Written and maintained by CASRAI Editorial Board
Last updated
VisualDx is a point-of-care visual clinical decision support (CDS) platform used by emergency medicine physicians, primary care clinicians, hospitalists, dermatologists, infectious disease specialists, and medical educators to construct differential diagnoses and compare peer-reviewed clinical photographs across dermatologic, infectious, immunologic, genetic, ophthalmology, and drug-induced conditions. By combining a structured diagnostic ontology with a curated multi-skin-tone medical image library and computer-vision lesion analysis, VisualDx bridges diagnostic gaps when frontline clinicians evaluate complex cutaneous and systemic presentations.
What Is VisualDx? Core Clinical Decision Support Architecture
Unlike text-only clinical reference encyclopedias that require the clinician to already suspect a specific disease name before searching, VisualDx is architected around backward-chaining visual and symptomatic entry points:
- Differential Builder: Clinicians enter a primary morphological finding (such as maculopapular rash, purpura, targetoid lesion, vesicles/bullae, or retiform purpura), affected body location, patient age/sex, systemic symptoms (such as fever, arthralgia, or mucosal involvement), immunologic status, geographic travel history, or environmental exposures. The inference engine computes a ranked differential diagnosis displaying side-by-side clinical images of how each candidate disease presents across early, classic, and severe stages.
- DermExpert (Embedded Machine Learning Image Analysis): Within the VisualDx mobile application (iOS and Android), clinicians using DermExpert can photograph a patient’s skin lesion at the bedside. An on-device/cloud convolutional neural network analyzes the visual morphology of the lesion, estimates the most probable lesion types and diagnostic categories, and launches the VisualDx Differential Builder pre-populated with those visual findings for physician review.
- Adverse Drug Events (Medication Reaction Checker): When a patient presents with a suspected cutaneous drug eruption (ranging from benign morbilliform drug exanthems to life-threatening Stevens-Johnson syndrome [SJS], toxic epidermal necrolysis [TEN], or DRESS syndrome), the clinician enters the patient’s active medication list to rank which drugs have documented literature associations with the observed rash morphology.
| VisualDx Component | Clinical Input | Diagnostic & Workflow Output |
|---|---|---|
| Differential Builder | Lesion morphology, body distribution, systemic signs, travel/exposure history | Weighted differential diagnosis with multi-image visual comparison and diagnostic pearls |
| DermExpert AI | Smartphone clinical photograph of a skin rash or lesion | Automated lesion morphology classification feeding into the Differential Builder |
| Adverse Drug Events | Patient’s active medication list + observed cutaneous reaction pattern | Ranked culprit medication probabilities and reaction severity guidance |
| Patient Handouts | Confirmed diagnosis selected by the treating clinician | Plain-language patient education summaries paired with skin-tone-matched reference images |
Fitzpatrick Skin Types IV–VI Representation and Project IMPACT
A well-documented limitation in dermatologic education and clinical AI dataset curation (highlighted in TRIPOD+AI fairness guidelines and hospital AI governance audits) is the historical underrepresentation of pigmented skin (**Fitzpatrick skin phototypes IV, V, and VI**) in medical textbooks—where erythema, cyanosis, violaceous plaques, and target lesions appear markedly different than on lighter skin (Fitzpatrick types I–III).
VisualDx addresses this diagnostic equity gap through two structural mechanisms:
- Pigmented Skin Image Curation: Nearly 30% of VisualDx’s curated medical image collection specifically depicts dermatologic and systemic conditions on Fitzpatrick skin types IV–VI, verified by an international editorial board of dermatologists and infectious disease specialists.
- Skin of Color Filtering & Project IMPACT: Clinicians can toggle a persistent “Skin of Color” view filter inside the Differential Builder to prioritize dark-skin reference photographs first. In parallel, VisualDx coordinates Project IMPACT (Improving Medicine’s Power to Address Care and Treatment), a global collaborative initiative dedicated to eliminating racial and ethnic disparities in dermatologic diagnosis.
SMART on FHIR, HL7 Infobutton, and EHR Integration Workflows
To operate inside acute and ambulatory clinical workflows without manual data re-entry, VisualDx integrates directly with major enterprise EHR systems—including Epic Systems (via the Epic Showroom / App Market), Oracle Health (Cerner), and MEDITECH:
- SMART on FHIR Contextual Launch: Launching VisualDx from an open EHR chart invokes HL7 FHIR R4 US Core resources (`Patient`, `Condition`, `MedicationRequest`, and `Observation`) to automatically pass the patient’s age, biological sex, active problem list, and current medications into the VisualDx workspace.
- HL7 Context-Aware Knowledge Retrieval (Infobutton Standard): When a clinician clicks an Infobutton link next to a SNOMED CT or ICD-10-CM diagnosis code on the EHR problem list or order set, VisualDx opens directly to the corresponding monograph, displaying diagnostic criteria, biopsy/laboratory workup steps, and visual look-alike differentials.
Frequently Asked Questions
What is VisualDx?
VisualDx is a point-of-care visual clinical decision support (CDS) system that helps clinicians build differential diagnoses, evaluate drug reactions, and compare peer-reviewed clinical images across dermatologic, infectious, and systemic diseases—featuring the DermExpert AI lesion analyzer, extensive Fitzpatrick skin type IV–VI representation, and SMART on FHIR EHR integration.
How does DermExpert work inside VisualDx?
DermExpert is an embedded machine learning feature in the VisualDx mobile app that allows a clinician to take a photograph of a skin lesion or rash. The algorithm classifies the primary lesion morphology and populates the VisualDx Differential Builder so the clinician can combine visual features with clinical history to narrow the differential diagnosis.
How does VisualDx integrate with Epic and other EHR systems?
VisualDx integrates with Epic, Oracle Health, and other certified EHRs using SMART on FHIR (HL7 FHIR R4) and the HL7 Infobutton standard, automatically reading patient demographics, SNOMED CT / ICD-10 problem lists, and active medications from the open chart.
