Photograph the report
Upload a photo or scan. The text reader finds the results table and pulls out each value and the patient's sex.
Haematology decision support
Photograph a CBC report or type the values in. MediMatrix checks every parameter against its reference range, screens for six common blood disorders, and lists the next tests worth ordering — in English and Bangla.
How it works
The report photo is deleted as soon as its values have been read.
Upload a photo or scan. The text reader finds the results table and pulls out each value and the patient's sex.
Check what was read and fix anything the camera got wrong. Each field shows where the value sits against its normal range as you type.
Six trained models screen the full panel. Anything flagged comes with a first test to order and what to do next.
What it screens for
Each screen is a gradient-boosted model trained on 18 CBC parameters. The output is a class, not a diagnosis.
| Screen | Possible results |
|---|---|
| Anemia | NoneNormocyticMicrocyticMacrocytic |
| Leukocytosis | NormalNeutrophiliaLymphocytosisMonocytosisEosinophiliaBasophiliaMixed |
| Leukopenia | NoneNeutropeniaLymphopeniaLeukopeniaCombinations |
| Thrombocytopenia | NormalMildModerateSevere |
| Thrombocytosis | NegativePositive |
| Polycythemia | NegativePositive |
The team
Faculty advisor
Senior Lecturer, Department of Electrical and Computer Engineering. Machine learning & AI.
Data scientist
Data preprocessing and analysis. Django, Python.
Web developer
Interface design and front-end development. Django, JavaScript, HTML/CSS.
ML developer
Model training and deployment. Python, scikit-learn, XGBoost.