liveBook · Manning

Series UID

As explained in Deep Learning with PyTorch, Second Edition

A unique DICOM identifier used to name and associate the files belonging to one CT scan series.

A series UID identifies one CT scan series and associates the samples or files that belong to it. In a dataset, it can also act as a practical filter for selecting one scan. section 12.5.2 section 13.7.1

Filtering a dataset to one scan

Pass a truthy series_uid to LunaDataset, then retain only candidate records whose series_uid matches the requested value. This makes a single problematic CT scan easier to visualize or debug. section 12.5.2

class LunaDataset(Dataset):
  def __init__(self,
         val_stride=0,
         isValSet_bool=None,
         series_uid=None,
      ):
    self.candidateInfo_list = copy.copy(getCandidateInfoList())     #1

    if series_uid:
      self.candidateInfo_list = [
        x for x in self.candidateInfo_list if x.series_uid == series_uid
      ]

The dataset constructor copies candidate information and filters it with a direct series_uid equality test. section 12.5.2

The identifier also provides a file-integrity check: each series should have one .mhd file and one .raw file, with directory counts matching expectations. section 13.7.1

How Deep Learning with PyTorch, Second Edition covers it

How deep it goes
Defines the identifier operationally through dataset filtering and file checks, without describing its generation or internal format.
Examples
Uses LunaDataset filtering, paired .mhd and .raw files, and shell listings that verify directory contents.
What you can do after
Diagnose/debug with it by isolating one CT scan and checking whether its expected files are present.
How it connects
Links CT scans with datasets, training sets, validation sets, and paired scan files.

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Deep Learning with PyTorch, Second Edition. The section links point to the parts of the book it draws on.