Breaking tasks down

[ditwrd]

Data Preprocessing

Good to have:

• Add git hook for pytest
• Add CI for pytest (and pylint?)
• Better file naming to make it sort from lowest (001,002,...,xxx)

Notes: Save to hdf5 for each step

Gettting Segmentation mask

1. Convert plist to python dict + test
2. Clean dict (remove zero value, remove unused data, floating to integer convertion)+ test
3. Split feature for binary and multiclass + test
4. Make binary and multiclass feature into segmentation mask for all patient data+ test

Getting CT Scan Image

1. Get CT Scan image
2. Get CT Scan HU representation

TFRecords

1. Make TF Records per patient (Multiclass feature, binary feature and image)

Addons

1. Calculate agatston score + test

[ditwrd]

Model

Model - UNet++ Base Model Model - MobileNetV2 Blocks Model - Custom Loss (Log Cosh Dice Loss)

Postprocessing

Postprocessing - other algorithm + Connected Component Labelling Postprocessing - Agatston Scoring Algorithm Postprocessing - Summarize result