edisonbatucan/specialproblem

Repo for our special problem 1 and 2

Special Problem (Repo for our special problem 1 and 2)

Dataset

From SPIE-AAPM-NCI PROSTATEx Challenges (PROSTATEx)

• https://wiki.cancerimagingarchive.net/pages/viewpage.action?pageId=23691656

Data Preparation

dicom to tiff conversion

pip install dicom2jpg

dicom_path = '../dicom'
output_path = '../out_dicom'
dicom2jpg.dicom2tiff(dicom_path, target_root=output_path, anonymous=False, multiprocessing=True)

Use Labelme Tool

 Save .json mask in ./json folder

Run labelme_to_json_batch.py
 Get mask -> Run tiff_convert.py

Data Partition :   (80 : 20, 'Train' : 'Validation')

Read Data

Pre-processing

Run dataprep.py

Prostate : Read -> Resize -> Histogram Equalization -> save to .npy

Mask : Read -> Resize and Grayscale -> save to .npy

Model Architecture

Model Accuracy and Loss

Predictions :

References

• https://pyimagesearch.com/2022/02/21/u-net-image-segmentation-in-keras/
• https://github.com/jocicmarko/ultrasound-nerve-segmentation.git
• https://towardsdatascience.com/metrics-to-evaluate-your-semantic-segmentation-model-6bcb99639aa2#