2.1. Trustworthy Interactive Visualization for Healthcare Data
Interactive data visualization in the healthcare sector is dependent on advanced current technology, which enables professionals from a variety of professions to effectively take decisions in medical diagnosis [
21,
22,
23,
24]. Interactive visualization tools help healthcare practitioners comprehend trends that have occurred in the past as well as those that are occurring in the present, in addition to helping them predict and anticipate future trends and directions. Interactive healthcare data visualization, in its broadest sense, refers to the practice of displaying information and healthcare data for medical diagnosis purposes in a variety of formats, including graphs, charts, diagrams, and photographs [
25,
26,
27,
28]. These interactive healthcare data visualization approaches can make it simple for healthcare providers to recognize and comprehend patterns, trends, and outliers in healthcare data [
2].
Trustworthiness issues in interactive visualization techniques have become increasingly relevant in many areas of healthcare, particularly with regard to assisting medical professionals in the formulation of vitally important clinical choices for the health of patients and communities. A healthcare organization is able to transform raw healthcare data into graphs and then exhibit them in charts by utilizing a variety of approaches for healthcare data visualization [
14]. This enables the organization to perform rapid analysis of trends and patterns.
A trustworthy tool or piece of software provides efficient and interactive visualization by utilizing systems associated with healthcare, examining threats, and responding to incidents swiftly and instinctively. To solve the many challenges that come with healthcare data visualization, specialists and researchers have brought a variety of assurance issues to light [
17]. The overall level of trustworthiness of healthcare interactive visualization tools can be broken down into their component parts. The combination of these characteristics can vary depending on who is receiving treatment. For consumers, for instance, trustworthiness in healthcare visualization tools consists primarily of a perceived amount of control and privacy. On the other hand, for healthcare professionals, trustworthiness in visualization involves a far bigger and more diverse collection of concerns, such as reliability and a transparent healthcare data storage policy. When compared to the sets of elements that affect trustworthiness in the general healthcare data visualization domain, the sets of factors that affect trustworthiness in a healthcare portal are unique. In order to ensure that interventions are trustworthy, it is necessary to carry out independent research on the topic of trustworthiness in healthcare interactive visualization tools as a separate topic.
A case study was conducted on a few different alternatives in order to enhance the performance of healthcare interactive visualization tools and rank the characteristics of healthcare interactive visualization tools in descending order of importance [
8]. A consensus was reached among the authors regarding the characteristics to be used in the evaluation of healthcare interactive visualization tools, and these conclusions informed the decision-making process for identifying and selecting the criteria. For the purpose of this work, five characteristics of trustworthiness, four characteristics of visualization design, and three characteristics of interactive widgets were taken into consideration for the idealness assessment. The decision that experts in a particular field make collectively is what drives the process of alternative selection. Further, alternatives were chosen as per the popularity of interactive visualization tools among healthcare professionals. The authors of the study used fuzzy logic for this evaluation, so each healthcare data visualization was assigned a value between 0 and 1 for each characteristic.
Moreover, with regard to our specified characteristic set, each healthcare data visualization tool gained a value between 0 and 1 for each characteristic. In addition, the results of the evaluators’ subjective cognition in linguistic words for each healthcare data visualization characteristic were based on the scale and the opinions of the experts, both of which are explained in the methodology portion of the paper. This work presents, on the basis of the identified characteristic set, the process of evaluating the six tools used for healthcare as well as the quantitative outcomes of that evaluation. The characteristics that were found and the alternatives are shown in
Figure 2. The figure that follows discusses both the subsection description and the significance of the traits that were identified.
Characteristics at the first level are further divided into sub-characteristics. Trustworthiness is affected by confidentiality, integrity, reliability, availability, and performance [
19]. Visualization design is affected by availability, user friendliness, performance, and operational cost. Interactive widgets in a visualization tool depend on its characteristics, such as reliability, user friendliness, and operational cost. The first-level characteristics depend on each other for the best interactive visualization tool building. Hence, after creating its dependencies, it becomes a network-like structure problem. For such a multiple-criteria network-like structure problem, the Analytic Network Process (ANP) is a mathematical theory that was developed by Saaty [
28] that is helpful for forecasting and presenting the influence of multiple decision criteria, their interactions, and their relative weights. ANP was named after the nature of its problem, which is a network-like structure of different criteria. To calculate the different weights of characteristics associated with this problem, we used ANP with fuzzy logic and TOPSIS for ranking the alternatives.
2.1.1. Trustworthiness
Interactive visualization tools are having an increasingly substantial impact on clinical decisions and diagnoses as a result of the significant increase in the number of healthcare solutions. On the other hand, there is scant evidence to support the notion that software can be trusted. An application or tool is said to have trustworthiness when it possesses the qualities that make it trustworthy for others, such as healthcare professionals. There are different characteristics of trustworthiness that contribute to building a trustworthy visualization tool. We identified five important characteristics, the details of which are as follows:
Confidentiality: To maintain healthcare data privacy, it must be protected from digital and physical intrusion. Confidentiality is closely tied to other aspects of information privacy, such as who can see, share, and use specific pieces of healthcare data. Information with a low level of confidentiality may be considered “public” or innocuous if disclosed to a larger audience. High-confidentiality information must be protected against disclosure for the sake of avoiding identity theft, account and system compromise, legal or reputational damage, and other undesirable outcomes. The importance of privacy in keeping healthcare data visualization trustworthy is thus demonstrated.
Integrity: Healthcare data integrity is vital because so much depends on it. A dataset error in healthcare data can impact a clinician’s decision making. Healthcare data integrity means accuracy, completeness, and consistency. Healthcare data integrity includes security for regulatory compliance. It decreases consumer assurance and trustworthiness. There are several data integrity threats. Copy-transferred data should not be modified between updates. Error-checking and validation maintain data integrity when sent or reproduced without alteration.
Reliability: For a clinician or healthcare expert to use a visualization tool to ensure the accuracy of healthcare data, the data must be complete and correct. This is what is meant by “data reliability.” One of the main goals of data integrity programs, which are also used to maintain data security, data quality, and regulatory compliance, is to make sure data is trustworthy. The term “data reliability” refers to how consistent information is between different databases, apps, or platforms. It also has to do with how trustworthy the source of information is. If the data are trustworthy enough, a trustworthy figure will always be right. Thus, it can be seen as a sign of honesty.
Availability: The frequency with which healthcare data can be used by a healthcare provider organization or a partner is a measure of data availability. A healthcare provider organization runs more smoothly when clinicians have access to data at all times. Important aspects of data availability include access to the data and a steady stream of data. Data that cannot be accessed are as good as useless. As a result, data availability is the only factor that can influence the trustworthiness of interactive data visualization. If workers were to have problems obtaining firm data, productivity would take a hit. Data accessibility is critical for most modern enterprises. The good news is that, by following data availability best practices, a forward-thinking firm may enjoy all the advantages of having sufficient data availability.
2.1.2. Visualization Design
User-Friendly: The goal of good visualization tool design is to ensure that a product is not only functional but also pleasurable and simple to use for healthcare experts. The interactive visualizations could be understood by every level of health expert, be it a researcher in healthcare or a doctor. The goal is to guarantee that every consumer of an interactive visualization tool is completely content with their experience using the product. An improved user experience is the result of a design that facilitates the user’s goals and duties. The design must be simplified, the instructions must be clear and succinct, and the learning curve must be minimized.
Operational Cost: Visualization tool developers have never been in a position where they are not under pressure to cut expenses and maximize investment. This pressure has only risen as a result of COVID-19, as more and more hospitals and healthcare organizations have sped up their digitalization efforts in an effort to preserve a viable, virtual corporate presence. These operational cost optimizations have a direct impact on the design, trustworthiness, and interactive widgets of tools. Hence, operational cost is an important criterion for measuring a trustworthy and interactive visualization tool for healthcare professionals.
2.1.3. Interactive Widgets
The goal of interactive visualization is to improve the way in which people engage with information by using graphical representations of healthcare data. The term “interactive visuals” can also apply to the graphical displays that are employed by various technologies for analytics and business intelligence. The majority of the time, these representations are implemented in the form of interactive widgets. These widgets offer a simple method for comprehending insights that may be based on data that are constantly shifting. In order for healthcare data visualizations to be called interactive, they need to incorporate some sort of human input (such as the ability to click on a button or move a slider), and their response times need to be fast enough to demonstrate a genuine connection between the healthcare data input and the visual output.
Performance: Problems with performance are characterized by a reaction time for output that is significantly longer than the time that is anticipated for its execution in a healthcare interactive visualization tool. The performance could be caused by untrustworthy third-party healthcare data, such as databases or hardware, or it could be caused by the design of the visualization. Hence, performance issues can be affected by trustworthiness, visualization design, and interactive widgets as well.
All of the features that were covered earlier have some bearing on interactive visualization tools. In addition, each of the mentioned traits, by virtue of the implicit specifications that they carry, plays an important part in the ideality of interactive visualization for healthcare as a whole. The authors of the study began by compiling a list of the multiple characteristics that were pertinent to the investigation. Following that, a conversation was had with the team of healthcare and security tool development experts about finalizing the characteristics set. After having a group discussion about all of the detected features, the specialists deleted any characteristics that were deemed unnecessary or inconsistent. The individual disagreements that arose among the experts over the characteristic selection were brought to a minimum, and at the end of this expert group debate, a collection of trustworthy healthcare data visualization characteristics was decided upon. So, each of these characteristics was taken into consideration for this analysis.
2.2. Methodology
The Analytic Network Process (ANP) of MCDM can solve networked decision-making problems. T.L. Saaty invented the technique in 1965. The approach has changed and improved since then [
28]. It estimates the relative relevance of criteria (characteristics). It helps specialists choose the judgement that best fits their goal and understanding of the situation. Fuzzy data improve this process and helps elicit more accurate healthcare data [
29].
Fuzzy sets are crucial here. Fuzzy sets accurately reflect decision makers’ foggy preferences. Fuzzy logic can eliminate doubts when defining an element’s membership in a fixed set is challenging. This is very useful for categorizing elements. Fuzzy logic rarely addresses such issues [
30]. TOPSIS also produces the highest alternate rating [
8]. Its main draw is this ability. Thus, integrating fuzzy logic with the ANP-TOPSIS methodology makes this study more effective and allows it to be used to evaluate the efficacy of interactive visualization achieved through the software of a hybrid medical expert system.
This article evaluates each visualization tool’s success using the ANP-TOPSIS main methods. The fuzzy TOPSIS approach first determines the weight of each characteristic for the fuzzy ANP method. To establish correct weights, correlations between attributes are examined. After determining these attributes’ key weights, the fuzzy TOPSIS technique is used to assess decision makers’ social and economic performance and risk. These methods require interactive visualization specialists and a hybrid medical expert system. The specialists must have at least three years of experience selecting, managing, and conceptualizing interactive visualizations utilizing a hybrid medical expert system.
Figure 3 shows the integrated fuzzy ANP TOPSIS method:
2.2.1. Fuzzy Analytic Network Process
There is a significant potential for fuzzy-based ANP-TOPSIS to tackle MCDM challenges generated by imprecise and uncertain healthcare data [
28,
30]. When applied in a fuzzy environment, ANP yields characteristic weights that are more accurate, which in turn leads to outcomes that are more beneficial [
28]. The TOPSIS method, when applied to problems involving MCDM, is one of the better-known techniques for ranking available solutions [
28]. In this particular study, a total of 5 interactive visualization tools were used as alternatives and 12 characteristics of interactive visualization were used as criteria. When identifying and selecting the qualities, both the views of the specialists and well-known research works were taken into consideration.
Thomas Saaty [
28] came up with the idea of incorporating the ANP, which is an extension of the AHP (Analytic Hierarchy Process). It enhances the capacity to handle interactions and dependencies between characteristics and sub-characteristics, which might alter the weights that are allocated to them. This can be a challenge when trying to build a model. Despite the many attempts to alter it so that it can account for erroneous human judgments, the ANP is severely constrained when it comes to evidentiary assessment fractions. The fundamental concept that underpins this is that the aggregation strategy utilized by the ANP is a reasonably straightforward one that can be applied to intervals as well as locally fuzzy priorities. However, in order to carry out the supermatrix priority derivation technique in the ANP, one must be familiar with complex real-number matrix operations. Nevertheless, none of the currently available methods for determining interval or fuzzy local priorities give results that can be incorporated into the calculation of the ANP matrix. This is the case despite the fact that these methods have been developed. In order to solve the unpredictability that surrounded the preferences of experts in a pairwise comparison matrix, the fuzzy ANP (PWCM) was developed. The following is a rundown of the various stages of the fuzzy ANP strategy that will be used in the calculation of the weights of the qualities:
Step 1: The process of constructing the network of the problem involves characterizing the relationships that exist among its numerous parts.
Step 2: In the following step of the procedure, triangular fuzzy numbers (TFNs) based on the scale proposed by Saaty [
28] are used to compare various linked properties of the network pairwise and create comparison matrixes. The linguistic term “comparison” is used in this study, and the appropriate TFNs were assigned [
30].
Step 3: Creating the Supermatrix.
In order to construct the supermatrix, it is necessary to ascertain the relative importance of each characteristic and sub-characteristic. As each comparison matrix’s characteristics have a triangular fuzzy structure, the respective weights are calculated using Lotfi Zadeh’s proposed extent analysis method [
28]. This method consists of the following steps and is necessary because of the triangular, fuzzy membership structure.
Step 3.1: Finding the value of the fuzzy artificial degree for each of the following qualities is the first step in Zadeh’s technique:
Let us assume that
Pw is a pairwise comparison matrix in (Equations (1)–(5)):
Step 3.2: Calculating the relative likelihood of each Si over other characteristics (Equation (6)).
Step 3.3: When using Equations (7) and (8), it can be difficult to determine how much weight each characteristic ought to be given in terms of the probability that a convex fuzzy number is greater than k other convex fuzzy numbers.
Step 3.4: Estimating the normalized weight vector using (Equation (9)).
Once the weights of each pairwise comparison matrix have been computed, the supermatrix can be created according to Equation (10):
Step 4: Computing the actual weight vector of each sub-characteristic (Equation (11)).
2.2.2. Fuzzy Technique for Order of Preference by Similarity to Ideal Solutions
The fuzzy TOPSIS method was first put forth by M. M. D. Widianta et al. [
31], and since then it has been widely used for evaluating alternatives in a range of contexts. This technique can be used to rank choices according to how similar or close they are to the ideal response. On this page [
29], there is a lot of information about how to use fuzzy TOPSIS.
Step 1: Calculating the decision matrix that has been normalized.
Supposing that
Dnd is the normalized fuzzy decision matrix (Equation (12)):
Every aspect of decision making, with the exception of the decision-making process itself, has been standardized according to the category that best fits each principle. The principle decides whether it is a benefit principle or a cost principle, which means that an increase in magnitude is advantageous in the first category; however, in the second category, a decrease in size is advantageous. To calculate each one, the following equations are used, which are classified by each component of the normalized decision-making process (Equation (13)):
where
is the maximum
for the benefit characteristic (Equation (14)).
where
is the minimum
for the cost characteristic.
Step 2: Calculating the weighted normalized decision matrix.
The weight of the characteristic is represented as wi, and the weighted normalized decision matrix is determined as follows (Equation (15)):
Step 3: The fuzzy positive ideal and the negative ideal as potential answers are suggested in this step.
for the benefit characteristic; for the cost characteristic.
Where is the maximum , is the minimum , and i = 1,….., r; j = 1,……, s.
Step 4: Calculating the distance of each alternative from positive ideal and negative ideal solutions (Equations (16) and (17)).
The distance between two TFNs can be calculated (Equation (18)):
Step 5: This step is performed to calculate the closeness coefficient factor (CCF) (Equation (19)).
Step 6: Prioritizing the alternatives.
The algorithm that is used to determine the closeness coefficient gives more weight to the choice that has the highest CCFj value; hence, the one that has the highest CCFj value is the one that comes out on top in the ranking list.