VINSIA: Visual Navigator for Surgical Information Access

Abstract

Information access at the point of care presents a different set of requirements than those for traditional search engines. Critical care in remote (e.g., battle field) and rural settings not only requires access to clinical guidelines and medical libraries with surgical precision but also with minimal user effort and time. Our development of a graphical, anatomy-driven navigator called Visual Navigator for Surgical Information Access (VINSIA) fulfills the goal for providing evidence-based clinical decision support, specifically in perioperative and critical care settings, to allow rapid and precise information access through a portable stand-alone system. It comes with a set of unique characteristics: (a) a high precision, interactive visual interface driven by human anatomy; (b) direct linkage of anatomical structures to associated content such as clinical guidelines, literature, and medical libraries; and (c) an administrative content management interface allowing only an accredited, expert-level curator to edit and update the clinical content to ensure accuracy and currency. We believe that the deployment of VINSIA will improve quality, safety, and evidence-based standardization of patient care.

INTRODUCTION

The exponential expansion of clinical information, coupled with evidence-based expert and professional society guidelines and individual evidence-based practice styles make immediate access to reference materials at the point-of-care imperative.¹ Although Web-based search engines are available, information access at the point of care² presents a different set of challenges than those for traditional information retrieval. For example, in perioperative and critical care settings, access to clinical content and guidelines must be both precise and expedient. The interface for access must be highly efficient due to time constraints in these environments, which is radically different from the performance requirements of traditional search engines where the emphasis is more on coverage (high recall) but less on quality (low precision). This low precision translates into a significant time lag when attempting to locate specific information for users.

There are two typical modes of information access.³ The first is targeted, where the user knows what to look for, comes with specific knowledge about the objective and tries to retrieve full information. The second is explorative, also referred to as informational. In this case, queries often cannot be easily or precisely formulated, due either to the user’s limited knowledge of the subject area or to the use of synonyms and ambiguous terms in the subject area. Table I summarizes the comparisons between different kinds of information access systems in terms of input effort, response speed, precision, recall, quality, and portability. Note that “input effort” is an aspect largely ignored in traditional information retrieval system. Information access at the point of care possesses all the characteristics shown in the last row of Table I, and is designed to (1) follow the logical thought processes of clinicians, (2) facilitate efficient navigation to desired references, and (3) facilitate availability at the point of care. Systems empowered with such characteristics will have significant impact in the quest to improve quality, safety, and evidence-based standardization of patient care.⁴

TABLE I.

Comparisons Between Different Information Access Systems

	Input Effort	Response Speed	Precision	Recall	Content Quality	Portability
Web Search Engine	Various	Fast (1st page!)	Low	High	Various	No
Digital Library (Targeted)	Various	Fast	High	High	N/A	No
Digital Library (Explorative)	Medium	Various	High	Medium	High	No
Information Access at Point of Care	Minimum	Fast	100%	100%	High	Yes

The Visual Navigator for Surgical Information Access (VINSIA) system we present in this article falls into the category of information access at the point of care. It is a framework for information access initially targeting perioperative and critical care settings, but is also designed to be easily expandable and deployed in other diverse clinical settings and medical specialties that possess similar information access characteristics.⁵ VINSIA is envisioned ultimately to integrate and complement other clinical systems (electronic health records) by recognizing individual patient medical comorbidities, filtering and presenting appropriate evidence-based guidelines⁶ to the caregiver for management. These guidelines will be categorized as described in later sections to benefit from anatomic body system organization. In addition, VINSIA will provide access to the literature base establishing the foundation for these guidelines and care paths in an easily navigated anatomy-based system.

Rural, community, or even some military situations where timely access to expert resources may be compromised defines a fundamental target for development of VINSIA. Several features of VINSIA are consistent with recent recommendations for development of computerized clinical decision support and knowledge management systems in community settings.^7,8 Some of the distinguishing characteristics of VINSIA include

• (1)A precise, interactive, and visual interface to convey detailed anatomic and organ system-based information quickly at the point of care
• (2)A design to provide an optimal set of user experience metrics, such as state-of-art human–computer interactions⁹ to retrieve the desired content
• (3)Public and private libraries to allow access to expert-derived content and individual user-preferred information, respectively
• (4)Assurance of information quality enabled by requiring accredited, expert-level users to edit and update the content
• (5)Annotated graphical anatomic content tagged with terminological standards such as the Foundational Model of Anatomy (FMA)¹⁰ to ensure semantic consistency across the whole system.

METHODS

The ultimate goal of VINSIA is to serve clinicians by providing evidence-based clinical decision-making support, so as to improve quality of patient care. The whole development procedure of it was fully based on interactions between clinicians and developers through weekly meetings, to best illustrate clinical perspectives.

Technically, VINSIA is developed on the open-source Semantic MediaWiki (SMW),¹¹ which is an extension of the popular MediaWiki system that has been used to create the Wikipedia.¹² SMW was developed at the Karlsruhe Institute of Technology by incorporating support from the World Wide Web Consortium Semantic Web technologies. In contrast to standard Web content, the Web content in SMW is semantically enriched and labeled, which can be interpreted not only by humans but also by computers in an accurate and consistent manner.¹³ The use of “semantic labels,” where the label names are defined in third-party formal terminologies such as ontologies,¹⁰ will allow softwares to automatically and precisely locate information in a Web resource. In addition, SMW is a flexible system that allows new functionalities to be created in the form of extensions. Hence, VINSIA was implemented on SMW, which allowed us to leverage its extensive capabilities.

Just like Amazon Books are classified to support both the targeted and the informational mode of information access, so they can be located through virtually any of the imaginable angles, information in VINSIA is organized along several facets:

• (1)User type—Physician, nurse, pharmacist: captures different types of permissions for user access control
• (2)Clinical Locale: captures content for physical locations in a clinical setting where the information is used or needed
• (3)Anatomy—Body Organ System: captures the anatomic information based on the FMA¹⁰ and follows a clinical organ system-based approach to information retrieval
• (4)Type of Information and Topic: captures the specific types of information—the desired target

These four facets form a tree structure (Fig. 1), and each tree node is called a “category” in SMW. All the information in VINSIA is stored along this structure through two steps: (1) SMW¹¹ stores the information into pages. (2) One of its functionalities called “Category” helps organizing pages into different categories. For example, if there is one page whose content is relevant to the “Aortic Valve,” it is easy to put it under the category “Aortic Valve” by labeling it. One page can be labeled with more than one category at the same time. Moreover, one of MediaWiki’s extensions named “CategoryTree”¹⁴ is able to give a dynamic view of all the category structures as a tree (Fig. 1).

FIGURE 1.

All of the information is organized by the category functionality in SMW.

The root of all the categories is called “Library” (Fig. 1). We keep two kinds of libraries to store all the information: Public Library and Private Library. The Public Library contains expert-level, high-quality content and is open to all the users. Resources in the Public Library are strictly categorized along the tree structure based on the facets mentioned earlier. Although there is only one Public Library, every user can have his own Private Library. When a new user is created, a brand new Private Library will be created for that user. Each Private Library contains specific user-chosen content only accessible by its owner. Different from the Public Library, VINSIA provides users with the flexibility to reorganize the tree structures of their own Private Libraries.

Figure 2 shows the whole information cycle in VINSIA: For the information upload side (left part of Fig. 2), only qualified and confirmed users (content experts) are allowed to upload files to VINSIA through the upload page (Fig. 3). When doing that, they can choose the Public Library and further choose the specific categories the file belongs to. Users with limited access can upload files to their own Private Library, but not the Public Library. For the information access side (right part of Fig. 2), resources in VINSIA that a user may be able to access are controlled by the user management system. VINSIA gives different access rights to different users based on their roles and user groups.

FIGURE 2.

The information cycle: users can upload information to different categories in VINSIA and then easily access them.

FIGURE 3.

On the upload page, users can choose multiple categories the file belongs to, then the file will go to the corresponding categories automatically.

Next, we give a detailed description about the four facets:

User

The extension ConfirmAccount¹⁵ of SMW is used to control user account creation so that only qualified users can get accounts in VINSIA after the system administrator reviews their profiles and accepted their requests to create accounts. After creation, the users will be put into different user groups. As mentioned before, different users in different groups are given different access rights by the system administrator, and only confirmed users (content experts) can upload files to the Public Library of VINSIA. Contrary to Wikipedia¹² or other public-accessible wiki sites, the content on each page of VINSIA is only editable by users with corresponding privileges, to guarantee the quality of the whole knowledge base in VINSIA.

Clinical Locale

As shown in Figure 1, this category indicates the associated physical locations in a clinical setting, which can be used to filter for information relevant to the specific locale. Moreover, because this facet concerns actual physical addresses instead of virtual topics, it can help VINSIA automatically detect the Locale (e.g., through the IP address) and narrow down accessible information beforehand, thereby, reducing users’ input effort.

Anatomy—Body Organ System

The visual navigation functionality is accomplished by the Anatomy facet: high quality, multilevel anatomic images are used for intuitive guidance. For a high-level image, we mark different anatomic zones on it as links to corresponding low-level anatomy pages. By simply clicking on some part of the image representing the anatomy (body organ system), users are led to another subcategory page with information about the selected anatomic feature.

In our pilot implementation, we primarily focus on the cardiovascular organ system. The direct subcategories of the heart can be found in the tree-structured representation (Fig. 1) of the anatomic features. The granularity and structure of the anatomic features are defined based on the class structure of the FMA ontology,¹⁰ with some modifications according to user feedback.

Type of Information and Topic

This category organizes information by type and topic, which includes those that are used on a daily basis in a clinical setting (Fig. 1):

• (1)Communication: It is the key to safety in complex clinical care environments involving multiple caregivers, ^16–18 providing the ability to rapidly retrieve contact information for specific clinicians or clinical care locales, which facilitates enhanced communication.
• (2)Clinical Management: It includes the basic information, guidelines, care paths, and suggested care approach information housed in a well-organized, intuitive framework that can enhance standardization of care.¹⁹ This information is at the heart of patient care and directly impacts clinical outcomes.^4,6,20
• (3)Education: It is an important aspect of clinical decision making. Access to the evidence base supporting the Clinical Guidelines and Carepaths housed in the Clinical Management section promotes improved retrieval and understanding of in-depth guidelines and education-related information.
• (4)Research: It is useful for situations where clinical research is in progress. In addition, being consistent with the Clinical Management and Education sections allows easy identification of projects and potential patients who meet the inclusion criteria for clinical studies.
• (5)Informatics and Technology: It is an area dedicated to pertinent information about the information systems that support clinical care. The technology component serves to organize and make available the information regarding the myriad clinical devices and equipment.
• (6)Process and Administration: It is an area dedicated to organize the administrative information necessary to run a division, department, or group. In addition, information on the processes involved in care is available in this section.
• (7)Quality and Safety: It is used to house information relating to patient safety, quality improvement efforts, and assessments. Details of quality improvement projects and initiatives are contained in this ever-expanding section.

The 4 facets and all the elements in them we described earlier are defined according to the current clinical system in the University Hospitals.²¹ Nevertheless, they can be easily extended and modified to fit into various settings.

Information assurance is a critical part in VINSIA, which directly impacts the quality of patient care. Apart from the user access control functionality provided by SMW and its extensions, VINSIA also has a mechanism to trail all the historical records of all the documents. Besides the technical parts, we have two more layers above it to further guarantee the content quality and safety in VINSIA: (1) Every page in VINSIA has a subpage called the “discussion” page, which provides users with a place for debate and discussion. Before tagging the page into the categories in the Public Library, final decision will be made by the moderator, who is usually a subject matter expert in the corresponding field. (2) For content already in the public domain, VINSIA collects feedback on a weekly/monthly basis, for updating after being reviewed by the content experts. As required by Federal Regulations on Electronic Records, VINSIA “employs procedures and controls designed to ensure the authenticity, integrity, and, as appropriate, the confidentiality of electronic records from the point of their creation to the point of their receipt.”²²

RESULTS

When a query is conducted, the Clinical Locale is, by default, the place where the query takes place or can be selected if in a remote location. The Anatomy facet provides the default navigation modality. With a few selections from the hierarchy, the user arrives at the desired level in the standard anatomic structure. Once a specific anatomic structure is identified, the Topic facet provides access to specific contents desired. For example, to access content specifically on aortic valve, the user selects the “heart” on the “human body,” with the following sequence: “cardiac valves”→“aortic valve”→“clinical management” (Fig. 4). This entirely visual access mode involves no typing at all. Following the navigation trail of the user, VINSIA will take the user to the appropriate page named “Clinical Management on Aortic Valve.”

FIGURE 4.

To access contents specifically on aortic valve, click the “heart” on the “human body,” followed by the click sequence “cardiac valves” → “aortic valve” → “clinical management”. This access mode only involves mouse clicking without any typing.

State-of-the-art human-to-machine technologies, such as screen-touching, were also implemented in VINSIA. As shown by the upright sub-image in Figure 4, navigation through different heart images only requires finger swiping on mobile devices, such as ITOUCH or IPAD, which empowers VINSIA with the portability provided by those devices and makes it more acceptable to end users.

The example above shows the explorative way of accessing information. A challenge for the explorative mode is that for cross-category information, it is hard to decide under which category the information should be placed, since either way will induce the risk of missing the target. VINSIA solves this problem easily by tagging multiple categories to the same page. For example, although the page “Clinical Management on Aortic Valve” in Figure 4 was explored under the category “heart,” it can also be found below the “clinical management” category, because the page is tagged with both categories.

Actually, besides the explorative way of accessing information, VINSIA also realizes the targeted mode of information access: For users who have certain knowledge about the tree structure (Fig. 1). VINSIA provides a direct search interface, powered by an extension of SMW called Multi-category,²³ as illustrated in Figure 5. After selecting the category names, as well as the options to include or exclude some pages, the user can easily get the result.

FIGURE 5.

Instead of being driven by anatomies, user can also search information based on categories.

VINSIA is also capable of storing tremendous amounts of information. In our pilot site, we only chose a small portion of two common but important organ systems: Heart and Lungs. Each of them has at least 3 levels of subcategories. In total, we get 74 anatomic categories. Since Locale has 11 subcategories and Topic has 7, the total number of pages is at least 74 × 11 × 7. Because SMW provides the functionality to add pages and categories very easily, it is predictable for VINSIA to maintain a huge volume of data in the future.

DISCUSSION

There are many existing anatomy navigation systems, such as the InnerBody Explorer,²⁴ VISIBLE BODY,²⁵ and Human-Anatomy.²⁶ However, they do not provide a medical information–retrieval mechanism based on the anatomy terms. The impact of this is that instead of related information, only the terminologies are shown when hovering over different regions of the body organ system. These anatomy navigation systems are good for exploring the human body but are not designed for anatomy-based information access.

There also exist medical support systems making use of anatomic images. For example, the Virtual Soldier Project funded by Defense Advanced Research Projects Agency aims to use both geometric data derived from images and canonical anatomic knowledge to create physiological representations of individual soldiers. The vision of this project is to use these whole-body representations to improve medical diagnosis on and off the battlefield.²⁷ At the time of injury, an information system would read the images, offer advice about the nature of the wounds, the patient’s prognosis, and requirements for therapy.

Rubin et al²⁸ developed a method to integrate patient-specific geometric data and anatomic knowledge and used that knowledge to reason about penetrating injuries. The researchers built a three-dimensional geometric model of a virtual human from segmented images by linking regions in this model to entities in a comprehensive ontology of anatomy containing organ identities, adjacencies, and other information.

VINSIA stands out from other systems in that it provides the following:

• (1)A Medical Information Retrieval System
• (2)An Information Base for Human Anatomy
• (3)An Image Aided Search Engine
• (4)A Semantic Search Engine
• (5)A User-friendly Portable System

In the future, building a comprehensive and ontology-based medical image database will be essential for VINSIA. However, a large-scale medical image base is difficult to collect as well as annotate. We have recently developed a semisupervised approach to combine visual object–detection technologies with medical ontology to automatically mine Web images and retrieved a large number of disease, organ, and drug tablet images with minimal human labeling.²⁹ For each medical term, we seek to provide a set of high-quality images with relevant contents, creating a rich information resource that could be readily used by VINSIA. The images are organized by semantic relations, which allows semantic search in VINSIA and in the knowledge base itself.

CONCLUSIONS

VINSIA provides a high-precision, visual, anatomy-based medical information retrieval system for clinicians in perioperative and critical care environments. It maintains high content quality, provides fast response, and is adjustable to portable mobile devices. The portability of VINSIA enables it to be used in diverse environments, including the battlefield. We believe that a new generation of high-precision, high-performance information access systems such as VINSIA will significantly facilitate the quest to improve quality, safety, and evidence-based standardization of care.