a b/README.md 

  1
  
<h1>About Dataset</h1>





  
  

  2
  





  
  

  3
  
<h2>Abstract</h2>





  
  

  4
  
<p>





  
  

  5
  
This dataset comprises FDG-PET/CT and radiotherapy planning CT imaging data of <strong>298 patients</strong> from four institutions in Québec, diagnosed with histologically proven head-and-neck cancer (H&amp;N). The dataset includes pre-treatment FDG-PET/CT scans conducted between April 2006 and November 2014. A median interval of <strong>18 days</strong> (range: 6–66 days) was maintained before treatment. For de-identification, dates in the TCIA images were altered while preserving the time intervals between serial scans.





  
  

  6
  
</p>





  
  

  7
  





  
  

  8
  
<h2>Data Description</h2>





  
  

  9
  
<p>





  
  

  10
  
This subset focuses exclusively on the <strong>CT scans</strong> from the original dataset. It provides valuable insights for researchers studying head-and-neck cancer, offering a comprehensive collection of imaging data crucial for understanding tumour characteristics and treatment outcomes.





  
  

  11
  
</p>





  
  

  12
  





  
  

  13
  
<h2>Overview of Processing Workflow</h2>





  
  

  14
  
<p>This processing workflow is designed to handle CT scan images in DICOM format. It consists of two major phases:</p>





  
  

  15
  





  
  

  16
  
<h3>Phase 1: Normalization and Conversion (DICOM to PNG)</h3>





  
  

  17
  
<ul>





  
  

  18
  
  <li><strong>Global Min-Max Calculation:</strong> Calculates the global minimum and maximum pixel values across all DICOM files for standardized normalization.</li>





  
  

  19
  
</ul>





  
  

  20
  





  
  

  21
  
<h4>DICOM Image Processing:</h4>





  
  

  22
  
<ul>





  
  

  23
  
  <li>Each DICOM file is normalized using the global min-max values.</li>





  
  

  24
  
  <li><strong>CLAHE</strong> (Contrast Limited Adaptive Histogram Equalization) is applied to enhance contrast.</li>





  
  

  25
  
  <li><strong>Histogram stretching</strong> is used for further contrast enhancement.</li>





  
  

  26
  
</ul>





  
  

  27
  





  
  

  28
  
<h4>Conversion to PNG:</h4>





  
  

  29
  
<ul>





  
  

  30
  
  <li>Processed images are converted into <strong>PNG format</strong> for easier handling.</li>





  
  

  31
  
  <li>Each image is uniquely named based on its slice index.</li>





  
  

  32
  
</ul>





  
  

  33
  





  
  

  34
  
<h3>Phase 2: Upscaling and Additional Image Processing (PNG Images)</h3>





  
  

  35
  





  
  

  36
  
<h4>Image Upscaling:</h4>





  
  

  37
  
<ul>





  
  

  38
  
  <li>Each PNG is upscaled to a higher resolution (e.g., <strong>2048×2048 pixels</strong>) using <strong>Lanczos resampling</strong>.</li>





  
  

  39
  
</ul>





  
  

  40
  





  
  

  41
  
<h4>Additional Image Processing:</h4>





  
  

  42
  
<ul>





  
  

  43
  
  <li><strong>Gaussian blur</strong> is applied to reduce noise and artifacts.</li>





  
  

  44
  
  <li><strong>Antialiasing</strong> is performed to improve visual clarity.</li>





  
  

  45
  
</ul>





  
  

  46
  





  
  

  47
  
<h2>Publication Citation</h2>





  
  

  48
  
<p>





  
  

  49
  
Vallières, M. et al. <em>Radiomics strategies for risk assessment of tumour failure in head-and-neck cancer.</em> Sci Rep 7, 10117 (2017). DOI: <a href="https://doi.org/10.1038/s41598-017-10371-5" target="_blank">10.1038/s41598-017-10371-5</a>





  
  

  50
  
</p>





  
  

  51
  





  
  

  52
  
<h2>License</h2>





  
  

  53
  
<p>Creative Commons Attribution 3.0 Unported License</p>





  
  

  54
  





  
  

  55
  
<h2>More Info</h2>





  
  

  56
  
<p>Dataset URL: <a href="https://wiki.cancerimagingarchive.net/display/Public/Head-Neck-PET-CT" target="_blank">https://wiki.cancerimagingarchive.net/display/Public/Head-Neck-PET-CT</a></p>