VisualDX

Introduction

Typical clinical decision support software requires a user to use text-words to describe symptoms and patient findings. This can create problems if users are unfamiliar with or use variations of the software's “correct” terminology. VisualDX solves this problem by creating a graphically based interface for inputting visual symptoms, thus helping users quickly answer the question, “What is this?”

VisualDx was first developed by Logical Images for pediatric, adult, and geriatric dermatologic conditions, which can be notoriously difficult for non-dermatologists to diagnose. A 1999 article found a significant difference in the diagnostic skills of dermatologists (93% correct), compared to primary care physicians (52% correct), when viewing images of the most common skin diseases (P < 0.001) [1]. VisualDx was designed to meet the needs of users who may not see dermatological manifestations everyday: primary care physicians, emergency room physicians, dentists, infectious disease specialists, and public health workers.

Since then, the product has expanded to include modules on oral lesions, pulmonary infections, and terrorism recognition, among others. In total, the knowledgebase has 21 modules containing more than 14,000 images that span over 800 diseases. That includes more than 1,700 dark skin images, a group often underrepresented in dermatology resources. The images come from a variety of sources, including university and individual archives. In addition, each condition has an associated handbook-length monograph including testing and management pearls. This unreferenced text comes from textbooks, journal articles, and medical experts.

Access

VisualDx is Java-based and can be accessed through a stand-alone application or online through the application hosted either on the Logical Images' Web server or an institution's Web server. The choice of access will depend on the institution's needs and resources. This reviewer used only the Logical Images–hosted Web version, which loaded quickly and ran well with only a minimal lag. For institutions with high-speed Internet access, this version should work well.

Searching

When users first connect to VisualDX, the introductory screen they see lists the various content modules. These modules simplify the process of navigating the database and offer search options customized for the content. The graphical search interface can be bypassed if the diagnosis is known by entering it directly into the top right search box. Otherwise, the appropriate module can be chosen based on patient demographics and clinical findings. For situations where the module choice may be unclear, there is a Help Me Select a Module utility, but, as the modules overlap significantly, module choice is unlikely to exclude potential matches or diagnoses.

For the purposes of this review, the patient is a light-skinned female adult patient who has smooth papules and erythema on her face and hands and has been losing weight. After selecting a module (in this case Adult Rash), the first step is to indicate the type of lesion. Rather than just using descriptive text, VisualDx brings up a grid with illustrations of different lesion types (Figure 1). In this case, smooth papule and erythema were chosen.

Figure 1.

VisualDx grid with illustrations of different lesion types

The next step is to select the distribution of the rash. Just as with the type of lesion, VisualDx shows images of different distribution patterns and allows users to select between extensive and limited patterns. Unfortunately, the color scheme chosen for this display can make the pattern difficult to distinguish on some monitors. In this case, photo-distributed was selected.

The final step is to enter any other findings or symptoms. These can either be selected from a hierarchical list in the Findings drop-down menu or typed into the field labeled Enter Findings. The options here are extensive and include medical, medication, and social history as well as physical, laboratory, and imaging findings. An autocomplete feature suggests potential matches as they are entered in the Enter Findings box and only allows users to enter findings that exist in the list. In this case, weight loss was entered.

As lesion type, distribution, and findings are entered into the system, VisualDx displays images of potential diagnoses ranked by the number of criteria matched. For instance, at the end of the above example, systemic lupus erythematosus is listed first with four matches followed by several other diagnoses with three or fewer matches. Because isolated findings could be unrelated to the condition being diagnosed, it is valuable that VisualDx lists all the potential diagnoses, rather than just those that match all criteria. Results can be sorted to show emergency status (life-threatening conditions) first, rather than number of criteria met. Users can also choose to view only selected diagnoses from the full list.

The images and monograph for a given diagnosis can be accessed from the results screen. Where traditional textbooks or atlases describe symptoms and offer “classic” or “severe” images, VisualDx provides many versions of typical and atypical presentations. This allows users to see the true range of variation of disease due to age, severity, time, or skin tone. Offering this breadth of manifestations allows a clinician to more comfortably match a diagnosis to a particular patient. Images can be viewed one at a time, simultaneously, or enlarged when double-clicked. The pictures and their associated monographs can be easily and clearly printed.

For users wanting more assistance with using VisualDx, a help system is provided. While basic and slightly awkward, it covers the essentials of using the interface.

Future directions

VisualDx has continually expanded their product since its inception, and a new cellulitis module should be online soon. Long-term plans include modules on ophthalmology, emergency plain film radiography, emergency ultrasound, and bites, stings, and envenomations. The possibilities for the interface seem endless; with expansion into otoscopic, endoscopic, and CT or MRI images, this tool could only become more powerful.

Conclusion

Overall, VisualDx works very well. The interface is easy to use and walks the user through the search process; the image collection is excellent; and the availability of multiple access routes would appeal to a variety of users. Lack of references, both in provenance on the images and citations on the monographs, is VisualDx's biggest flaw. External links to resources like PubMed or an attempt to integrate evidence-based medicine would also increase the overall value of the product.

A straightforward and intuitive tool, VisualDx provides a unique graphical interface to support diagnostic clinical decision making. With a wide selection of high-quality images, it would be a useful addition to the toolbox of primary care and emergency physicians, dentists, and public health workers.

Note: VisualDx's Terrorism Recognition module meets the preparedness requirements of the Centers for Disease Control and Prevention and Health Resources and Services Administration for public health departments and emergency rooms. As such, the product can be purchased using the preparedness funds made available to these institutions.