Case study
QAF
Quality Assessment Framework for Photon Counting CT
Cloud-native SaaS that standardizes image quality assessment for Photon Counting CT. Researchers upload multi-energy DICOM datasets, run validated quantitative analyses (SNR, NPS, MTF, material decomposition), and export reproducible reports from one browser-based platform.
Product overview — full frame
01 — Challenge
The problem we solved
- Fragmented desktop tools and custom MATLAB scripts for quality metrics
- Poor reproducibility across labs and institutions
- Complex multi-energy DICOM management with no unified workflow
- Limited standardization for collaborative research
02 — Goals
What success looked like
- Unified cloud platform for quantitative IQ analysis
- Standardize methodologies across institutions
- Automate validated algorithms with exportable reports
- Secure collaboration for clinical and research teams
03 — Signals
Capability profile
Measured strengths across delivery, innovation, and operational maturity.
10+
Analysis modules
~500MB
Upload capacity
5
Account tiers
High
Reproducibility focus
04 — System design
Architecture
Multi-layer cloud analysis platform
Server-rendered Django monolith with AJAX JSON endpoints — presentation, processing, storage, and reporting separated for maintainability.
- 101
Presentation
Layer 1 of 5Django templatesBootstrap 5Plotly.jsAJAX / fetch - 202
Application
Layer 2 of 5Accounts & OTPStudy workflowsTier gatingSession state - 303
Processing
Layer 3 of 5PyDICOMNumPy / SciPyNPS / MTF engines.npy volume cache - 404
Storage
Layer 4 of 5PostgreSQLMEDIA_ROOT DICOMAnalysisResult JSONPlot assets - 505
Ops
Layer 5 of 5GunicornNginxGitHub Actionssystemd
05 — Stack
Technology stack
Production tools powering this product — icons for quick recognition.
Python
Django
PostgreSQL
NumPy
Pandas
Bootstrap
JavaScript
Nginx
GitHub Actions
06 — Capabilities
Key features
Multi-energy DICOM upload & study indexing
SNR, 2D/3D NPS, MTF, material ID & quantification
Interactive Plotly visualizations + Matplotlib exports
Tiered accounts (Free → Enterprise) with analysis gates
Study history and reproducible AnalysisResult storage
07 — Deep dive
Structure & detail
Core analysis modules
| Module | Output | Use case |
|---|---|---|
| SNR | Quantitative + plot | Noise characterization |
| 2D / 3D NPS | Spectrum + metrics | Noise texture analysis |
| MTF | Resolution curve | Spatial resolution QA |
| Material decomposition | ID + quantification | Spectral CT research |
| Energy bin comparison | Comparative charts | Multi-energy protocols |
Data model
| Entity | Relationship | Notes |
|---|---|---|
| User | 1 → N Studies | Custom accounts.User + tiers |
| Study | 1 → N AnalysisResult | UUID PKs |
| AnalysisResult | metrics JSON + plot | SNR, NPS, MTF, MIQ… |
| DICOM files | Filesystem paths | Not FileField — MEDIA_ROOT |
08 — Outcomes
Business impact
Research velocity
Single workflow replaces multi-tool desktop pipelines
Standardization
Shared metrics enable cross-institution comparison
Collaboration
Cloud studies + exportable publication-ready plots
Closing note
Production research SaaS that bridges clinical PCD-CT workflows and academic quality assessment with validated metrics and cloud collaboration.
