---

About

Algorithm Facts

Summary

This algorithm performs an Automatic Quality Assessment of lung CT scans. It provides both qualitative and quantitative tests to assess image quality. Furthermore, it checks whether the lung has been acquired completely by the scan.

Mechanism

Input¶

This algorithm expects an anonymized DICOM series of a lung CT scan as input. Relevant DICOM tags need to be available for the algorithm to work correctly:

• (0008, 0008) Image Type
• (0020, 0013) Instance Number
• (0028, 1052) Rescale Intercept
• (0028, 1053) Rescale Slope
• (0028, 1054) Rescale Type

Results¶

The output of the algorithm can roughly be subdivided into four categories:

1. Image Metadata

   • Contains descriptive information extracted from DICOM tags

   • Subdivided into Image Description, Image Volume Information, and CT Metadata

2. Qualitative QA Tests

   • Checks for volume integrity (e.g. no missing slices)

   • Checks if rescale information is available

   • Validates the image type (e.g., the image should not be a maximum intensity projection)

3. Organ Coverage Test: Checks whether the lung has been acquired completely.

4. Quantitative QA Tests

   • Several quantitative quality assessments, e.g.

     • Signal-to-noise ratio

     • A custom-developed quality index

     • Sharpness

     • Streakings

   • Noise levels per tissue class (we als report volumes for different tissue classes)

   • Image resolution: Comparison of nominal to intrinsic resolution

When inspecting your results using our integrated medical image viewer, you can get further information on these tests by reading the tooltips in the sidebar.

How to interpret the results¶

When you open the results in our integrated viewer, the sidebar will contain the results of the algorithm. Four each of the three test categories (qualitative QA, organ coverage, quantitative QA), indicators show whether the prerequisites for the tests are fulfilled and whether the overall results of the tests are positive or negative.

The quantitative test results (e.g. signal-to-noise ratio) are hard to interpret for a single image. The usefulness of these quantities lies in comparing different images against each other. For example, they can be used to monitor image quality for a certain scanner over time.

Uses and Directions

This algorithm was developed for research purposes only.

Warnings

⚠️ This algorithm is not meant for clinical use.