaluation scores is 0.24, indicating that innovation capability is not concentrated among a few leading firms. The Theil index is 0.1657 (<0.2), further confirming relatively equal innovation levels across enterprises.

To delineate the positioning of enterprises more precisely within global technological competition, we classify them into three tiers (high, medium and low) based on a 2:3:5 ratio. High-tier enterprises represent the technological forefront of their nations. Figure 4 presents the country-level distribution of these high-innovation enterprises.

Figure 4.

Proportion of high technological innovation level enterprises in each country.

The United States has the highest concentration of high-innovation enterprises, accounting for approximately 70% of the total. Japan, China, and Germany rank second, third, and fourth, respectively. However, they lag considerably behind the U.S. in number. Figure 5 presents a heat map comparing indicator distributions across enterprise tiers.

Figure 5.

Heatmap of indicators for technology leaders and technology followers.

Figure 5 reveals that, in technologically advanced countries, nearly all indicators are positive and possess relatively high values (1.15), establishing a virtuous cycle of “high patent output-strong network centrality-deep knowledge accumulation.” In contrast, the indicators of technologically backward countries exhibit a significant “unevenness.” The patent-related indicators and some network centrality indicators remain positive, suggesting that these countries still demonstrate positive performance in the “basic output” of innovation activities and network connections, presumably during the catch-up stage. The significant negative value of closeness centrality indicates that technologically backward countries have a marginal position in the innovation network. This situation makes it difficult for them to rapidly approach and access core knowledge sources, effecting low efficiency of knowledge flow. The concurrent negative values of “width of knowledge” and “depth of knowledge” suggest that their knowledge base is severely restricted, lacking accumulation of core technologies, and their knowledge structure is vulnerable.

To explore the proportion of countries with high, medium, and low levels of technological innovation, we conducted a country-wise analysis (Figure 6).

Figure 6.

Proportion of countries at each innovation level.

Figure 6 illustrates that the United States exhibits the highest proportion in high-tech innovation and the lowest in low-tech innovation. Countries such as Japan, Germany, the UK, and Israel show comparable enterprise shares across innovation levels. In China and South Korea, low-tech innovation enterprises outnumber those in high-tech sectors. Leading countries occupy both ends of the “smile curve,” aiming to preserve system integration advantages. For instance, over 90% of German medical device enterprises are SMEs, yet many dominate niche markets as “hidden champions.” Government policy centers on “connectivity”—linking enterprises with universities and hospitals to build self-reinforcing innovation networks. Latecomer countries initially prioritize manufacturing scale to meet basic medical needs. In China, about 80% of medical device enterprises generate annual sales below a few million yuan, focusing on low-value-added products. While effective in the “follower innovation” stage, this strategy risks path dependency.

To identify national strengths, we further disaggregated the data by country for each secondary indicator (Figures 7 –9).

Figure 7.

Comparison of technology patentometrics indicators among different countries.

Figure 8.

Comparison of technological innovation contributions indicators among different countries (the betweenness centrality indicator is placed on the secondary coordinate axis).

Figure 9.

Comparison of knowledge base characteristics indicators among different countries.

A national breakdown of key indicators reveals notably that, the U.S. dominates in technological innovation outcomes within the medical device sector, leading all other countries in patent counts, citations, and patent families. It also maintains a central position in the knowledge flow network and leads in both the breadth and depth of knowledge coverage. While Japan is a significant contributor, other major players—including China, South Korea, Germany, and Israel—collectively form a second tier.

To better understand enterprises’ differences, we analyzed enterprise performance using the indicators of technological innovation (partially shown in Figure 10). It is shown that high-level enterprises scored significantly on technological innovation contributions indicators.

Figure 10.

Comparison of indicators of technological innovation level (Part).

Ablation of Enterprises of Technological Innovation

The analysis reveals that many firms contribute minimally to medical device innovation, evidenced by sporadic patenting activity or by holding patents with incidental medical-device-related codes. To focus on significant contributors, we refined the sample. First, from the enterprise citation network, we removed edges with a weight below 10, thereby eliminating enterprises with fewer than 10 knowledge flow interactions. Isolated nodes were then excluded, yielding a core set of 1,778 enterprises with substantial contributions. For these enterprises, we retrieved their indicator values and calculated their technological innovation scores and rankings, which are presented in Table 11.

Table 11.

Technological Innovation Level of Enterprises After Ablation (Part).

Enterprise	Scores before ablation	Ranking before ablation	Scores after ablation	Ranking after ablation	Ranking changes
GENERAL ELECTRIC COMPANY	0.132466	14	0.494989	1	13
SIEMENS HEALTHCARE GMBH	0.172617	8	0.45111	2	6
FUJIFILM CORPORATION	0.412417	2	0.298053	3	1
SIEMENS MEDICAL SOLUTIONS USA INC	0.144311	11	0.292853	4	7
MEDTRONIC INC	0.212095	4	0.224829	5	1
KONINKLIJKE PHILIPS ELECTRONICS N.V	0.437045	1	0.223665	6	5
TOSHIBA MEDICAL SYSTEMS CORPORATION	0.12753	16	0.215944	7	9
GE MEDICAL SYSTEMS GLOBAL TECHNOLOGY COMPANY LLC	0.231966	3	0.213705	8	5
BOSTON SCIENTIFIC SCIMED INC	0.092425	43	0.21247	9	34
HITACHI LTD	0.145435	9	0.196903	10	1

Ablation caused minor ranking shifts among most top enterprises, except for BOSTON SCIENTIFIC SCIMED INC. Removing peripheral nodes weakened certain firms’ knowledge contributions, lowering their rankings. While post-ablation results better highlight core knowledge interactions and innovation contributions, this method overlooks firms with weaker capabilities or those focused on narrow technical niches.

Robustness Analysis

Discussion on the Proportion of Weighted Fusion

Based on expert input, AHP and entropy weights were combined at a 0.6:0.4 ratio. To assess robustness, we conducted a sensitivity analysis by varying α (Formula 9) across 0, .2, .4, .5, .6, .8, and 1. For each value, we computed and ranked enterprise innovation scores. Using α = .5 as the benchmark, we calculated Pearson correlation coefficients to evaluate ranking stability across weight combinations (Table 12).

Table 12.

Correlation of Evaluation and Ranking of Enterprises’ Technological Innovation Levels Under Different Weight Proportions (For Robustness Analysis).

Enterprises	α = 0	α = .2	α = .4	α = .5	α = .6	α = .8	α = 1.0	Min.	Max.
PHOTOCURE ASA CO LTD	12,268	12,543	12,647	12,729	12,737	12,740	12,741	12,268	12,741
LIGHT SCIENCES CORP	11,871	11,818	11,765	11,740	11,731	11,701	11,676	11,676	11,871
ORALUM LLC	12,908	12,876	12,844	12,830	12,829	12,824	12,816	12,816	12,908
CRAFT DENTAL CO LTD	10,456	10,605	10,613	10,607	10,604	10,605	10,603	10,456	10,613
IMPLANTDENT CO LTD	10,886	10,667	10,637	10,632	10,628	10,618	10,616	10,616	10,886
BEIJING KANG KE DA TECH CO LTD	10,743	10,602	10,585	10,580	10,575	10,571	10,568	10,568	10,743
CCIC SOUTHERN ELECTRONIC PRODUCT TESTING SHENZHEN CO LTD	8,050	8,420	8,430	8,427	8,423	8,423	8,425	8,050	8,430
JILIN PROVINCE SAFETY INSPECTION AND TESTING CO LTD	13,237	13,234	13,234	13,234	13,234	13,234	13,234	13,234	13,237
INST OF RADIATION MEDICINE SHANDONG ACAD OF MEDICAL SCIENCES	8,854	8,381	8,293	8,265	8,247	8,237	8,219	8,219	8,854
NANJING GUOKE MEDICAL TECH DEVELOPMENT CO LTD	4,182	4,279	4,280	4,259	4,258	4,269	4,268	4,182	4,280
SHENZHEN YUANMENG PREC TECHNOLOGY RES INST	9,718	9,345	9,287	9,272	9,263	9,249	9,241	12,268	12,741
Pearson’s correlation coefficient	.9507	.9960	.9997	1.0000	.9999	.9995	.9991

Table 12 shows that across different AHP-entropy weight ratios, the correlation coefficients for enterprise innovation rankings consistently exceed .95, demonstrating strong stability.

Robustness Tests and Uncertainty Analysis

To assess result robustness, we performed consistency tests using varying time windows, weighting schemes, and normalization methods. Table 13 shows that rankings of technological innovation levels remain highly stable across different evaluation methods.

Table 13.

Rank Consistency Test (For Robustness Analysis).

Evaluation methodology	Mean value of Kendall’s τ	Standard deviation of Kendall’s τ	95% CI lower limit	95% CI upper limit	Consistency categories
Different time windows
2020–2024 vs. 2012–2024	.5240	.0318	0.4597	0.5832	Moderate Consistency
Different weighting schemes
AHP vs. AHP+ Entropy weight	.9809	.0006	0.9798	0.9820	Very Strong Consistency
Entropy weight vs. AHP+ Entropy weight	.8770	.0026	0.8719	0.8820	Very Strong Consistency
Different normalization methods
z-score vs. min-max	.6616	.0045	0.6529	0.6703	Strong Consistency

Method Comparison and Benchmark Testing

To demonstrate the advantages of our method, we compared it with two alternatives: ranking based solely on patent and citation counts, and the mainstream TOPSIS method (Table 14). Table 15 presents the ranking correlations across methods.

Table 14.

Ranking of Enterprises’ Technological Innovation Level Evaluation by Various Methods (For Robustness Analysis).

Enterprises	Patent holding quantity	Patent citation quantity	TOPSIS	AHP+ entropy weight
KONINKLIJKE PHILIPS ELECTRONICS N.V	18	30	1	1
FUJIFILM CORPORATION	9	55	2	2
GE MEDICAL SYSTEMS GLOBAL TECHNOLOGY COMPANY LLC	16	13	3	3
MEDTRONIC INC	22	4	4	4
EASTMAN KODAK COMPANY	62	91	5	6
TOSHIBA CORP	14	122	6	5
MEDTRONIC NAVIGATION INC	41	11	7	7
SIEMENS HEALTHCARE GMBH	3	7	8	8
GENERAL ELECTRIC COMPANY	1	1	9	14
THE BOARD OF TRUSTEES OF THE LELAND STANFORD JUNIOR UNIVERSITY	118	196	10	13

Table 15.

Correlation of Evaluation Ranks of Enterprises’ Technological Innovation Levels by Different Methods (For Robustness Analysis).

Evaluation indicator or methodology	Patent holding quantity	Patent citation quantity	TOPSIS	AHP+ entropy weight
Patent holding quantity	1
Patent citation quantity	.797**	1
TOPSIS	.164**	.182**	1
AHP+ Entropy weight	.233**	.283**	.987**	1

Note. The correlation is significant at the .01 level of confidence (two-tailed).

p < 0.01.

Table 15 shows significant correlations across methods. Though rankings based solely on patent and citation counts correlate relatively weakly with our results, the correlation with TOPSIS—a mainstream evaluation method—is .987, indirectly supporting the scientific validity of our proposed approach.

Summary

This study constructs a knowledge flow network from patent citations in the medical device sector to measure enterprise technological innovation levels. From patentometric, innovation contribution, and knowledge-based perspectives, we developed nine tertiary indicators. Weights were assigned using a combined AHP and entropy weight method. We assessed innovation capabilities based on their contributions to industry advancement. The results inform competitive landscape analysis and guide enterprise innovation strategy.

Research Conclusions and Discussion

Research Conclusions

This study assessed the technological innovation levels of 4,116 medical device enterprises with patents from 2012 to 2024. Although robustness checks using the 2020 to 2024 window confirmed our core findings, evaluating innovation capabilities requires capturing long-term innovation activities and knowledge accumulation. A longer period also ensures greater completeness of citation data. Thus, our main conclusions are based on the 2012 to 2024 window.

The medical device sector is highly competitive, with technology-driven firms concentrated in the U.S., Japan, and Europe. Based on innovation level assessments, these enterprises dominate the market, exhibiting a significant lead over others. They serve as key drivers and benchmarks for global technological advancement.

Latecomer countries also demonstrate strong innovation capacity. Evaluation results show that most leading firms were established earlier and have extensive R&D experience in medical technology such as KONINKLIJKE PHILIPS ELECTRONICS N.V (founded 1891). Meanwhile, SHANGHAI UNITED IMAGING HEALTHCARE CO LTD from China is ranked 29th, founded in 2011, achieving rapid technological growth within just 13 years.

Although many enterprises participate in the knowledge flow network, their contributions remain limited. Technological innovation levels follow a long-tail distribution, with substantial disparities across enterprises. Beyond the top 20, all firms score below 0.1.

Discussion

1. By integrating patent statistics with knowledge flow and knowledge base considerations, this study highlights the role of knowledge in enterprise innovation. From a social network perspective (Coleman, 2000), enterprises with high centrality acquire technological knowledge and resources more efficiently, expanding innovation search breadth and improving external knowledge utilization (Ahuja et al., 2008). Betweenness centrality, reflecting an enterprise’s bridging function, enables it to control knowledge flow between subnetworks, facilitating exploratory innovation and breakthrough recombination (Zaheer & Bell, 2005). In fast-paced information environments, network topology proximity matters, but position effects must be interpreted alongside knowledge flow mechanisms. Our results assign weight (0.297) to knowledge base indicators, consistent with knowledge-based theory. Diverse knowledge elements create broader recombination opportunities, increasing breakthrough potential (Katila & Ahuja, 2002) and knowledge depth enhances innovation reliability and forms core problem-solving capabilities (Rosenkopf & Nerkar, 2001). From the perspective of dynamic capabilities, knowledge depth sustains competitive advantage on mature technological trajectories (Teece et al., 1997).

It should be noted that our interpretation of the ranking results through the lenses of Social Network Theory and the Knowledge-Based View is intended to illustrate structural correlations and theoretical alignment between the indicators and the evaluation results, rather than inferring causal effects of network position or knowledge base on innovation performance.

2. Our findings engage in multiple dialogues with existing theories and literature. First, the sustained leadership of U.S., Japanese, and European firms supports the national innovation systems (NIS) theory, whereby institutional advantages underpin technological dominance (Ferrer-Serrano et al., 2021; Pavitt, 1998). Second, the coexistence of latecomer catch-up and incumbent path dependence reflects the complexity of medical device innovation. Emerging market firms pursue niche catch-up through new technological trajectories (K. Lee & Malerba, 2017), while incumbents leverage complementary assets and accumulated experience to sustain advantages (Kang & Li, 2025). Third, the long-tail distribution of innovation performance—where most firms contribute minimally despite network participation—quantifies extreme innovation asymmetry. Knowledge influence remains concentrated among core hubs (e.g., the U.S.), as reflected in high betweenness centrality, leaving peripheral countries with a significant knowledge transfer gap (Corsino et al., 2019). Collectively, these findings depict a global innovation ecosystem shaped by structural advantages, differentiated catch-up paths, and concentrated knowledge power.

Suggestions

Fostering collaboration is essential for enhancing technological innovation, especially for latecomer firms. In China, for example, medical device companies are often small, numerous, and lack dominant players, core innovations, or visible contributions. Leveraging the strengths of these micro and small enterprises, particularly by integrating them into larger firms’ innovation ecosystems, can facilitate knowledge flow and base expansion through collaboration. This would generate high-quality innovation outputs and strengthen overall technological capabilities.

Well-structured innovation networks foster enterprise development and high-quality research environments, which enhance technological capabilities. For instance, broadening collaboration channels, embracing open innovation, and building robust international platforms to connect innovators across countries. Establishing innovation alliances, such as vertical alliances spanning raw materials, R&D, production, and application, and encouraging state-owned medical institutions to adopt domestic innovations can facilitate global knowledge circulation through coordinated efforts.

Enterprises with low innovation levels should collaborate with high-betweenness firms. In knowledge flow networks, such partners provide access to cutting-edge technological knowledge that can be absorbed and transformed through cooperation, which is a key pathway for low-innovation firms to accumulate knowledge, enrich their innovation base, and strengthen network embeddedness.

Research Limitations and Future Research Directions

Research Limitations

Given data acquisition constraints, this study constructs a knowledge flow network solely from patent citations to assess technological innovation, introducing several limitations. First, patents capture only explicit innovation; tacit innovations, such as process improvements, remain unobserved. What’s more, enterprises may withhold patents for confidential or software-related innovations. Second, knowledge flows which are generated from collaboration, personnel mobility, and IP transfers are not captured here. Third, reliance on patent data may underrepresent universities and research institutions strong in theoretical contributions. Finally, this study does not incorporate temporal dynamics.

Future Research Directions

Therefore, considering the above limitations, we propose several future research directions. First, incorporating clinical data on medical devices would enhance evaluation accuracy, although these data are difficult to obtain. Second, to construct more comprehensive knowledge flow networks, diverse data sources such as scientist mobility and enterprise collaboration should be integrated. Third, including datasets from theoretical research could broaden the scope of innovation assessment. Finally, introducing temporal dynamics would enable richer insights. For instance, constructing rolling 5-year networks to visualize structural evolution, or designing event studies around major technological breakthroughs or policy shifts can examine changes in enterprise knowledge acquisition and integration patterns. Such approaches would reveal network dynamics and provide temporal evidence extending our current cross-sectional findings.

Footnotes

Appendix A

Acknowledgements

This study does not include third-party institutions that require acknowledgments.

ORCID iD

Qianwen Cao

Ethical Considerations

This study does not involve experiments on humans or animals.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the [Philosophy and Social Science Foundation of China] under Grant [21&ZD119].

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data Availability Statement

Data sharing not applicable to this article as no datasets were generated or analyzed during the current study.

References

Ahuja

Lampert

C. M.

Tandon

(2008). Moving beyond Schumpeter: Management research on the determinants of technological innovation. Academy of Management Annals, 2, 1–98. https://doi.org/10.5465/19416520802211446

T. L.

Jiang

Wang

D. Z.

(2023). Measurement of critical technologies in China’s High-Tech manufacturing and the catch-up path: Lithium battery industry as an example. Economic Research Journal, 58(1), 192–208.

Bai

X. Y.

Duan

J. J.

Yin

Yang

(2024). Global quantitative analysis and visualization of big data and medical devices based on bibliometrics. Expert Systems with Applications, 254(11), 1–19. https://doi.org/10.1016/j.eswa.2024.124398

Bradley

Kim

Tian

(2017). Do unions affect innovation? Management Science, 63(7), 2251–2271. https://doi.org/10.1287/mnsc.2015.2414

Cai

Zhou

Liu

(2024). Patent analysis for identifying core technology and forecasting promising technology in medical imaging equipment. IEEE Transactions on Engineering Management, 71(9), 14375–14386. https://doi.org/10.1109/TEM.2024.3453652

Chang

S. B.

(2012). Using patent analysis to establish technological position: Two different strategic approaches. Technological Forecasting and Social Change, 79(1), 3–15. https://doi.org/10.1016/j.techfore.2011.07.002

Coleman

(2000). Social capital in the creation of human capital. In Lesser

E. L.

(Ed.), Knowledge and social capital (pp. 17–41). Butterworth-Heinemann. https://doi.org/10.1016/B978-0-7506-7222-1.50005-2

Corsino

Mariani

Torrisi

(2019). Firm strategic behavior and the measurement of knowledge flows with patent citations. Strategic Management Journal, 40(7), 1040–1069. https://doi.org/10.1002/smj.3016

Cricchio

Barabuffi

Crupi

Di Minin

(2025). China’s new knowledge brokers. A patent citations network analysis of the artificial intelligence open innovation ecosystem. Journal of Engineering and Technology Management, 79, Article 101870. https://doi.org/10.1016/j.jengtecman.2025.101870

10.

Cummings

Knott

(2018). Outside CEOs and innovation. Strategic Management Journal, 39(8), 2095–2119. https://doi.org/10.1002/smj.2792

11.

de Luca

L. M.

Atuahene-Gima

(2007). Market knowledge dimensions and cross-functional collaboration: Examining the different routes to product innovation performance. Journal of Marketing, 71(1), 95–112. https://doi.org/10.1509/jmkg.71.1.095

12.

de Melo

C. A.

da Silva

M. P.

Benedito

R. D.

(2021). Renewable energy technologies: Patent counts and considerations for energy and climate policy in Brazil. Climate and Development, 13(7), 630–643. https://doi.org/10.1080/17565529.2020.1848778

13.

Ferrer-Serrano

Latorre-Martínez

M. P.

Fuentelsaz

(2021). The European research landscape under the Horizon 2020 Lenses: The interaction between science centers, public institutions, and industry. The Journal Technology Transfer, 46, 828–853. https://doi.org/10.1007/s10961-020-09816-3

14.

French

Powell

Schulman

(2000). Using clustering strategies for creating authority files. Journal of the American Society for Information Science and Technology, 51(8), 774–786. https://doi.org/10.1002/(SICI)1097-4571(2000)51:8%3C774::AID-ASI90%3E3.0.CO;2-P

15.

Geisler

(2002). The metrics of technology evaluation: Where we stand and where we should go from here. International Journal of Technology Management, 24(4), 341–374. https://doi.org/10.1504/IJTM.2002.003060

16.

Guerrero-Bote

V. P.

Moed

H. F.

Moya-Anegón

(2021). New indicators of the technological impact of scientific production. Journal of Data and Information Science, 6(4), 36–61. https://doi.org/10.2478/jdis-2021-0028

17.

Haq

Davies

(2023). The person with maximum knowledge will win the race: Conceptualizing knowledge in microbusinesses. Journal of Small Business Management, 61(2), 295–321. https://doi.org/10.1080/00472778.2020.1768799

18.

Ibragimov

Mello-Thoms

(2024). The use of machine learning in eye tracking studies in medical imaging: A review. IEEE Journal of Biomedical and Health Informatics, 28(6), 3597–3612. https://doi.org/10.1109/JBHI.2024.3371893

19.

Jin

Wang

Chen

Wang

(2015). A study of the relationship between the knowledge base and the innovation performance under the organizational slack regulating. Management Decision, 53(10), 2202–2225. https://doi.org/10.1108/MD-05-2014-0253

20.

Kamasak

Bulutlar

(2010). The influence of knowledge sharing on innovation. European Business Review, 22(3), 306–317. https://doi.org/10.1108/09555341011040994

21.

Kang

(2025). How coopetition relationships impact breakthrough innovation for incumbent enterprises. Research-Technology Management, 68(5), 48–58. https://doi.org/10.1080/08956308.2025.2532319

22.

Katila

Ahuja

(2002). Something old, something new: A longitudinal study of search behavior and new product introduction. Academy of Management Journal, 45(6), 1183−1194. https://doi.org/10.5465/3069433

23.

Lai

K. K.

Chen

H. C.

Chang

Y. H.

Kumar

Bhatt

P. C.

(2020). A structured MPA approach to explore technological core competence, knowledge flow, and technology development through social network patentometrics. Journal of Knowledge Management, 25(2), 402–432. https://doi.org/10.1108/JKM-01-2020-0037

24.

Lai

Liang

Zhang

(2022). Shared-cathode closed bipolar electrochemiluminescence cloth-based chip for multiplex detection. Analytica Chimica Acta, 1206(8), 1–10. https://doi.org/10.1016/j.aca.2022.339446

25.

Lee

Yoon

(2015). The idiosyncrasy of research and development efficiency across types of small- and medium-sized enterprises: Evidence from Korea. R & D Management, 45(3), 250–266. https://doi.org/10.1111/radm.12082

26.

Lee

Malerba

(2017). Catch-up cycles and changes in industrial leadership: Windows of opportunity and responses of firms and countries in the evolution of sectoral systems. Research Policy, 46(2), 338–351. https://doi.org/10.1016/j.respol.2016.09.006

27.

Lee

Y. N.

(2015). Evaluating and extending innovation indicators for innovation policy. Research Evaluation, 24(4), 471–488. https://doi.org/10.1093/reseval/rvv017

28.

Peng

Zheng

Yuan

(2024). Cooperative R&D networks embeddedness and innovation resilience: The moderating role of environmental turbulence. European Journal of Innovation Management, 1–21. https://doi.org/10.1108/EJIM-11-2023-1031

29.

Z. B.

Sun

K. Y.

Zhao

J. J.

(2021). How does the company’s knowledge base moderating multi-source knowledge acquisition performance?-Threshold effect based on depth and breadth of knowledge. Studies in Science of Science, 39(2), 303–312.

30.

Liu

Cai

Xiang

X. Y.

(2014). The knowledge base evolution in industry: Based on patent intelligence analysis. Studies in Science of Science, 32(7), 1019–1028. https://doi.org/10.3969/j.issn.1003-2053.2014.07.009

31.

Luo

H. Y.

Lin

X. Y.

(2022). Evaluation of breakthrough technological innovation capability of petroleum equipment manufacturing enterprises based on the CFPR-TIFNs model. Complexity, 2022, 1–14. https://doi.org/10.1155/2022/6251754

32.

(2023). Evaluation of biomedical industry technological innovation ability based on a grey panel clustering model. Mathematical Biosciences and Engineering, 20(1), 1538–1557. https://doi.org/10.3934/mbe.2023070

33.

Wang

D. L.

(2023). Measurement and analysis of the scientific and technological innovation efficiency of industrial enterprises above a designated scale in China. SAGE Open, 13(3), 1–11. https://doi.org/10.1177/21582440231194578

34.

Mahmood

Saba

Rehman

Alamri

F. S

(2024). Harnessing the power of radiomics and deep learning for improved breast cancer diagnosis with multiparametric breast mammography. Expert Systems with Application, 249(9), 1–18. https://doi.org/10.1016/j.eswa.2024.123747

35.

Mannucci

P. V.

Yong

(2018). The differential impact of knowledge depth and knowledge breadth on creativity over individual careers. Academy of Management Journal, 61(5), 1741–1763. https://doi.org/10.5465/amj.2016.0529

36.

Pavitt

(1998). The social shaping of the national science base. Research Policy, 27(8), 793–805. https://doi.org/10.1016/S0048-7333(98)00091-2

37.

Pei

L. D.

Cheng

F. J.

Guo

S. Y.

Chen

Jin

Zhou

(2024). Trigonometric function-driven interval type-2 trapezoidal fuzzy information measures and their applications to multi-attribute decision-making. Engineering Applications of Artificial Intelligence, 135(9), 1–28. https://doi.org/10.1016/j.engappai.2024.108694

38.

Ponta

Puliga

Manzini

(2021). A measure of innovation performance: The Innovation Patent Index. Management Decision, 59(13), 73–98. https://doi.org/10.1108/MD-05-2020-0545

39.

Qiu

Wang

(2023). Patent citation network simplification and similarity evaluation based on technological inheritance. IEEE Transactions on Engineering Management, 70(12), 4144–4161. https://doi.org/10.1109/TEM.2021.3109134

40.

Rehman

Mahmood

Saba

(2024). Robust kidney carcinoma prognosis and characterization using Swin-ViT and DeepLabV3+with multi-model transfer learning. Applied Soft Computing, 170(2), 1–19. https://doi.org/10.1016/j.asoc.2024.112518

41.

Rosenkopf

Nerkar

(2001). Beyond local search: Boundary-spanning, exploration, and impact in the optical disk industry. Strategic Management Journal, 22, 287–306. https://doi.org/10.1002/smj.160

42.

Shukla

(2021). R&D spillover, firm size and in-house R&D: Panel data evidence from electronics goods sector in India. Singapore Economic Review, 71(1), 343–383. https://doi.org/10.1142/S0217590821500508

43.

X. L.

Jia

Xiang

H. X.

Zhu

(2023). Research progress in preparation, properties, and applications of medical protective fiber materials. Applied Materials Today, 32, Article 101792. https://doi.org/10.1016/j.apmt.2023.101792

44.

Teece

D. J.

Pisano

Shuen

(1997). Dynamic capabilities and strategic management. Strategic Management Journal, 18(7), 509–533. https://doi.org/10.1002/(SICI)1097-0266(199708)18:7<509::AID-SMJ882>3.0.CO;2-Z

45.

Velazquez-Cazares

M. G.

Gil-Lafuente

A. M.

Leon-Castro

Blanco-Mesa

(2021). Innovation capabilities measurement using fuzzy methodologies: A Colombian SMEs case. Computational Mathematical Organization Theory, 27(6), 384–413. https://doi.org/10.1007/s10588-020-09321-w

46.

Wang

X. L.

Daim

Huang

L. C.

Z. Q.

Shaikh

Kassi

D. F.

(2022). Monitoring the development trend and competition status of high technologies using patent analysis and bibliographic coupling: The case of electronic design automation technology. Technology in Society, 71(11), 1–11. https://doi.org/10.1016/j.techsoc.2022.102076

47.

Whittington

K. B.

Owen-Smith

Powell

W.W.

(2009). Networks, propinquity, and innovation in knowledge-intensive industries. Administrative Science Quarterly, 54(1), 90–122. https://doi.org/10.2189/asqu.2009.54.1.90

48.

Shanley

M. T.

(2009). Knowledge stock, exploration, and innovation: Research on the United States electromedical device industry. Journal of Business Research, 62(4), 474–483. https://doi.org/10.1016/j.jbusres.2007.12.004

49.

Duan

(2023). Medical knowledge graph: Data sources, construction, reasoning, and applications. Big Data Mining and Analytics, 6(2), 201–217. https://doi.org/10.26599/BDMA.2022.9020021

50.

H. Y.

Winnink

Yue

Z. H.

Liu

Z. Q.

Yuan

G. T.

(2020). Topic-linked innovation paths in science and technology. Journal of Informetrics, 14(2), Article 101014. https://doi.org/10.1016/j.joi.2020.101014

51.

Liu

Wang

Pang

(2025). Linkages among science, technology, and industry on the basis of main path analysis. Journal of Informetrics, 19(1), 1–22. https://doi.org/10.1016/j.joi.2024.101617

52.

Zhang

L. G.

Mao

X. L.

Bao

J. X.

(2021). Evaluating technological innovation of media companies from the perspective of technological ecosystem. Wireless Communications & Mobile Computing, 2021(12), 1–8. https://doi.org/10.1155/2021/4170619

53.

Yin

J. Y.

(2021). Study on evaluating innovation ability of high-tech industry based on particle swarm synthesis optimization. Tehnicki Vjesnik-Technical Gazette, 28(2), 616–623. https://doi.org/10.17559/TV-20190430051659

54.

L. P.

Duan

Y. L.

Fan

T. T.

(2020). Innovation performance of new products in China’s high-technology industry. International Journal of Production Economics, 219(1), 204–215. https://doi.org/10.1016/j.ijpe.2019.06.002

55.

Yuan

X. D.

Song

W. L.

(2022). Evaluating technology innovation capabilities of companies based on entropy-TOPSIS: The case of solar cell companies. Information Technology & Management, 23(2), 65–76. https://doi.org/10.1007/s10799-021-00344-6

56.

Zaheer

Bell

G. G.

(2005). Benefiting from network position: Firm capabilities, structural holes, and performance. Strategic Management Journal, 26(9), 809–825. https://doi.org/10.1002/smj.482

57.

Zhang

Zeng

Fao

(2023). Enterprise innovation evaluation method based on swarm optimization algorithm and artificial neural network. Neural Compute & Application, 35(36), 25143–25156. https://doi.org/10.1007/s00521-023-08317-2

58.

Zhang

R. Z.

Zhang

Ren

(2025). Discovering technology opportunities of latecomers based on RGNN and patent data: The example of Huawei in self-driving vehicle industry. Information Processing & Management, 62(1), 1–21. https://doi.org/10.1016/j.ipm.2024.103908

59.

Zhang

Rao

(2021). Key risks and development strategies for China’s high-end medical equipment innovations. Risk Management Healthcare Policy, 14(7), 3037–3056. https://doi.org/10.2147/RMHP.S306907

60.

Zhao

W. B.

Yin

Z. X.

Fan

T. R.

Luo

(2019). Research on influence spread of scientific research team based on scientific factor quantification of big data. International Journal of Distributed Sensor Networks, 15(4), 1–15. https://doi.org/10.1177/15501477198421

61.

Zhou

Yang

Liu

Gong

(2024). Digital technology adoption and innovation performance: A moderated mediation model. Technology Analysis & Strategic Management, 36(11), 3341–3356. https://doi.org/10.1080/09537325.2023.2209203