LSTAC -- Lesion Segmentation and TIC Analysis for CEUS

---

Implementation of A Deep Learning-based Framework for the Malignancy Analysis of Thyroid Lesions in Contrast-Enhanced Ultrasound Videos

Pre-requisites

---

• python >= 3.8
• torch >= 1.10

pip install -r requirements.txt

Usage

---

Segmentation

Download the yolov5-seg checkpoints from GitHub and put them in the segment folder. Datasets structure should be same as segmentation\datasets\thyroid-example.

cd segmentation
python segment/train.py --img 960 --batch 16 --device 0 --epochs 500 --cfg models/hub/yolov5s_modified.yaml --weights segment/yolov5s-seg.pt --data datasets/thyroid.yaml

Segment result:

Frame Extraction

Run frame_extact_all.ipynb to analysis all CEUS videos. Run merge_extracted_frames.ipynb to merge all extracted frames into the extract-key-frame folder.

Analysis result:

Classification

Make sure you have the extracted key frames in the extract-key-frame folder. extract-key-frame dataset folder structure should be like this:

extract-key-frame
├── cohort1
│   ├── all
│   │   ├── roi1
│   │   │   ├── 0
│   │   │   │   ├── imgs
│   │   │   ├── 1
│   │   │   │   ├── imgs
│   │   ├── target
│   │   │   ├── 0
│   │   │   │   ├── imgs
│   │   │   ├── 1
│   │   │   │   ├── imgs
│   ├── ultrasound
│   │   ├── same as 'all' folder
│   ├── contrast
│   │   ├── same as 'all' folder
├── cohort2
│   ├── ........

Classification training:

cd classification
# Argument details can be found in the `run.py` file.
python run.py --data ../extract-key-frame/ultrasound/ --save ./result/exp1 --model InceptionV3  --area roi1 --epoch 300 --lr 0.00001 --batch_size 32

Classification testing:

# Argument details can be found in the `test.py` file.
python test.py --data ../extract-key-frame  --path ./result