XClose

UCL Division of Medicine

Home
Menu

Medical Physics

The UCL Medical Physics Group undertakes basic research related to Nuclear Medicine and Multimodality Imaging.

This currently encompasses a range of projects including design of novel SPECT systems, development of new reconstruction algorithms and development of image analysis techniques that aim to reduce artefacts and improve quantification. The group is largely dependent on external funding through grants (EPSRC, EC, BRC) and industrial collaboration (Siemens, GE, GSK, National Physical Laboratory).

The group has joint research projects with the UCL Centre for Medical Image Computing with several individuals there now working on Nuclear Medicine related projects (along with Professor Simon Arridge). We attract many visiting scientists and organises regular seminars which are open to scientists in the London region.

Research group of the Institute of Nuclear Medicine Medical Physics group

Medical physics research

  1. INSERT: A new multi-modality SPECT/MRI system
  2. Respiratory motion correction
  3. Image reconstruction in emission tomography
  4. Kinetic Modelling
  1. Improved Quantification in the Lung
  2. Partial volume correction in neuroimaging
  3. TEXRAD Texture analysis: Quantifying Heterogeneity
  4. Open source software

INSERT: A new multi-modality SPECT/MRI system

The INSERT project (Integrated SPECT/MRI for Enhanced Stratification in Radio-chemotherapy) aims to create a new multi-modality imaging tool for simultaneous SPECT and MRI studies by the development of a custom SPECT insert for commercially available MRI scanners. It is funded by the EC under the FP7-HEALTH programme and involves a number of collaborating groups throughout Europe (see the EU Fact Sheet).

The system will be designed primarily for the study of brain tumours, mainly gliomas. The fundamental unit of the INSERT system is a compact gamma camera with high intrinsic resolution [ref. 1]. The principal task for our group is the detector-system and collimator design.

Whereas conventional SPECT systems consist of large rotating detectors, the INSERT system needs to fit inside the restricted space of the MRI bore. Also, the system should be stationary, as any detector motion would add complexity to the design and could interfere with the MRI signal. In relation to collimator design, a trade-off is always needed between sensitivity, resolution and FOV. For the current application, high sensitivity was considered more important than good spatial resolution.

The system optimisation is based on analytical calculations of sensitivity and resolution, as well as simulation and reconstruction of data corresponding to various geometric and anthropomorphic phantoms. For this purpose, we developed a novel projection algorithm based on angular blurring [ref. 2]. We have investigated the use of both multi-pinhole and multi-slit-slat collimators [refs. 3-6]. A sensitivity similar to or higher than the standard collimator could be obtained with both multi-pinhole and multi-slit-slat systems. The former gives better uniformity and trans-axial resolution, while the latter gives better axial resolution.

We propose a new type of slit-slat collimation for the INSERT system: the mini-slit-slat (MSS) collimator [ref. 7, fig. 1]. In contrast to a standard slit-slat collimator, in which the slit is continuous along the collimator, we explore the use of mini-slits to obtain improved angular sampling. In addition, we introduce the concept of interior slits, in which the slit aperture is located within the slat component, which thereby extend beyond the slit plane. If the resolution is fixed in both directions, the slat spacing can be increased, resulting in a sensitivity gain. The main advantage is allowing for minification without compromising the slat length and maximising the use of the space to optimise sensitivity.

Technical aspects of the system design are being addressed, such as collimator type [ref. 7], ring configuration [ref. 8], shielding [ref. 9], septal penetration, and geometrical calibration, and installation [fig. 1]. We have been using GATE to assess system performance. We are also currently performing experiments to compare mulit-pinhole and multi-slit-slat prototype collimators using a single detector module.

Reconstructions from simulated data for a multi-pinhole collimator

Multi-slit-slat collimator with 4 different phantoms (a uniform cylinder, a Derenzo phantom with rod-sizes of 6-11 mm, a Defrise phantom with parallel planes, and the Zubal brain phantom)

References
  1. Busca P, et al. (2014) Nucl. Instr. Meth. Phys. Res. A, 734:141-6.
  2. Bousse A, et al. (2013) Conf. Proc., Fully 3D reconstruction meeting.
  3. Erlandsson K, et al. (2013) Conf. Record,2013 IEEE Medical Imaging Conf.
  4. Salvado D, Erlandsson K, Bousse A, Occhipinti M, Busca P, Fiorini C, Hutton BF, "Collimator design for a clinical brain SPECT/MRI insert", Eur J Nucl Med Mol Imaging: Physics 1 (Suppl 1): A21, 2014.
  5. Erlandsson K, Busca P, Bukki T, Gola A, Salvado D, Butt A, Nemeth G, Piemonte C, Hutton B, Fiorini C, "Design considerations for INSERT: A new multi-modality SPECT/MRI system for preclinical and clinical imaging", Journal of Nuclear Medicine 55, 2014.
  6. Salvado D, Erlandsson K, Bousse A, Occhipinti M, Busca P, Fiorini C, Hutton BF, "Collimator Design for a Brain SPECT/MRI Insert", IEEE Transactions on Nuclear Science 62: 1716-1724, 2015.
  7. Salvado D, Erlandsson K, Bousse A, van Mullekom P, Hutton BF, "Novel Collimation for SPECT/MRI", IEEE Medical Imaging Conference, November 2014, Seattle, WA, USA.
  8. Erlandsson K, Salvado D, Bousse A, Hutton BF, "Evaluation of a Partial Ring Design for the INSERT SPECT/MRI System", Eur J Nucl Med Mol Imaging: Physics 2 (Suppl 1): A47, 2015.
  9. Salvado D, Erlandsson K, Hutton BF, "Shielding Requirements of a SPECT Insert for Installation in a PET/MRI System", IEEE Medical Imaging Conference, November 2015, San Diego, CA, USA.

Back to top


Practical Respiratory Motion Correction in Simultaneous PET/MR

Respiratory motion during PET acquisition leads to blurring in resulting images and quantification underestimation. Areas in the upper abdomen and thorax are adversely affected, especially lung/liver/pancreas lesions and the heart. In the case of lesions, PET images can provide information for monitoring of tumor response to therapy, diagnosis of malignancy via a standard uptake value (SUV) and calculation of lesion volumes for radiation planning, all of which are adversely affected by respiratory motion. We have several projects running related to respiratory motion detection and correction for PET/CT and PET/MR, with the two main projects listed below.

Data-driven extraction of a respiratory signal from raw PET data

The first step in correcting for respiratory motion is to obtain a signal related to the respiration. With this signal, the data can for instance be binned ("gated") into (nearly) motion free datasets prior to reconstruction by selecting the detected events with respect to the breathing phase. On current scanners the respiratory signal required to do so is obtained from an external device, whereas with data-driven (DD) methods it is directly obtained from the raw PET data, therefore avoiding the use of external equipment, which is expensive, needs prior setup and can cause patient discomfort.

Many DD methods have been developed over the years but we concentrate on using Principal Component Analysis (PCA) on dynamic PET sinograms of low spatial resolution [ref. 10].

This method is easy to implement and fast and has been shown to perform well compared to other methods in a preliminary study [ref. 11]. We are performing further evaluations, also in collaboration with the PET Center of Yale University.

However, many DD methods including PCA provide signals whose relationship with the physical motion is determined up to an arbitrary sign. This could cause inaccurate motion correction especially with multiple bed positions, where consistency between adjacent beds is needed, and attenuation correction artifacts in PET/CT.

We proposed new methods and compared them to a published registration-based method on FDG oncology patient data [ref. 12].

Respiratory motion corrected images through a pancreatic lesion

Respiratory motion correction line profile through a pancreatic lesion

Clinically practical respiratory motion correction in simultaneous PET/MR

The recent advent of PET/MR scanners allows us to exploit the simultaneity of the modalities by using high spatial resolution and high contrast MR images to track respiratory motion and use it to correct PET data, without increasing patient scan times and radiation exposure. In collaboration with Dr David Atkinson (UCL Centre for Medical Imaging), we proposed a practical respiratory motion correction regime that requires no external hardware to provide a respiratory signal and no change to the clinical protocol except a short extra MR sequence run after the clinical scan to provide a patient-specific motion model. The approach is anatomically general, applicable to any type of thorax/abdomen related motion caused by respiration e.g. lung, liver, pancreatic lesions and cardiac data.

The method [ref. 13] works by extracting a respiratory signal from the PET data (see also previous section), using this PET-derived signal to bin PET and MR data, constructing a motion model from the binned MR data and using this data to correct PET data for respiratory motion. We have validated our PET-derived respiratory signal by comparison with an absolute measure of diaphragmatic displacement via in MR pencil-beam navigator. Motion-corrected images were compared to non-corrected images qualitatively, and quantitatively with line profiles and change in SUV through regions of high tracer uptake (Fig. 4).

Current work focuses on building a continuous motion model from 1 min worth of dynamic MR data. This enables to extrapolate motion information even if the respiratory pattern during the PET or MRAC was not observed during the dynamic MR sequence, enabling use of all the PET data and avoiding artefacts when the MRAC was acquired at an atypical breathing stage.

References
  1. Thielemans K, Rathore S, Engbrant F, Razifar P, Device-less gating for PET/CT using PCA. Conf. Rec. IEEE NSS-MIC, Valencia, Spain (2011).
  2. Thielemans K, Schleyer P, Marsden PK, Manjeshwar RM, Wollenweber SD, Ganin A, Comparison of Different Methods for Data-driven Respiratory Gating of PET Data. proc. IEEE MIC 2013, Seoul, Korea.
  3. Bertolli O, Arridge S, Stearns CW, Wollenweber SD, Hutton BF, Thielemans K, "Sign determination methods for the respiratory signal in data-driven PET gating". IEEE Medical Imaging Conference, November 2015, San Diego, USA.
  4. Manber R, Thielemans K, Hutton B, Barnes A, Ourselin S, Arridge S, O'Meara C, Wan S, Atkinson D (2015). Practical PET Respiratory Motion Correction in Clinical PET/MR. J Nucl Med, Jun;56(6):890-6.

Back to top


Image reconstruction in emission tomography

Our group focuses on the reconstruction of images from single photon emission tomography (SPECT) and positron emission tomography (PET) raw data using iterative algorithms such as maximum likelihood expectation maximisation (MLEM) and Maximum A Posteriori (MAP) type algorithms.

Our current research is targeted towards accurate quantification, including aspects of system modelling, patient motion, and numerical convergence. Our research also includes reconstruction of multi-modality data such as PET/CT and PET/MRI.

We work closely with researchers from the UCL Centre for Medical Image Computing and the groups of Profs. Sébastien Ourselin and Simon Arridge. Since April 2015, we lead the EPSRC-funded Collaborative Computational Platform on Synergistic PET-MR Reconstruction.

Acceleration of MAP Image Reconstruction Algorithms

Traditional Maximum Likelihood (ML)-type image reconstruction algorithms suffer from problems with either reduced quantification or increased noise due to the ill-posedness of the ML inverse problem. Maximum a Posteriori (MAP) regularises the solution but can need a higher number of iterations to obtain a clinically useful image. To circumvent this, we have proposed a fast algorithm, LBFGS-B-PC [ref. 14]. The algorithm combines LBFGS-B with diagonal preconditioning. Preliminary results show good performance which is relatively independent of the choice of prior etc, which might make this algorithm a good candidate for use in a clinical context.

Joint activity / attenuation reconstruction in SPECT/PET

Image reconstruction in emission tomography needs an estimate of the attenuation of the photons. This is nowadays often obtained from CT, but there are situations where this is not possible or desirable (e.g. due to dose considerations or standalone-SPECT or PET-MR).

In SPECT, our team has investigated the utilisation of dual energy window data (i.e. photopeak and scatter windows) to jointly reconstruct the activity distribution image and the attenuation map, thus removing the need for CT measurement [ref. 15]. We are currently working on improving this approach with the utilisation maximum-likelihood methods and GPU-accelerated implementation using the CUDA library (NVidia), as well as priors. Another work in progress is the joint activity/attenuation reconstruction in PET using dynamic data.

Joint image reconstruction/motion estimation in respiratory gated PET/CT

Gating can be used in PET/CT to reduce the effect of respiratory motion. Obtaining gated CT images however considerably increases the dose delivered to the patient. To avoid the need for gated CT, we developed an algorithm that jointly estimates motion and activity distribution directly from the raw data in PET/CT using a single CT image [ref. 15]. The CT image is warped alongside the activity for attenuation correction. Figure 5 shows the motion corrected PET image using this algorithm. More recently, this approach was extended to TOF-PET/CT [ref. 17].

Top: non-motion corrected reconstructed PET image. Bottom: motion-corrected reconstructed PET image using the algorithm proposed in ref. 16.

Joint-Reconstruction in PET-MR

We have developed methods for joint reconstruction of multi-modality data [ref. 18] which will allow us to investigate the potential advantages of using common structural information in for instance PET/MR applications that need fast imaging. We have recently shown that the joint priors can be used for more "traditional" image reconstruction of PET data with anatomical information from MR [ref. 19].

We are also working on image reconstruction for the INSERT SPECT/MR system. Other PET/MR related projects including head motion correction and GPU-enabled image reconstruction with our UCL collaborators are described on CMIC.

To be able to develop and evaluate new image reconstruction algorithms, a considerable amount of time needs to be invested in the software. We therefore lead various Open-Source Software projects.

References
  1. Tsai Y, Bousse A, Ehrhardt MJ, Hutton BF, Arridge SR and Thielemans K, "Performance Evaluation of MAP Algorithms with Different Penalties, Object Geometries and Noise Levels", IEEE MIC Conf. Proc., 2015.
  2. Cade CS, Arridge S, Evans MJ, Hutton BF, "Use of measured scatter data for the attenuation correction of single photon emission tomography without transmission scanning," Med. Phys., vol. 40, no. 8, 2013. http://dx.doi.org/10.1118/1.4812686
  3. Bousse A, Bertolli O, Atkinson D, Arridge S, Ourselin S, Hutton BF, and Thielemans K, "Maximum-likelihood joint image reconstruction/motion estimation in attenuation-corrected respiratory gated PET/CT using a single attenuation map," IEEE Trans. Med. Imag., vol. 35, no. 1, pp. 217-228, 2016. http://dx.doi.org/10.1109/tmi.2015.2464156
  4. Bousse A, Bertolli O, Atkinson D, Arridge S, Ourselin S, Hutton BF, and Thielemans K, "Maximum-likelihood joint image reconstruction and motion estimation with misaligned attenuation in TOF-PET/CT," Phys. Med. Biol., vol. 61, pp. L11-19, 2016. http://dx.doi.org/10.1088/0031-9155/61/3/L11
  5. Ehrhardt MJ, Thielemans L, Pizarro L, Atkinson D, Ourselin S, Hutton BF, and Arridge SR, "Joint reconstruction of PET-MRI by exploiting structural similarity," Inverse Problems, vol. 31, 2015. http://dx.doi.org/10.1088/0266-5611/31/1/015001
  6. Ehrhardt MJ, Markiewicz P, Liljeroth M, Barnes A, Kolehmainen V, Duncan J, Pizarro L, Atkinson D, Hutton B, Ourselin S, Thielemans K, Arridge SR, "PET Reconstruction with an Anatomical MRI Prior using Parallel Level Sets.", IEEE Trans. Med. Imag., 2016 (in press) http://dx.doi.org/10.1109/tmi.2016.2549601

Back to top


Kinetic Modelling

In nuclear medicine studies, the uptake, retention, and washout of a tracer in a tissue region depends on physiological and biochemical parameters, such as blood flow, metabolic rate and receptor concentration. These parameters can be quantified by analysing time-activity curves obtained from dynamic PET or SPECT scans - a procedure known as kinetic modelling.

At the INM we have several ongoing projects that based on kinetic modelling. Some of these involve development of new methods by combining PET and MRI data (1,2 and 3 below), some involve optimization of acquisition and analysis methods for particular tracers (2 and 3).

Arterial Input Function estimation

A good measurement of the arterial input function (AIF) is required to estimate kinetic parameters accurately. The current gold standard method requires collection of arterial blood samples, which makes it impractical due to its invasiveness. Imaged derived input function (IDIF) is an alternative method used to extract the AIF from reconstructed PET images but has some limitations due to the restricted spatial resolution of PET scanners.

The main goal of this research is to develop non-invasive PET input function measurement techniques utilising MRI data. A practical IDIF method for brain images has been developed which only requires segmentation of the carotid arteries from TOF MR Angiography images [1].

A novel partial volume correction (PVC) method, called Single Target Correction method has been introduced, which is tailored for IDIF extraction methods as it only requires segmentation of the carotid arteries with no need for additional parcellation of background regions.

In addition, the information from partial volume corrected whole-blood TACs were combined with the simultaneous estimation approach to non-invasively analyse PET tracers where radiometabolites are present in the blood plasma.

This methodology has been tested with simulated FDG data [2] and 11C-SB207145 PET tracer [3] which is used for serotonin 4 receptor (5-HT4 receptor) studies and the input function needs to be corrected for metabolites present in the plasma.

Fig. 6. Brief description of IDIF extraction method, including segmentation, registration and PVC

ASL-incorporated kinetic modelling of PET data with reduced acquisition time

Pharmacokinetic analysis of PET data typically requires at least one hour of image acquisition, which poses a great disadvantage in clinical practice. In this work, we propose a novel approach to perform pharmacokinetic modelling with significantly reduced PET acquisition time, by incorporating the blood flow information from simultaneously acquired arterial spin labelling (ASL) MRI. Evaluation on clinical amyloid imaging data from an Alzheimer's disease study shows the proposed approach with simplified reference tissue model can achieve amyloid burden estimation from 30-min [18F]florbetapir data and 5-min ASL MR data acquired simultaneously which is comparable with the estimation from 60-min [18F]florbetapir data.

Fig. 7: Schematic of conventional dynamic PET acquisition time for pharmacokinetic modelling to estimate amyloid burden and the reduced time required for the proposed method.
Reference: Scott C et al. (2016) MICCAI, Athens, Greece.

Improved parameter-estimation with MRI-constrained PET kinetic modelling

We have investigated the potential of MRI-constrained PET kinetic modelling using simulated [18F]2-FDG data. The volume of distribution, Ve, for the extra-vascular extra-cellular space (EES) can be estimated by dynamic contrast enhanced (DCE) MRI, and then used to reduce the number of parameters to estimate in the PET model.

We used a 3 tissue-compartment model with 5 rate constants (3TC5k), to distinguish between EES and the intra-cellular space (ICS). The results from the simulations showed reductions in bias and variance with Ve constraint. The accuracy of the parameters estimated with our new modelling approach depends on the accuracy of the assumed Ve value.

In conclusion, we have shown that, by utilising information that could be obtained from DCE-MRI in the kinetic analysis of [18F]2-FDG-PET data, it is in principle possible to obtain better parameter estimates with a more complex model, which may provide additional information as compared to the standard model.

Fig 8: 3TC5k model (left), and results obtained with different models (right), showing reduced bias and variance using this with fixed Ve.

Quantification of [18F]-FMISO binding in colorectal tumours

In solid tumours, hypoxic regions can develop due to poor blood supply. Hypoxia is associated with poor treatment outcome. The PET tracer [18F]-FMISO can be used to map hypoxic tumour regions. Under hypoxic conditions, FMISO binds irreversibly to macromolecules and becomes trapped inside the cells. Due to the poor blood supply, tracer molecules typically need to diffuse across large distances before reaching the hypoxic region.

This results in a low initial uptake, followed by slow accumulation of tracer. Well perfused regions, on the other hand, will show high initial uptake, followed by wash-out.  As a result, the two time-activity curves may cross at some time-point.

For this reason, a dynamic imaging protocol may be advantageous compared to a single late static scan, which has been used in the past. We have analysed dynamic FMISO data from patients with colorectal tumours and compared various different data acquisition protocols and kinetic modelling approaches [ref. 21-22].

Some results are shown in the diagrams

Fig. 9. Parametric [18F]-FMISO images (left) and time-activity curves for hypoxic and normal VOIs (right).

References
  1. Erlandsson K, et al., (2016) IEEE Trans. Nucl. Sci., DOI: 10.1109/TNS.2015.2507444
  2. Erlandsson K et al., (2016), Quantification of [18F]-FMISO binding in colorectal tumours using dynamic PET data: VOI- and voxel-based approaches, SNM annual meeting, San Diego.
  3. Erlandsson K et al., (2016), A simplified acquisition protocol for dynamic [18F]-FMISO PET studies of colorectal tumours, SNM annual meeting, San Diego.

Back to top


Improved Quantification in the Lung for PET/CT in Idiopathic Pulmonary Fibrosis

Idiopathic Pulmonary Fibrosis (IPF) is an interstitial lung disease (ILD) characterised by an increase in the quantity of extra-cellular matrix and the destruction of parenchyma structure. The result is a reduction in the lung's ability to undergo gas exchange, ultimately causing systemic hypoxia and cardiac failure. As idiopathic suggests, there is no known cause of the disease and currently, treatment options are very limited. With a mean survival of 3 years after diagnosis, IPF has an extremely high mortality.

Clinical trials investigating IPF treatments are limited by an inability to track progression of disease. However, positron emission tomography / computerised tomography (PET/CT), a functional imaging methodology, may provide a disease progression biomarker making it a very promising technique for use in clinical studies.

PET/CT provides images of the concentration of a radiotracer in the body at a spatial resolution of approximately 4-7mm. While static imaging is the most regularly used imaging method, the parameter derived, the standardised uptake value (SUV) is sensitive to the time of the acquisition and the supply of the radiotracer to the tissue. Dynamic imaging however, makes use of the time variable tracer uptake and radiotracer supply, providing much more robust information on the physiological changes in the tissue. Use of dynamic imaging in diffuse lung diseases is rare and to obtain accurate parameter estimates in this patient group consideration must be given to how the respiratory cycle and the tissue fraction effect (TFE), a result of the time variable lung air and blood components, will impact the PET quantitation.

The Tissue Fraction Effect

As part of this work, a previously published correction method for the TFE was expanded to include correction for the blood component in static imaging and then further developed to allow air and blood corrections (ABC) of the kinetic parameter estimates. Use if the ABC was found to reduce bias due to variations in the quantity and uptake of the air and blood components, ensuring that the measured PET signal originated from the parenchyma. The result is a potential enhancement in the sensitivity of the measured parameter to changes induced by drug interventions [ref. 23]

These techniques were further reviewed using IPF patient data and found that correcting for the air and blood components in the lung inverted the relative SUV and influx rate constant estimates between the fibrotic and normal appearing tissue. This meant that while previously it was thought that there was an increased uptake of FDG in the regions of fibrosis in comparison to the normal appearing lung, after correction, this was no longer the case, changing the possible interpretation of the patient data [ref. 23].

Further work in this area is required to accurately determine and validate the fractional blood volume in the lung, required to apply the ABC. To do this, partial volume effects, both in the aorta and the lung itself need to be accounted for.

PET/CT Mismatch Due to the Respiratory Cycle

Consideration has been given to the effects of PET and CT attenuation map mismatch on PET parameter estimates, static and dynamic, in both the lung and a region representing a lung tumour. Errors associated with the mismatch were found to be predominantly local to the area of the mismatch and dependent on the activity distribution throughout the body in the field of view. The error dependency on activity distribution means that the effects of mismatch between PET and CT are specific to the radiotracer used and the time of acquisition. As a solution to these issues, a combination of CT information from a multi-scan protocol was suggested and, when used as the attenuation map for reconstructing the PET data was found to reduce PET quantitation errors.

Studies undertaken at the INM have allowed characterisation and correction of some of the uncertainties associated when imaging diffuse lung diseases, allowing for more stable and robust parameters to be determined [refs. 23-24]. However, more work needs to be done if a useful disease biomarker from dynamic PET data is to be found. The aim of our current research is therefore to investigate and attempt to provide solutions for the issues associated with obtaining quantitative static and kinetic PET parameter estimates in the lung in general and in IPF.

References
  1. Holman BF, Cuplov V, Millner L, Hutton BF, Maher TM, Groves AM, and Thielemans K. Improved correction for the tissue fraction effect in lung PET/CT imaging. Phys Med Biol, 60(18):7387-7402, 2015. http://dx.doi.org/10.1088/0031-9155/60/18/7387.
  2. Holman BF, Cuplov V, Hutton BF, Groves AM, and Thielemans K. The effect of respiratory induced density variations on non-TOF PET quantitation in the lung. Phys Med Biol, 61(8):3148-63, 2016. http://dx.doi.org/10.1088/0031-9155/61/8/3148.

Back to top


Partial volume correction

The quantitative accuracy of PET and SPECT is affected by a phenomenon known as the partial volume effect (PVE). These effects are caused by the limited resolution of the scanner result in blurred images. Most partial volume correction (PVC) techniques utilise anatomical information from other imaging systems such as MRI or CT [ref. 25 for review]. By combining information from the high-resolution MRI or CT with the lower resolution PET or SPECT, PVEs can be reduced or removed, resulting in better estimation of the tracer distribution. With the availability of simultaneous PET/MR scanners, the application of PVC, particularly for brain imaging, is now much more feasible [ref. 26].

At the INM, we have developed a range of PVC algorithms for PET and SPECT applications. One PVC method, developed specifically for SPECT, is incorporated into an iterative reconstruction algorithm [ref. 27]. Fig. X shows an example with the DATSCAN tracer, used in the study of Parkinson's disease. We also developed the region-based voxel-wise (RBV) correction [4], which was applied in dementia studies with PET tracers that can image amyloid plaques (a hallmark of Alzheimer's disease).

An alternative method is the iterative Yang (iY) technique [ref. 25], which is conceptually similar to RBV, but faster and simpler. Gaussian mixture model-based deconvolution (GMD) [ref. 29] performs PVC while also maintaining the contrast in areas of the PET which may not be observable in MRI, such as hypometabolism in epilepsy.

A method called Single Target Correction (STC) was also developed for the situation when only one image region needs to be corrected [ref. 30]. STC operates on a voxel-basis and does not require definition of any background regions.

A brain SPECT study on a healthy volunteer using the tracer DATSCAN without (top) and with PVC (bottom), including sagittal (left), transaxial (middle) and coronal (right) sections.

It is important to be able to compare different PVC approaches under realistic conditions [ref. 31]. We are therefore currently in the process of developing a framework for the generation of test datasets based on actual patient data, which would also be made available to the wider nuclear medicine community. We have already made some of our software for PVC available as Open-Source Software.

References
  1. Erlandsson K, et al. (2012) Phys. Med. Biol. 57:R119-R159.
  2. Erlandsson K, et al. (2016) PET Clin 11:161-177, http://dx.doi.org/10.1016/j.cpet.2015.09.002
  3. Erlandsson K, et al. (2011) Nucl. Instr. Meth. A: 648: S85-88.
  4. Thomas BA, et al. (2011) Eur J Nucl Med Mol Imaging, 38:1104-19.
  5. Bousse A, et al. (2012) Phys. Med. Biol., 57:6681-6705.
  6. Erlandsson K, Hutton BF, J Nucl Med, 55 (Supplement 1):2123, 2014.
  7. Hutton BF, et al. (2013) Nucl. Instr. Meth. A, 702:29-33.

Back to top


'TEXRAD' Texture analysis: Quantifying Heterogeneity to provide novel imaging biomarker in oncology

Heterogeneity is a key feature of malignancy associated with adverse tumour biology. Quantifying heterogeneity from imaging could provide a useful non-invasive biomarker. Heterogeneity on routinely acquired diagnostic images such as CT, MR, PET/CT, PET/MR, mammography, ultrasound can be quantified using texture analysis (TA) which may otherwise be imperceptible to the naked eye.

One such TA-approach is the filtration-histogram technique where the filtration step extracts and enhances features of different sizes (comprising of fine, medium and coarse texture scales corresponding to 2-6mm in size) followed by quantification using histogram and statistical-based parameters.

In our recent article [ref. 32] we describe the filtration-histogram technique of CTTA in more detail and demonstrated how these texture parameters relate to different aspects of heterogeneity such as number of objects, size, variation in density of the objects in relation to the background (parenchyma) of the tissue/lesion of interest.

Colorectal TEXRAD Texture Analysis, with a four-stage image (labelled A-D) and a histogram of medium filtered lesion image

As part of the qualification-process of CTTA as an imaging biomarker in oncology, a number of research-studies have demonstrated evidence towards:

  • Biological correlates (hypoxia, angiogenesis, fibrosis, proliferation, genetic mutation) e.g. lung cancer [refs. 33-35]
  • Technical validation (test-retest reproducibility and reliability assessment, multi-centre studies, degree of variability to different image acquisition parameters) e.g. lung cancer [refs. 36-37]
  • Clinical applications (e.g. lesion-characterisation, prognosis, treatment-response and prediction. e.g. lung-cancer [refs. 37-38]
  • Cost-effectiveness & clinical-utility (e.g. PET/CT lung MDT - clinical-adoption audit study); e.g. lung-cancer [refs. 39-40]

CTTA has demonstrated potential clinical application as an adjunct to radiological and oncological workflow in a number of cancer sites including lung, colorectal (primary & metastases), oesophageal, breast, prostate, head & neck, lymphoma, renal (primary & metastases), metastatic melanoma, soft-tissue sarcoma, gliomas, neuro-endocrine tumours etc. CTTA has demonstrated added value in comparison and in combination (multi-parametric) with known existing biomarkers (size, attenuation, perfusion, metabolic activity, clinical stage and other clinical, blood markers and patient demographics) in several oncological applications.

Recent applications in multi-parametric MR include identifying significant prostatic cancers within transition zone and amongst PIRADS 3 patients having significant cancers within the peripheral zone. Additionally, in rectal and breast cancer MRTA has demonstrated treatment response to chemo- and radiotherapy.

References
  1. Miles KA, Ganeshan B, Hayball MP. CT texture analysis using the filtration-histogram method: what do the measurements mean? Cancer Imaging. 2013 Sep 23;13(3):400-6. doi: 10.1102/1470-7330.2013.9045. Review. http://www.ncbi.nlm.nih.gov/pubmed/24061266
  2. Weiss GJ, Ganeshan B, Miles KA, Campbell DH, Cheung PY, Frank S, Korn RL. Noninvasive image texture analysis differentiates K-ras mutation from pan-wildtype NSCLC and is prognostic. PLoS One. 2014 Jul 2;9(7):e100244. doi: 10.1371/journal.pone.0100244. eCollection 2014. http://www.ncbi.nlm.nih.gov/pubmed/24987838
  3. Ganeshan B, Goh V, Mandeville HC, Ng QS, Hoskin PJ, Miles KA. Non-small cell lung cancer: histopathologic correlates for texture parameters at CT. Radiology. 2013 Jan;266(1):326-36. doi: 10.1148/radiol.12112428. Epub 2012 Nov 20. http://www.ncbi.nlm.nih.gov/pubmed/23169792
  4. Ganeshan B, Abaleke S, Young RC, Chatwin CR, Miles KA.Texture analysis of non-small cell lung cancer on unenhanced computed tomography: initial evidence for a relationship with tumour glucose metabolism and stage. Cancer Imaging. 2010 Jul 6;10:137-43. doi: 10.1102/1470-7330.2010.0021. http://www.ncbi.nlm.nih.gov/pubmed/20605762
  5. Chen SH*, Ganeshan B*, Fraioli F. Reproducibility of CT texture parameters by leveraging publicly available patient imaging datasets. British Journal of Radiology 2016 [under-review]
  6. Win T, Miles KA, Janes SM, Ganeshan B, Shastry M, Endozo R, Meagher M, Shortman RI, Wan S, Kayani I, Ell PJ, Groves AM. Tumor heterogeneity and permeability as measured on the CT component of PET/CT predict survival in patients with non-small cell lung cancer. Clin Cancer Res. 2013 Jul 1;19(13):3591-9. doi: 10.1158/1078-0432.CCR-12-1307. Epub 2013 May 9. http://www.ncbi.nlm.nih.gov/pubmed/23659970
  7. Ganeshan B, Panayiotou E, Burnand K, Dizdarevic S, Miles K. Tumour heterogeneity in non-small cell lung carcinoma assessed by CT texture analysis: a potential marker of survival. Eur Radiol. 2012 Apr;22(4):796-802. doi: 10.1007/s00330-011-2319-8. Epub 2011 Nov 17. http://www.ncbi.nlm.nih.gov/pubmed/22086561
  8. Miles KA, Ganeshan B. Selection of patients for advanced non-small cell lung cancer for chemotherapy: Potential cost-effectiveness of CT Texture Analysis. In European Society of Radiology 2012, Vienna, Austria.
  9. Miles KA. How to use CT texture analysis for prognostication of non-small cell lung cancer. Cancer Imaging. 2016 Apr 11;16(1):10. doi: 10.1186/s40644-016-0065-5. Review. http://www.ncbi.nlm.nih.gov/pubmed/27066905

Back to top