Visualizing and Quantifying Plant Root Distribution in Soils: A CT Scan and Machine Learning Approach

Environment Setup

Using Conda

1. Create a new conda environment:

conda create -n soilcores python=3.10

2. Activate the environment:

conda activate soilcores

3. Install required packages:

pip install -r requirements.txt

Using the GUI to Segment and Analyze Soil Core CT Scans

This section describes how to use the Soilcore NIfTI Viewer graphical interface to load, segment, threshold, and analyze 3D CT scans for root topology. We illustrate this with the sample file C1216.nii.gz.

---

Step 1: Upload the CT Scan Raw Data

Click on "Load NIfTI File" and choose a .nii or .nii.gz file from your dataset (e.g., C1216.nii.gz).

---

Step 2: Select the Soil Core File

From the file browser, locate your CT scan file (e.g., C1216.nii.gz in the folder test/) and click Open.

---

Step 3: Visualize the Raw Data

Once loaded, the middle slice of the scan will be shown. Use the "Axis" slider to change the orientation and the bottom slider to scroll through slices.

---

Step 4: Choose a Segmentation Model

Select one of the available models in the Segmentation Model dropdown. For example, choose UNETR (AdamW, 100k, 2 heads).

Click "Run Segmentation" to start. A message will appear once the segmentation is complete.

---

Step 5: Apply a Threshold

Adjust the Lower Threshold and Upper Threshold sliders to filter out unwanted noise or background.

---

Step 6: Choose a Path to Save Thresholded Images

Click "Save Thresholded Images", then choose the directory where the binary PNG slices will be stored. A confirmation will appear once the images are saved.

---

Step 7: Select Folder with Saved Thresholded Images

Use the "Select Single Folder" or "Select Multiple Folders" buttons under Root Topology Analysis to load the folder(s) with the thresholded images.

---

Step 8: Get Root Topology Results

Click "Analyze Root Topology". A bar chart will appear showing the total root length for different diameter ranges. CSV and PNG files are also saved in the output directory.