Ir para o conteúdo principal

Clinical Trial Workflow and Data Integration

This page describes the end-to-end workflow for psoriasis severity assessment in a clinical trial using the Legit.Health platform, from protocol design through to EDC data delivery.

Workflow overview​

The workflow has two phases: a one-time study setup, and a per-visit cycle that repeats at every assessment.

Study setup (once per study)

  1. Protocol design: Configure the study with Legit.Health
  2. Site setup: Deploy the platform to investigator sites

Per-visit cycle (each assessment)

  1. Patient enrolment: Register the patient in the platform
  2. Image capture: Guided full-body photography at the site or at home, with real-time DIQA quality control
  3. AI scoring and review: Automated PASI, reviewed by the investigator, with response milestones tracked across visits
  4. Data export: Structured delivery to the EDC system

1. Protocol design​

Before the study begins, Legit.Health works with the sponsor or CRO to configure the study protocol:

ConfigurationOptions
Scoring systemAutomated PASI, per-component scores, or custom composite
Scoring scopeGlobal (whole-body) PASI or local PASI from the available close-ups
Body areasFull body (4 perspectives + 4 close-ups), scalp focus, palmoplantar, or custom
Response milestonesWhich PASI response thresholds are flagged automatically: PASI 50, PASI 75, PASI 90, PASI 100
Visit scheduleAssessment timepoints aligned with the study calendar
Capture modelIn-clinic, decentralised (home-based), or hybrid
Alert thresholdsConfigurable PASI change thresholds for automated notifications

For studies with significant dark skin representation, performance is available stratified by Fitzpatrick group to inform protocol design; see Performance across skin types.

A study-specific investigator manual is generated for each trial, providing site personnel with step-by-step instructions, example images, and a knowledge test to confirm training. For decentralised protocols, a patient-facing manual is generated as well, with capture instructions in every required language.

The scoring model version is locked at study initiation. No mid-study model updates occur, so every patient is scored by the same validated model and endpoint integrity is preserved across the trial. Any model update requires full re-validation before deployment.

2. Site setup​

The Legit.Health clinical trials platform is deployed as a web application accessible from any browser. Each investigator site receives:

  • Login credentials for all study personnel
  • Pre-configured protocol (body areas, endpoints, scoring method)
  • Investigator manual (digital, accessible from the platform)
  • Training resources and knowledge test

No hardware installation is required; investigators use their existing smartphones for image capture.

3. Patient enrolment​

After enrolling a patient in the study per the clinical protocol, the investigator creates the patient record in the Legit.Health platform. Each patient is identified by a study-specific pseudonymised identifier; no personal data (name, date of birth) is stored in the platform.

4. Image capture​

The investigator or patient captures the standard set of photographs using the Legit.Health mobile application: 4 full-body perspectives for affected-area segmentation across the four PASI regions, and 4 close-ups of the most significant lesions for intensity scoring. The application provides:

  • Perspective guidance: visual silhouettes showing the required body area and angle
  • Real-time quality control: the DIQA quality gate checks each image at capture and flags substandard images for immediate recapture, so only adequate images proceed to scoring

For the standard 8-image protocol (4 perspectives + 4 close-ups), the entire capture process takes approximately 3 to 5 minutes.

Capture can be performed on-site by the investigator or, for decentralised and hybrid designs, by the patient at home. See Decentralised and hybrid trial support.

Photographs are anonymised before they are stored or exported, masking faces so no re-identifiable data leaves the site. See anonymisation.

5. AI scoring and review​

Once images are submitted, the AI processes them automatically in approximately <2 seconds and generates the PASI report, which the investigator reviews directly in the platform. There is no central reader and no delay.

The report contains:

  • Per-region affected area via pixel-level segmentation, with the segmentation mask per body region
  • Per-region intensity scores (0 to 4) for erythema, desquamation and induration
  • Per-region PASI contribution, weighted by each region's share of body surface
  • Global PASI score on the standard 0 to 72 scale
  • DIQA quality score per image

The investigator verifies that the AI's assessment aligns with their clinical observation. If any image is inadequate, they can recapture and resubmit; only the latest submission for each visit is retained.

From the follow-up visits onward, the same report tracks how severity evolves across the study.

PASI response milestones

Because every visit is scored by the same pipeline, response against the patient's own baseline is computed automatically: PASI 50, 75, 90 and 100 are flagged at each visit, together with the absolute change from baseline and the per-region trend.

PASI 75 and PASI 90 are the usual primary efficacy endpoints in psoriasis trials, with PASI 100 as a secondary one. Deriving them from the same pipeline that produced the scores removes the reconciliation step between the raw score and the response derivation.

6. Data export and EDC integration​

Where the in-platform report (step 5) is the investigator's human-readable view, the export is a structured, analysis-ready dataset built to load straight into the sponsor's EDC. Each assessment is delivered as column-level records, not a re-rendering of the report.

Exported data fields​

Field names below are representative; the exact schema is agreed during study setup.

GroupFields
Identifiersstudy_code, site_code, subject_code, visit_code, diagnostic_report_id, timestamp, condition_code
Source imagesoriginal_image_path and original_segmentation_image_path for every perspective and close-up, so each score is traceable to the source photograph and the AI mask
Regional scoresAffected-area contribution and intensity contribution per PASI region (head, trunk, upper extremities, lower extremities), plus the global_score
Per-sign detailOne row per body region and visual sign (erythema, desquamation, induration), each with its value (0 to 4) and per-image quality score
Response flagsPASI 50, 75, 90 and 100 achievement per visit, with absolute and percentage change from baseline

Every field is delivered per visit, so the same structure supports both single-visit analysis and longitudinal review.

Integration methods​

  • API-based: Automated data flow from the Legit.Health platform to the sponsor’s EDC system via RESTful API
  • CSV export: Structured CSV files for manual import into EDC systems
  • CRF field mapping: Data fields pre-mapped to the sponsor’s Case Report Form structure, configured during protocol design
  • QuantifiCare platform integration: For studies using the combined QuantifiCare + Legit.Health solution, scoring data flows through QuantifiCare’s platform to the sponsor’s data management system

All exported data is structured for direct mapping to CRF fields in standard EDC systems.

The integration is API-based and system-agnostic; it works with any major EDC system. Export formats include RESTful API, CSV/Excel, and structured JSON.

Decentralised and hybrid trial support​

Hybrid designs

Some study designs combine investigator-captured and patient-captured images:

  • Investigator timepoints: Primary visits (e.g., baseline, month 3, month 6, month 12) where images are captured by trained site personnel at the clinic
  • Patient timepoints: Intermediate visits (e.g., monthly check-ins) where the patient captures images at home using the mobile application

The same AI scoring pipeline processes both types of images. DIQA quality control is applied identically, ensuring that patient-captured images meet the same quality standards as investigator-captured images.

Remote patient capture

For fully decentralised protocols:

  • Patients download the Legit.Health mobile application
  • The app guides them through the capture process with the same perspective silhouettes and DIQA quality checks used at investigator sites
  • Captured images are transmitted securely to the platform for AI scoring
  • The investigator can review scores remotely
Benefits for sponsors
Decentralised and hybrid capture
  • Reduced site burden

    Fewer in-clinic visits needed for severity assessment

  • Increased data frequency

    More timepoints without increasing site workload

  • Patient convenience

    Assessments from home reduce travel burden

  • Continuous monitoring

    More frequent assessments enable finer-grained treatment response detection

For psoriasis trials, decentralised capture lets patients photograph their own affected areas at home, which matters more here than in most indications because a complete PASI needs full-body coverage at every visit. The DIQA quality gate holds home-captured images to the same standard as site-captured images.