LazyLuna

Lazy Luna is software for Cardiovascular Magnetic Resonance Imaging (CMR) analysis, which provides a backend for coders and a GUI tool for multilevel reader comparison.

Table of Contents

• Description
• Requirements
• Project Overview
• First Use
• License

Description

Lazy Luna offers a number of base classes for multilevel reader comparison. Multilevel refers to comparisons on the image level (such as contour comparisons), on the patient level (such as differences between assessed clinical parameters) and the reader level (pertaining to statistical trends of reader differences).

Requirements

A specific environment for Lazy Luna is recommended. The environment should operate on Python 3.8 or higher with the following libraries installed:

• Pydicom 2.2.0 or higher
• Numpy 1.20.1
• Shapely 1.8.0 or higher
• Rasterio 1.2.10
• Pandas 1.2.4

The Graphical user interface requires:

• PyQt5 12.11.0
• Matplotlib 3.3.4
• Seaborn 0.11.1

Project Overview

Class Structure

Sequence or use-case specific classes can be inherited from these base classes in order to address specific or novel uses. The classes interact with each other in order to allow for these multilevel comparisons as described in the class diagram below. A more detailed elaboration of the software architecture is documented in this paper.

Graphical User Interface

Lazy Luna offers a GUI for reader and case comparison. This allows for loading selected cases from two readers, assessing their differences statistically and tracing them to contour origins.

First Use

As described in the paper, a first user experience with Lazy Luna can be realized by comparing a neural network's segmentations of the EMIDEC image dataset to EMIDEC's gold standard contours.

1. Install an environment for Lazy Luna
   1. Install Lazy Luna
   2. Install LL's requirements
2. Download the EMIDEC training Dataset from here
   1. Unzip the dataset into a folder as in figure below
3. Prepare Lazy Luna Data Folder
   1. Zipped annotations for the UNet are automatically downloaded with Lazy Luna's code in /docs.
      1. Copy "case_slice_to_sops.json" in /docs into the directory as in figure below
      2. Unzip "Emidec_AI" in /docs into the directory as in figure below
4. Run the Emidec_Converter jupyter Notebook in /Notebooks
   1. This converts EMIDEC's Nifti images into DICOM format and the gold standard and AI contours into Lazy Luna Annotation format
5. Open the Analyzer Tool

License

If the code was helpful to you please cite my paper.