Semi-Quantitative DCE-MRI parameters estimation

SeQ-DCEMRI Extension for 3D Slicer

Overview • Key Features • Installation and Setup • User Guide • Future Work • Acknowledgments • License • References

TL;DR

Jump straight into the User Guide

Overview

SeQ-DCEMRI is a slicer extension created to derive semi-quantitative parametric maps from signal intensity analysis of Dynamic Contrast-Enhanced Magnetic Resonance Imaging (DCE-MRI) datasets.

Semi-quantitative DCE-MRI refers to the analysis of DCE-MRI data using metrics that summarize the enhancement patterns of tissues over time after contrast injection. It focuses on relative signal changes, such as the rate and extent of contrast uptake and washout, rather than pharmacokinetic models derived from $$T_1$$ quantitative measurements. Common parameters include Peak Enhancement (PE), which measures the maximum signal increase with respect to pre-contrast baseline signal levels, Signal Enhancement Ratio (SER), which compares the signal intensity at different time points to infer tissue vascularity and permeability, and Functional Tumour Volume (FTV), which refers to the volume of tumour tissue that shows specific patterns of contrast enhancement, typically associated with active tumour regions. The semi-quantitative approach offers a practical and less complex way to evaluate tissue behaviour, particularly in oncology, without requiring detailed pharmacokinetic (PK) modelling [1]. However, this can also be combined with PK modelling and other imaging modalities, to explore a more robust radiomic approach.

Key Features

The SeQ-DCEMRI Slicer extension is based on the three time-point (3TP) analysis method to calculate the FTV [2,3]. It offers flexibility in selecting the pre-contrast, early and late post-contrast time points, and allows for optimising FTV measurements by modifying the PE and SER thresholds[4].

SeQ-DCEMRI uses the Slicer Sequences module to manage 4D datasets. It can process any DCEMRI dataset that can be loaded, or combined into a sequence. If the Sequence Registration module is installed, it also gives the option to use it to register the dataset prior to the analysis.

In terms of visualisation, SeQ-DCEMRI allows to scroll through the image time-point, display the image subtraction between two time-points, toggle the markup box and segmentation mask, as well as allows to select between a defaul hanging protocol and 3D render view.

By embedding the Segment Editor into the SeQ-DCEMRI GUI, the user can define a precise region-of-interest (ROI) covering the tumour, or any other tissue of interest, within the markup box.

The sliders to select the background image intensity (pre-contrast), PE and SER thresholds allow for testing and optimising FTV measurements.

The output maps: 3D Maximum Intensity Projection (MIP), Percentage of Enhancement and Signal Enhancement Ratio are combined into a sequence (OutputSequence), that can be managed by the Slicer Sequences module.

The FTV is reported in the form of a table and a colour-labelled image overlaid to the MIP volume.

Comparison with Breast DCE-MRI FTV

The algorithms used by SeQ-DCEMRI are equivalent to those used by the Breast DCE-MRI FTV Slicer extension. Although we don't provide the option to add omit regions, our extension offers the ability to manually delineate an ROI to more precisely derive the FTV parameters. Like the Breast DCE-MRI FTV, the background signal threshold is calculated over the markup box, but the PE, SER and FTV parameters are calculated over the segment mask drawn as ROI (see the User Guide for details).

Installation and Setup

The earliest 3D Slicer version compatible with SeQ-DCEMRI is 5.6.1. It has been tested with versions 5.6.1, 5.6.2 (latest stable release r32448) on MacOS (Sonoma 14.6.1) and Windows 11, and 5.7.0 (preview release r32969) on MacOS (Sonoma 14.6.1).

Currently, this extension is under development and is not yet available in the Slicer Extensions Catalogue, hence it has to be installed manually in testing mode by following the instructions here

User Guide

1. Ensure the extension is installed by following the Installation and Setup instructions.
2. Add the sample dataset to the local database. The sample data contains a synthetic FTV phantom (work in progress):
   1. Download the sample datasets from here
   2. Add the sample data to the DICOM database
   3. Load the image data into the scene

2. Process the sequence data

   1. Open the SeQ-DCEMRI GUI, if the input data is a sequence, it should appear in the drop-down list of the Inputs section
   2. If you want to register the data prior to the analysis, click the Register Sequence to open the Sequence Registration module^{1 2}
   3. Define the Pre-Contrast, Early Post-Contrast and Late Post-Contrast time points from the sliders in the Identify relevant timepoints. These values define the 3 time-points (3TP) parameters required by the algorithm to calculate the PE and SER maps[1]
   4. When processing real data, it may be useful to display the subtraction image between the pre-contrast and any of the post-contrast images, which can be done in the Display Subtraction Image Volumes
   5. By default, the software creates an empty segment mask (Segment_1), which can be modified manually by using the effects available in the Segment Editor.
      1. If the default mask is not modified, the software creates an ROI that matches the markup box.
      2. If multiple segmentations are created, you can select which one to use for the analysis in the drop-down list in the section Parameters
      3. Only one segmentation mask is used for analysis
   6. Select an image Background threshold to cut out pixel values in the 95th percentile that are below that threshold. This is intended to eliminate very low $T_1$ baseline values. This threshold is applied over the markup box in the pre-contrast image.
   7. Select the PE threshold to cut out PE values below that. PE is calculated from the pre-contrast and early post-contrast images (selected in the Identify relevant timepoints section) as follows¹:

   $$PE[\%] = 100*{S(t_{EARLY\ post-contrast}) - S(t_{PRE-contrast}) \over S(t_{PRE-contrast})}$$

   8. Select whether to use a pre-defined SER range or a single value. The pre-defined range is consistent with the values used by Li et al. [5]. By adding the option to select a single threshold value, it is possible to iterate over the results to find an optimum FTV, as reported by [4]. When selecting a single threshold, $$SER_{THRESH}$$, 3 intervals are defined. $$0 < SER ≤ SER_{THRESH}$$; is calculated from the 3TP images: pre-contrast, early and late post-contrast as follows¹:

   $$SER = {S(t_{EARLY\ post-contrast}) - S(t_{PRE-contrast}) \over S(t_{LATE\ post-contrast}) - S(t_{PRE-contrast})}$$

   9. Once all the parameters are set, click the button Click to Process to run the analysis.

3. The results can be reviewed by scrolling through the slices or time-points. The analysis can also be re-run by modifying any parameters in the Parameters section.

Limitations and future extensions

• To avoid duplicated extensions, we suggest merging this one with the existing Breast DCE-MRI FTV module. One of the main advantages of doing that would be the ability to generate PDF reports with the results while allowing flexibility in the input data.
• Currently, there are conflicts with the segmentation masks and markup boxes when processing multiple datasets. The current workaround is to close the scene and/or restart 3D Slicer to process a different dataset.
• When processing the registered dataset, the timing information is lost, so there is no consistency between the pre-registered (raw) image time points and the registered time points, as it gets assigned the default 1.2min sampling time. This can be solved by transferring the time information when registering the data.

Help and Support

We'll be more than happy to get any feedback, so please feel free to create an issue to let us know any comments/questions/bugs/etc. about this extension.

Acknowledgments

This project has been supported by ...

License Information

This project is licensed under the terms of the Slicer License

References

[1] Hylton, Nola M. and Blume, Jeffrey D. and Bernreuter, Wanda K. and Pisano, Etta D. and Rosen, Mark A. and Morris, Elizabeth A. and Weatherall, Paul T. and Lehman, Constance D. and Newstead, Gillian M. and Polin, Sandra and Marques, Helga S. and Esserman, Laura J. and Schnall, Mitchell D. (2012). Locally advanced breast cancer: MR imaging for prediction of response to neoadjuvant chemotherapy—results from ACRIN 6657/I-SPY TRIAL. Radiology, 263(3), 663-672.

[2] Degani, Hadassa and Gusis, Vadim and Weinstein, Daphna and Fields, Scott and Strano, Shalom (1997). Mapping pathophysiological features of breast tumors by MRI at high spatial resolution. Nature Medicine, 3(7), 780-782.

[3] Furman-Haran, Edna and Degani, Hadassa (2002). Parametric Analysis of Breast MRI. Journal of Computer Assisted Tomography, 26(3), 376-386.

[4] Musall, Benjamin C. and Abdelhafez, Abeer H. and Adrada, Beatriz E. and Candelaria, Rosalind P. and Mohamed, Rania M.M. and Boge, Medine and Le-Petross, Huong and Arribas, Elsa and Lane, Deanna L. and Spak, David A. and Leung, Jessica W.T. and Hwang, Ken-Pin and Son, Jong Bum and Elshafeey, Nabil A. and Mahmoud, Hagar S. and Wei, Peng and Sun, Jia and Zhang, Shu and White, Jason B. and Ravenberg, Elizabeth E. and Litton, Jennifer K. and Damodaran, Senthil and Thompson, Alastair M. and Moulder, Stacy L. and Yang, Wei T. and Pagel, Mark D. and Rauch, Gaiane M. and Ma, Jingfei (2021). Functional Tumor Volume by Fast Dynamic Contrast-Enhanced MRI for Predicting Neoadjuvant Systemic Therapy Response in Triple-Negative Breast Cancer. Journal of Magnetic Resonance Imaging, 54(1), 251-260.

[5] Li, Wen and Arasu, Vignesh and Newitt, David C. and Jones, Ella F. and Wilmes, Lisa and Gibbs, Jessica and Kornak, John and Joe, Bonnie N. and Esserman, Laura J. and Hylton, Nola M. (2016). Effect of MR Imaging Contrast Thresholds on Prediction of Neoadjuvant Chemotherapy Response in Breast Can- cer Subtypes: A Subgroup Analysis of the ACRIN 6657/I-SPY 1 TRIAL. Tomography, 2(4):378–387.

Footnotes

Footnotes

1. This requires the module to be already installed, if not, follow the instructions here ↩ ↩² ↩³

2. This will take you out of the SeQ-DCEMRI module, to come back, follow the instructions in the warning window when clicking Register Sequence ↩