Predicting Obstructive Sleep Apnea syndrome (OSA) based on Computerized Tomography (CT)

Input CT Dicom files

Choose a [zip] file for DICOM files of the CT or Drag it here.

I agree to the terms and conditions.

Input Clinical Information

Age (years):

(At the time of CT scan)

Height (cm):

Weight (kg):

Sex:

Prediction:

Start index of airway:

(Superior border of the hard palate)

End index of airway:

(Inferior end of epiglottis)

Instructions guide

Select the number of classes for your classification task.
For predicting significant OSA (two-class classification), the output is “Significant OSA” or “No significant OSA”.
For predicting OSA severity (four-class classification), the output is “Normal”, “Mild OSA”, “Moderate OSA”, or “Severe OSA”.
Choose the number of bits for the analysis, which will determine the precision of your computation.
Enter the start index of the airway (superior border of the hard palate) where the analysis should begin.
Enter the end index of the airway (inferior end of epiglottis) where the analysis should end.
The name of DICOM files in zip file should end with index (four digits) of slices. e.g., 11223344_0001.dcm (index 1), 11223344_0002.dcm (index 2), 11223344_0003.dcm (index 3), ......., 11223344_0125.dcm (index 125), and files should be sequentially ordered (ascending order).
Drag and drop or choose a [.zip] file containing your CT scan data.
Review the uploaded file name to ensure the correct file has been selected.
Read and accept the terms and conditions by checking the checkbox.
Click the 'Analysis' button to submit your data for processing.
One the analysis is completed the results will be displayed on the screen.

Please follow these guidelines when uploading files:

The file must be in .zip, .gz, .tar, and .7z format.
The maximum file size is 100MB.

Precautions for use

The test can only be performed once every 30 second, and repeated requests may return previous results.
The algorithm for highlighting the airway is based on Otsu's Method and adaptive thresholding. Its primary goal is to enhance the performance of deep learning algorithms rather than produce an exact reconstruction of the airways. For more details, please refer to the paper.
The training phase of the model is as described in the paper, and it continues to undergo fine-tuning along with the acquisition of additional data.

For more information about the deep learning models used for predicting obstructive sleep apnea based on computed tomography scans, please refer to our research paper on GitHub.

Current version: 0.9.0 (2024-01-21)

Contact: Prof. Jinyoup Kim (kjyoup0622@gmail.com) for medical support and Prof. Hyoun-Joong Kong (gongcop7@snu.ac.kr) for technical support.