-------------------------------------- ------------------------------------| 1. DATA SOURCES | | 2. INGESTION & PREPROCESSING |
| NIH blood smear datasets, Kaggle | | Images undergo normalization, |
| hematology datasets, and PhysioNet | | denoising, and resizing. |
| imaging archives form foundational | | Classical DICOM pipelines extract|
| inputs. Patient demographics | --> | pixel metadata. Quantum-related |
| corpus. Synthetic generators | | preprocessing enriches features |
| create underrepresented variations | | with encoded phase/amplitude. |
| All data validated for quality | | Patient-wise splitting ensures |
| and checked for distortions. | | no identity leakage. |
-------------------------------------- ------------------------------------ /
/
v ---------------------------------- | 3. HYBRID FEATURE ENGINEERING |
| CNNs extract spatial |
| representations. VQC-based |
| quantum circuits encode |
| high-dimensional correlations. |
| Combined features highlight |
| subtle shape and texture cues. |
---------------------------------- | v------------------------------------| 4. MODEL TRAINING (Hybrid |
| Classical + Quantum) |
| Hybrid architectures combine CNN |
| backbones with quantum layers. |
| Cross-validation strengthens |
| generalization. Quantum circuit |
| tuning improves entanglement |
| patterns. |
------------------------------------ | | v ------------------------------------ | 5. INFERENCE & DEPLOYMENT |
| Hybrid models deployed via |
| FastAPI services with optimized |
| batch scoring. Quantum inference |
| paths simulate circuit |
| evaluation. Outputs include |
| predictions and visual cues |
| highlighting influential regions.|
------------------------------------ /
/
v-------------------------------------| 6. MONITORING & FEEDBACK LOOP |
| Prediction drift and imaging |
| distribution shifts monitored. |
| Clinician feedback refines |
| calibration thresholds. Updated |
| data triggers retraining. Enhanced|
| reliability of medical |
| predictions over time. |
-------------------------------------