FlowLeap
An integrated platform combining AI-powered medical imaging analysis with high-performance CFD simulation for clinical and engineering applications.
What is FlowLeap?
FlowLeap bridges the gap between medical imaging and engineering simulation. It provides an integrated environment where clinicians can analyze patient anatomy and engineers can simulate fluid dynamics — all from the same interface.
Two workflows, one platform
Navigating the Interface
FlowLeap's interface consists of three main areas: the left sidebar on the left, the 3D viewport (3D viewer) in the center, and the top toolbar. A chatbot widget provides contextual help throughout.
The left sidebar (left sidebar)
The left sidebar is your data navigator. It organises patients and projects in a hierarchical tree:
| Mode | Level 1 | Level 2 | Level 3 | Level 4 |
|---|---|---|---|---|
| Medical | Patient | Event (EID) | Acquisition (AID) | Series → Anatomy, Spirometries, Reinflations |
| Engineering | Project | Device (EID) | Version (AID) | Geometry, Materials, Zones, BCs, Simulations |
The 3D viewport (3D viewer)
The central canvas is powered by the rendering engine and supports real-time interaction:
- Rotate: left-click + drag
- Pan: middle-click + drag (or Shift + left-click)
- Zoom: scroll wheel
- Quad View / MPR: switch to three orthogonal planes + 3D in the top toolbar
Tip: Use the DivLayoutLens toggle at the bottom-center of the viewport to enable lens/focus view for inspecting small regions.
Switching between modes
Click the mode selector in the top-left toolbar to switch between Patient mode (medical) and Project mode (engineering). Data persists across mode switches.
Medical Workflow Overview Medical
The Medical Workflow enables clinicians and researchers to move from raw imaging data to quantitative clinical insights in five stages:
Data hierarchy
FlowLeap organises all medical data in a four-level tree stored in the database:
Patient (ID)
└── Event (EID) — e.g. "2024-01-15_CT"
└── Acquisition (AID) — e.g. "CTscan"
└── Series (SID) — e.g. "scan001"
├── Image Data
├── Anatomy ← segmentations & 3D meshes
├── Virtual Spirometries
└── Reinflations
Clinical applications
Preoperative Planning
Visualise anatomy and predict post-resection function before surgery.
Respiratory Assessment
Evaluate lung function without invasive breathing tests.
Surgical Outcome Prediction
Predict lobar reinflation after lobectomy with the SURGELUNG module.
Research
Correlate anatomy with function across patient populations.
Step 1 · Uploading DICOM Data Medical
Import CT or MRI scan data into FlowLeap by creating a patient record and uploading DICOM files. FlowLeap accepts DICOM folders, ZIP archives, and VTI volumes.
Step-by-step
Create or select a patient
In the left sidebar, find the Patients section. Click + New Patient and fill in the patient ID, name, date of birth and gender. Or click an existing patient to load their record.
Create an Event
Click + Add Event on the patient node. Use a descriptive ID such as 2024-01-15_CT (date + modality). Events group imaging sessions chronologically.
Open the Radiology Import panel
Select the event and click Upload Data in the toolbar, or right-click → Upload DICOM. The Radiology Import panel opens (shown above).
Drop or browse your files
Drag a DICOM folder, ZIP archive, or VTI file onto the dotted drop zone. Supported formats: .zip .vtl .nii .nrrd .mhd .dicom .png .jpg. A progress bar confirms the upload.
Verify the scan in the viewer
Click the uploaded series in the left sidebar to open it. The viewer shows three orthogonal planes (axial, coronal, sagittal) + a 3D volume. Use the Window/Level dropdown presets (Lung, Soft Tissue, Bone) to optimise contrast.
Tip: For large datasets (>1 GB), compress to ZIP before uploading to reduce transfer time and ensure all files arrive as a single bundle.
Step 2 · Organ Segmentation Medical
Segmentation isolates anatomical structures within the 3D volume. Navigate to Modules → Patient Mode → Segment and Reconstruct to open the segmentation panel.
the AI segmentation engine — AI-powered segmentation
the AI segmentation engine uses deep-learning models to automatically identify multiple organs in one pass. It is the recommended starting point for all segmentation tasks.
Select a target
In the Segmentation section, open the Target dropdown. Available targets include lungs, cardiac, and many vascular structures. Select the anatomy you need.
Run the AI segmentation engine
Click the circular run button below the target selector. Processing takes 10 seconds to 3 minutes depending on target complexity. A progress indicator shows status.
Review the segment list
When complete, all detected structures appear in the Merge segments list (e.g. lung_upper_lobe_left, lung_lower_lobe_right, trachea, pulmonary_vein). Use the eye icons to toggle visibility per structure.
Manual tools
Use manual tools to refine the AI segmentation engine results or segment structures the AI model misses:
| Tool | Use | Key parameter |
|---|---|---|
| Region Growing | Expand from a seed point within intensity range | Lower/Upper threshold (HU) |
| Thresholding | Select all voxels within an HU range | Min/Max HU values |
| Brush | Paint mask manually slice-by-slice | Brush size (px), 3D mode |
| Eraser | Remove painted regions | Eraser size |
| Scissors/Contour | Draw precise boundaries | Contour closure |
Saving segmentation
Click Save Segmentation → enter a name → choose Anatomy Database (stores in patient record) or export as NIfTI / NRRD. Saved segmentations appear under the series node in the left sidebar.
Before proceeding: verify in the 3D preview that segmentation covers the full organ, has no large holes, and has no significant overlap with adjacent structures. Use Mask Refinement tools (island removal, hole filling, smoothing) to clean up.
Step 3 · 3D Reconstruction Medical
3D reconstruction converts the voxel mask into a continuous, watertight surface mesh. Open the Reconstruction tab inside the Segment and Reconstruct panel.
Reconstruction pipeline
Marching Cubes
Extracts the initial surface from the binary mask. Key settings: Iso-Value (default 0.5 for binary masks) and Decimate Target — the percentage of triangles to retain (default 25%, range 5–100%).
Smoothing
Eliminates staircase artefacts from voxel-to-surface conversion. Three methods available: Laplacian (fast, may shrink slightly), Taubin (no shrinkage), Sinc (highest quality, recommended). Set Iterations (15–20 is a good default).
Hole Filling
Set a maximum hole diameter (mm) and click Fill Holes. Check Validate Mesh afterwards — the mesh should report as Watertight and Manifold before proceeding.
Save or Export
Click Save Reconstruction to store in the database (accessible via left sidebar → Series → Anatomy). Or export as VTP, STL, or OBJ for external tools.
Mesh quality targets
| Parameter | Target range | Note |
|---|---|---|
| Triangle count | 50 K – 200 K | Balances quality and performance |
| Min quality | > 0.7 | Shown in the Mesh Info panel |
| Watertight | ✓ | Required for volume calculation and simulation |
| Manifold | ✓ | No edge shared by more than 2 triangles |
Step 4 · Virtual Spirometry Medical
Virtual spirometry computes standard respiratory function metrics (FEV1, FVC, FEV1/FVC) from reconstructed lung geometry — no patient effort or spirometer required. Navigate to Modules → Patient Mode → Virtual Spirometry.
Setup
Load lung geometry
Click Load Anatomy — the system searches for reconstructed lungs in the current acquisition. Select left, right, or both. Optionally load airway geometry (improves accuracy of flow predictions).
Enter patient parameters
Fill in Age, Height, Weight, and Gender in the Patient Parameters section. These are used to normalise results against population predicted values. The BMI field auto-calculates.
Set mechanical parameters
The key settings are Compliance (default 80 ml/cmH₂O for healthy lung) and Damping (default 0.08 m/s²). Enable Fibrosis (Expt.) to model stiffer tissue. Choose the breathing manoeuvre: Forced expiration or Slow VC.
Run the analysis
Click the run button. Processing takes 10–60 minutes depending on quality. Results populate in the Virtual Spirometries node in the left sidebar when complete.
Interpreting results
| Metric | Normal range | Pattern |
|---|---|---|
| FEV1 % predicted | ≥ 80% | <80% = reduced |
| FVC % predicted | ≥ 80% | <80% = restricted |
| FEV1/FVC | ≥ 70% | <70% = obstructive |
Important: Virtual spirometry is a clinical decision support tool. All diagnostic decisions must be made by qualified healthcare professionals and validated against actual spirometry where available.
Step 5 · Reinflation Analysis Medical
The SURGELUNG module predicts how remaining lung lobes will expand to fill the pleural space after lobectomy. Navigate to Modules → Patient Mode → Reinflation Analysis.
Workflow
Load preoperative anatomy
Click Load Preoperative Lungs. Lobar segmentation is required — if not done, use the fissure detection tool or segment lobes manually first.
Define the surgical resection
Select Resection Type (single lobectomy, bilobectomy, segmentectomy) and choose which lobe to remove. Click Preview Resection to visualise the surgical boundary in the viewport.
Set biomechanical parameters
Use a preset (Healthy Adult, COPD/Emphysema, Pulmonary Fibrosis, Elderly) or set tissue elasticity (Young's modulus), pleural pressure, and boundary conditions manually.
Run the simulation
Select quality (Fast 10–20 min / Standard 20–40 min / High 40–90 min) and click Run Reinflation Analysis. Results are saved under the Reinflations node in the left sidebar.
Review and export
The results panel shows volume changes per lobe, predicted postoperative FEV1 and FVC, a before/after 3D comparison, and an expansion colour map. Export as PDF report or CSV for clinical documentation.
Risk thresholds
| Risk level | Predicted postop FEV1 | Net volume loss |
|---|---|---|
| ✅ Low | > 60% predicted | < 20% |
| ⚠️ Moderate | 40–60% predicted | 20–30% |
| 🔴 High | < 40% predicted (<800 mL) | > 30% |
Engineering Workflow Overview Engineering
The Engineering Workflow is designed for biomedical engineers and device manufacturers who need to simulate fluid dynamics inside anatomical or device geometries using the the simulation solver CFD/DEM engine.
Geometry Import Engineering
Import your device or anatomical geometry into an Engineering project. FlowLeap supports STL and VTP files, as well as direct import from OnShape via the CAD integration.
Material assignment
After importing geometry, assign physical fluid properties to the internal volume:
| Property | Blood (default) | Unit |
|---|---|---|
| Density | 700 | kg/m³ |
| Viscosity | 2.e-6 | Pa·s |
| Thermal conductivity | 5.1e-6 | W/mK |
| Specific heat | 5.2e-7 | J/kgK |
CFD Simulation Engineering
Configure and launch a the simulation solver CFD simulation job. FlowLeap provides a Slow ↔ Rapidity slider to trade off accuracy vs. compute time, with an automatic workload estimate before you commit.
Zone definition
Boundary conditions
Set inlet/outlet conditions for each face of the geometry. FlowLeap supports:
- Velocity Inlet — specify flow rate or velocity profile
- Pressure Outlet — specify static pressure at outflow
- Wall — no-slip or slip boundary
- Symmetry — for half-model simulations
Results Visualization Engineering
Once a simulation job completes (status: 100% green), open the Results section in the simulation panel to visualise scalar fields and animations.
Available scalars
| Scalar | Description |
|---|---|
| Velocity [m/s] | Flow speed at each point; blue = slow, red = fast |
| Pressure [Pa] | Static pressure distribution |
| Wall Shear Stress | Tangential stress on vessel/device walls |
| Temperature [K] | For thermal simulations |
| Particle density | For DEM particle tracking |
Export
Use the Clipping sub-panel for cross-sectional views, the Animation sub-panel to step through transient results, and the download button (↓ icon near outlet markers) to export results as VTP files for post-processing in ParaView or similar tools.
