Software Lab

The Software lab mainly focusses on PET, SPECT, MRI and CT imaging. We have gained experience in advanced image reconstruction, system simulation and multimodal image analysis.

Current research

Partial volume effect correction in PET imaging

The partial volume effect is the deterioration of the image quality in a medical scan due to the limited resolution of the imaging system and the sampling of the image in discrete voxels. This results in a blurring of the image, and small structures showing a lower activity. Our group implements and develops post-reconstruction correction methods for this effect. This is done using information from the PET-scan itself (iterative deconvolution), or using anatomical information from a coregistered MRI-scan.

Figure 1: Phantom study for activity recovery in PET scans. Left: virtual PET scan; middle: PET-based correction (iterative deconvolution); right: MRI-based correction (Rousset)

Contact: Stijn Bonte

Go to top

Quantitative imaging for automated brain tumour diagnosis

In Belgium, 800 people are diagnosed with a primary brain tumour each year. Their life expectancy depends strongly on the tumour type, and ranges from several decades for low-grade tumours to only about one year for the most malignant types. An accurate pathological diagnosis is therefore of primary importance. In clinical practice the workflow from initial consultation to therapy involves many steps within a multidisciplinary team of radiologists, neurosurgeons, neuropathologists, nuclear medicine professionals, radiotherapists, etc. In each step of this pipeline, there is a degree of subjectivity involved in the decision process. However, in literature there is growing evidence that a quantitative, and thus objective, approach can yield a correct diagnosis and therapy prediction. The goal of this project is therefore to investigate how a quantitative analysis of medical images (both MRI and PET) can contribute to a personalised medicine. In this project, we will implement a method for automated brain tumour diagnosis, using textural analysis in combination with other quantitative tumour parameters, such as volume, shape and histogram features. Using advanced classification algorithms, such as Support Vector Machines or Convolutional Neural Networks, an algorithm for automated diagnosis will be implemented.

Figure 1: magnetic resonance imaging of a brain tumour with different sequences (T1-weighted, T2-weighted, MPRAGE, FLAIR)

Contact: Stijn Bonte

Go to top

PET attenuation correction with a supplemental transmission source

To obtain quantitative PET images, on needs to correct for the number of photons that were absorbed in the patient’s body. This so-called attenuation correction (AC) is in clinical settings mostly done by subsequently acquiring a CT scan. A simple bilinear scaling can convert CT Hounsfield units into PET linear attenuation coefficients. However, recently PET/MRI scanners have gained popularity while there exists no such relation between MR intensities and photon attenuation. Therefore, MR-based attenuation correction (MRAC) is a lot harder and not yet optimized for whole body PET imaging. The addition of a transmission source with a PET tracer can aid to find the attenuation correction values while keeping the dose to the patient low.
The research at Medisip lab focuses on the minimally required design for a transmission source. This is investigated in a theoretical way as well as with computer simulations. Future research will include real measurements with the theoretically optimal design.

Figure: Design of a transmission source that – in theory – allows to fully reconstruct the AC map.

Contact: Ester D’Hoe

Go to top

PET image reconstruction software QETIR

QETIR, which stands for Quantitative Emission Tomography Iterative Reconstruction is image reconstruction software for PET developed at Medisip. It utilizes listmode reconstruction as opposed to the more commonly used sinogram reconstruction. Implemented algorithms include (Time-of-Flight) MLEM, OSEM, MLTR, MLAA and MLAA+. Data from any scanner can be reconstructed if the scanner geometry is provided. Researchers interested in QETIR software can contact MEDISIP lab.

Figure: Example of reconstructed PET image by the QETIR software

Contact: Ester D’Hoe

Go to top

Joint reconstruction of activity and attenuation in listmode TOF PET

Joint reconstruction algorithms such as MLAA are capable to reconstruct both a PET image and the attenuation correction values (AC values) from Time-of-Flight PET data only. Although the inverse problem is ill-posed, the redundancy of the Time-of-Flight data allows to find AC values up to a constant scale factor. Current Medisip research focusses on listmode implementation of these algorithms and their extension to the case of prior information and to the case with background (scatter and random events).

Figure: Listmode-MLAA algorithm with background.

Contact: Ester D’Hoe

Go to top

Positron range and acollinearity modeling for hybrid PET/CT and PET/MR imaging

The development of integrated devices to combine Positron Emission Tomography (PET) and Magnetic Resonance (MR) imaging has been a constant trend in multimodal imaging for the last years. However, a number of approaches have shown that positron range and the acollinearity are fundamental limitations of PET image resolution. According to these studies, its effect is especially relevant with some radionuclides currently used in clinical and preclinical studies such as 82Rb, 124I and 68Ga. The aim of this project is threefold: 1 – Modeling the positron range and acollinearity of clinically relevant PET isotopes for different tissue types with and without the presence of a magnetic field. 2 – Validation of the positron model using phantom measurements with different isotopes and (pre)clinical PET/MRI and PET/CT systems. 3 – Use of the attenuation map derived from CT, MRI or PET only data to account for tissue dependent positron range during reconstruction.

Contact: Paulo Caribé

Go to top

Internal radiotherapy dosimetry

Internal vectorised radiotherapy are promising therapeutic modalities, increasingly used in the treatment of unresectable tumors. Optimization of the therapeutic outcome by maximizing the dose delivered to the tumor, while keeping the dose delivered to organs at risk at acceptable levels is the main goal of dosimetry. Systematically implemented dosimetry is essential in order to fully exploit the therapeutic potential of internal vectorised radiotherapy. Furthermore, accurate and precise dosimetric estimations are crucial for establishing a correlation between absorbed dose and treatment efficacy and safety. However, accuracy and precision in dosimetry may be hampered by potential systematic and stochastic errors respectively. The aim of this project is to assess the potential errors committed on the dosimetric calculations and to evaluate their biological impact on tumor dose-response and organ dose-toxicity correlation.

Contact: Gwennaëlle Marin

Go to top

Quantitative conductivity and permittivity imaging using magnetic resonance electrical properties tomogragphy (MREPT)

The frequency-dependent electrical properties (Conductivity and Permittivity) of biological tissue provide important diagnostic information, e.g., tumor characterization (usually the electrical properties of tumor will be higher than that of normal tissues) and also play an important role in estimating the local Specific Absorption Rate (SAR). Magnetic Resonance Electrical Properties Tomography (MREPT) is a method to map the conductivity and permittivity of the tissue using MRI. RF coil of a standard MR system is sufficient for EPT, whereas other methods uses separate electrodes, currents or RF probes for determining the electrical properties of the tissue. This makes MREPT a very promising technique to obtain the conductivity and permittivity in vivo without the need of additional hardware. We will be investigating the non-invasive measurements of the dielectric parameters using MREPT and the implementation of the sequence on the MRI scanner.

Contact: Prakash Vasudevan

Go to top

Past research

PET/MRI attenuation correction

Accurate attenuation correction is necessary to obtain quantitative PET images. In clinical practice, attenuation maps are usually derived from a low-dose CT scan or a PET transmission scan using a rotating source. In combined PET/MRI these methods will most likely not be used. In our group two approaches are investigated to derive the attenuation coefficient at 511 keV. The first approach uses special MR sequences (UTE) in combination with segmentation techniques to derive an attenuation map. In another appraoch we try to derive the attenuation map from a transmission scan using an annulus shaped transmission source inside the Field-of-View (FOV) of the PET scanner.

Figure 1 : MR-based attenuation correction using
Ultrashort Echo Time Sequences
Figure 2 : Transmission based attenuation correction>

Contact: , , Ester D’Hoe

Go to top

PET system simulation

System simulations can help to determine the influence of PET design parameters such as spatial resolution, TOF resolution, energy resolution, and detection efficiency. In this project the simulations are performed using GATE, a standard Monte Carlo simulation tool for emission tomography. In particular the performance of the PET system suitable to operate in an MR environment is investigated

Figure 1 : The scanner geometry of the simulated systems. Figure 2 : Influence of different design decisions on PET performance.

Contact: , Ester D’Hoe, Ekaterina Mikhaylova

Go to top

Innovative Collimator for Compact Stationary SPECT

In Single Photon Emission Computed Tomography (SPECT) scanners, collimators are used in front of the detectors to help determine the direction from which gamma-rays originate. In order to reconstruct a 3D image of the radioisotope distribution, we need to measure radiation coming from the patient at different angles, and for this there are two main strategies: rotating parallel-hole collimators and stationary pinhole collimators.

We came up with and developed a new collimator concept, that marries the two approaches: a stationary parallel-hole collimator. Because it is stationary, it should be considerably cheaper, less time-consuming and provide better quality images than non-stationary SPECT scanners. On the other hand, because of its innovative design based on parallel-holes, we can image the same volume as other currently available stationary SPECT systems using a much more compact scanner, therefore saving both space and material costs.

Figure 1: Rotating Parallel-hole Collimator   Figure 2: Stationary Pinhole Collimator   Figure 3: Our New Compact Stationary Parallel-hole Collimator


Go to top

SPECT System Modeling and Simulations

In any Single Photon Emission Computed Tomography (SPECT) scanner, an accurate model of its response is crucial to correctly predict its performance and to get the best reconstruction outcome. This is why we work on improving current methods for analytical modeling of SPECT systems, in particular for the geometrical collimator response in the innovative scanner designs which we develop in our group, which can have quite complex geometries.

On the other hand, we also use statistical software packages (Monte Carlo-based simulation software GATE) that allow us to perform realistic simulations of our system. It is more difficult to use these for the reconstruction itself, but we use them to produce realistic projection data to reconstruct and validate our analytical modeling.

Figure 1: System modeling using the Monte Carlo-based simulation software GATE Figure 2: (a) A point projection using our analytical system modeling. (b) A point projection using the GATE simulation software. (c) A line profile through the dashed line in both the analytical (blue) and GATE (red), where we can see that the two match quite well.


Go to top

The new paediatric TOF-PET: a simulation study

Children are the most challenging and vulnerable nuclear medicine patients due to great differences in their height, weight and physical development. They are more sensitive to consequences of the radiation dose expose than adults and have a higher risk to develop metastatic disease. In order to avoid or at least significantly reduce any risk during examinations, pediatric patients need a suitable diagnostic system that is as fast, as safe and as accurate as possible. Nowadays, there is no a commercial medical imaging modality especially dedicated for children.

The project aims to evaluate performance of a new pediatric TOF-PET scanner that combines the most recent achievements in radiation detectors, electronics and data processing methods. It is based on monolithic LYSO scintillators coupled to recently developed digital SiPM photodetectors in dual-sided configuration. A distinctive feature of the new pediatric PET is its 150 ps TOF resolution which allows to obtain images of low noise and high contrast. The novel pediatric PET aims to make examinations of children extremely fast, safe and precise.

The figures below show the recent simulation study results performed with voxelized phantoms of children of different age. The results show that the new paediatric PET is able to obtain high quality images in short acquisition times and with ultra-low doses (10-fold reduction).

Figure 1: The new paediatric PET with
a 10-year-old boy phantom.
Figure 2: Simulation of a scan of a 10-year-old boy
with tumor in lung and liver. Total activity is 4 MBq,
the acquisition time is 12 min.
Figure 3: Simulation of a scan of a 1-year-old boy
with focal hyperinsulinism. Total activity is 3 MBq,
the acquisition time is 12 min.
Figure 4: Simulation of a brain scan of a 10-year-old boy.
Total activity in the brain is 3.7 MBq,
the acquisition time is 12 min.

Contact: Ekaterina Mikhaylova

Go to top

Iterative CT reconstruction

Reducing the image noise in in-vivo micro Computerized Tomography (μCT) is important to improve the image quality and maintain diagnostic confidence. Theoretically, the images can be improved by using longer acquisition times. However, because of the limitations on administrated dose and anesthetics other approaches need to be considered. The goal of this work is to find regularization for sparse-view reconstruction and image denoising in general.

Figure 1: Comparison between differenant regularization schemes used in sparse view CT reconstruction


Go to top

SPECT/MRI compatibility

In the last 10 years, there has been a growing interest in multimodality imaging. Integrating different modalities in a single device is advantageous for several reasons: improved patient’s comfort, better diagnosis, and more efficient hospital logistics. Single Photon Emission Tomography (SPECT) and Magnetic Resonance Imaging (MRI) are two complementary modalities: SPECT is a functional imaging technique (Fig. 1a) while MRI is structural (Fig. 1c). A SPECT device consists of two main components: a detector and a collimator. We focus on the MRI-compatibility of SPECT collimators and more specifically 3D-printed tungsten collimators.

Figure 1: SPECT(a), SPECT/MRI(b), MR(c) Figure 2: 3D-printed collimator next to 7T-MRI coil

Contact: Karen Van Audenhaege

Go to top

Brain SPECT Imaging

Single Photon Emission Computed Tomography (SPECT) is a nuclear imaging technique often used for functional cerebral imaging. Current whole-body systems have shown to be clinically useful in a wide range of neurological diseases like dementia, cerebrovascular diseases and epilepsy. However, a dedicated brain SPECT system with a better sensitivity and resolution could result in better diagnosis and more effective treatment. A SPECT device consists of two main components: a detector and a collimator. We focus on innovative collimator design and optimized a multi-pinhole collimator insert for an existing ring of LaBr3 detectors.

Figure 1: SPECT device with detector and collimator   Figure 2: SPECT spatial resolution phantom   Figure 3: SPECT brain study

Contact: Karen Van Audenhaege

Go to top

CT/MR phantoms for quality assessment

Magnetic Resonance Imaging (MRI) is a radiation-free imaging technique. Therefore, when possible, should be preferred over other imaging modalities. However, Computed Tomography (CT) is still the gold standard in some organ imaging (lung, bone..) more challenging for MRI, and new MRI techniques should be validated against CT. We investigate the development of realistic CT/MR phantoms for quality assessment.

Figure 1: (a) MR (T2-weighted) and (b) CT image of lung phantom

Contact: Shandra Gutierrez Diaz

Go to top

Detector design in Hadron Therapy

Hadron therapy offers some advantages over conventional therapy. However, small errors in particle range may significantly increase dose delivered to healthy tissue or produce incomplete coverage of the target volume. Consequently, developing a method to verify the delivered dose and the Bragg peak location during treatment is crucial. Recent investigations have shown that the prompt gamma photons produced during the patient irradiation with hadron beams are in good correlation with the dose delivered to the patient and therefore, they can be used for in-beam range monitoring. Slit collimated gamma camera has been proposed as a system for prompt gamma imaging. Image reconstruction can potentially improve the accuracy of that system. The purpose of this work is to study and implement a reconstruction process for in-beam range monitoring using slit-collimated gamma camera. Obtained results will guide the design of the slit-collimated camera. The results indicate that the range monitoring is possible with slit camera using ML-EM reconstruction. The average deviation obtained using reconstructed profiles is 2.06 mm.

Figure 1: Proton beam (blue), PMMA phantom (yellow),
slit collimator (grey) and LYSO scintillator (red).
Figure 2: The falloff position measured on the reconstructed
profiles versus actual shift.
The green line shows the ideal case.

Contact: Faruk Diblen

Go to top