Since its introduction in 2000, PET/CT has become a widespread and effective imaging tool for the diagnosis and management of patients with cancer. Today, over 5,000 combined PET/CT systems are in clinical operation worldwide. As a core component of PET/CT, PET is a highly sensitive imaging technique capable of detecting as few as one million cancer cells [1]. Its inherent quantitative nature enables accurate, reproducible measurements of radiopharmaceutical uptake in the tumour during diagnostic work-up, therapy and treatment follow-up. Technological advances in PET imaging technology go hand in hand with the development of highly specific PET probes, with most of them only now being available for on-site use.

Nuclear medicine is an integral part of modern healthcare. It uses biomolecules tagged with radioactive isotopes that can recognize different molecular targets in the body or seek out hallmarks of malignant and benign conditions. The distribution of trace amounts of a radioactively labelled molecule, known as a radiotracer, can be mapped and followed noninvasively anywhere in the human body using a dedicated system, such as a single photon emission computed tomography (SPECT) system or a positron emission tomography (PET) system. This trace approach provides longitudinal sets of volumetric and quantitative images that can be used to diagnose a wide range of diseases and/or assess response to disease-specific treatments. Replacing these diagnostic radionuclides with alternative nuclides that emit a different type of radiation converts imaging tracers into drugs that deliver potent and targeted molecular treatment. Recent advances have extended the range of molecular radiotherapies, which are now applied across a broad spectrum of diseases from arthritis and benign thyroid diseases to many types of cancer. As such, nuclear medicine is in a key position to bring personalized or image-guided therapy into routine clinical practice. This key position is supported by the rapidly expanding fields of radiochemistry and radiopharmacy with novel kitlike or cassette-type labelling techniques which allow shorter turn-aroundtimesfromtheidentificationoftherapeutictargets to the design and evaluation of associated radiotracers, resulting in lower development costs. Over the past 5 years, significant technological breakthroughs in imaging hardware and image processing algorithms have also been introduced, enabling specific radiotracer imaging for improved characterization of targeted cellular and functional processes in clinical

Poster: "ECR 2014 / C-1135 / Dose reduction in time-of-flight PET/MRI" by: "C. Rubbert1, A. Kohan2, J. L. Vercher-Conejero2, M. Narayanan3, A. Kalemis4, K. A. Herrmann2, P. Faulhaber2, N. Avril2, R. Muzic2; 1Dusseldorf/DE, 2Cleveland, OH/US, 3Highland Heights, OH/US, 4Guildford/UK"

PET and MRI are established clinical tools which provide complementary information, but clinical workflow limits widespread clinical application of both modalities in combination. The two modalities are usually situated in different hospital departments and operated and reported independently, and patients are referred for both scans, often consecutively. With the advent of PET/MR as a new hybrid imaging modality there is now a possibility of addressing these concerns. There are two different design philosophies for integrated PET/MR imaging-positioning PET inside the MRI magnet or in tandem, similar to PET/CT. The Ingenuity TF PET/MR by Philips Healthcare is a sequential PET/MR tomograph combining state-of-the-art time-of-flight PET and high-field MRI with parallel transmission capabilities. In this review article we describe the technology implemented in the system, for example RF and magnetic shielding, MR-based attenuation correction, peculiarities in scatter correction, MR system optimisation, and the philosophy behind its design. Furthermore, we provide an overview of how the system has been used during the last two years, and expectations of how the use of PET/MR may continue in the years to come. On the basis of these observations and experiences we discuss the utility of the system, clinical workflow and acquisition times, and possible ways of optimization.

1522 Objectives Head and Neck (HN) cancer staging may benefit from new hybrid technology such as PET/MRI. The dilemma between a sufficiently long PET acquisition for good quality 2 mm voxel size reconstructions with neurovascular MRI coil attenuation and a short acquisition time due to limited patient cooperation constitutes a challenge in clinical routine. Using an Ingenuity TF PET/MR we investigated the impact of reduced acquisition time on lesion detectability, SUV and SNR measurements. Methods A retrospective reconstruction of listmode PET data (370MBq of F18-FDG injected, acquisition one hour after injection) was performed on 7 patients with histologically proven HN squamous-cell carcinoma (lesion size : 1 to 3 cm) and with neck dissection performed for cervical lymph node metastases (lesion size : 0.3 to 2.1 cm). The initial acquisition protocol was 6 min/bed in the HN area. Reconstructions were performed with 5, 4, 3, 2 and 1.5 min by clipping the listmode data. The relaxation parameters were adapted in order to keep a similar signal-to-noise ratio between reconstructions. Mean and maximum SUV as well as SNR were measured for each patient in the following areas: lung, cerebellum, cervical muscles, tumor site and lymph nodes. Results SUVmean, SUVmax and SNR were not significantly different for all acquisition times except for the SNR of muscle which was significantly decreased for 1.5 min compared to 6 min. Lesions were noticeable on all the reconstructions. All positive lymph nodes on pathology reports detected on the 6min reconstruction (down to 6 mm) were still visible on the 2 min reconstruction but lymph nodes smaller than 7 mm were missed in 1.5 min reconstruction. Conclusions Our preliminary data suggest that in HN squamous cell carcinoma it is possible to reduce the PET acquisition time to 2 min without affecting sensitivity and specificity for lesion detection.

PET/MRI is a new hybrid modality which is increasingly being used in clinical settings, although both clinical evaluation and technical optimization are still an ongoing process. Initial experience with this new imaging device proves promising for oncologic applications. Other clinical indications in the field of cardiac imaging and neuroimaging are also being explored. This article aims to review the current status of PET/MRI and its value in oncologic applications, and summarizes our own preliminary experience in this field.

Accurate quantification of tumour tracer uptake is essential for therapy monitoring by sequential PET imaging. In this study we investigated to what extent a reduction in administered activity, synonymous with an overall reduction in repeated patient exposure, compromised the accuracy of quantitative measures using time-of-flight PET/CT.

This paper presents a novel data-driven method for image intensity normalisation, which is a prerequisite step for any kind of image comparison. The method involves a novel application of the Siddon algorithm that was developed initially for fast reconstruction of tomographic images and is based on a linear normalisation model with either one or two parameters. The latter are estimated by maximising the line integral, computed using the Siddon algorithm, in the 2D joint intensity distribution space of image pairs. The proposed normalisation method, referred to as Siddon Line Integral Maximisation (SLIM), was compared with three other methodologies, namely background ratio (BAR) scaling, linear fitting and proportional scaling, using a large number of synthesised datasets. SLIM was also compared with BAR normalisation when applied to phantom data and two clinical examples. The new method was found to be more accurate and less biased than its counterparts for the range of characteristics selected for the synthesised data. These findings were in agreement with the results from the analysis of the experimental and clinical data.

Physiological gating in nuclear medicine image acquisition was introduced over 30 years ago to subdivide data from the beating heart into short time frames to minimize motion blurring and permit evaluation of contractile parameters. It has since been widely applied in planar gamma camera imaging, SPECT, positron tomography (PET) and anatomical modalities such as x-ray CT and MRI, mostly for cardiac or respiratory investigations. However, the gating capability of gamma cameras and PET scanners can be employed to produce multiply partitioned, statistically independent projection data that can be used in various ways such as to study the effect of varying total acquired counts or time, or administered radioactivity, on image quality and multiple observations for statistical image analyses. Externally triggered gating essentially provides ‘something for nothing’ as no data are lost and a ‘non-gated’ data set is easily synthesized post hoc, and there are few reasons for not acquiring the data in this manner (e.g., slightly longer processing time, extra disk space, etc). We present a number of examples where externally triggered gating and partitioning of image data has been useful.

V/Q (Ventilation / Perfusion) tests are used to diagnose pulmonary emboli predominantly using planar images. This study presents a quantitative, objective statistical analysis of SPECT V/Q scans for the detection of ventilation-perfusion mismatch. The method is based on a voxelwise statistical assessment of the V/Q images differences. A new acquisition protocol (externally triggered gating) is introduced for acquiring multiple frames for the voxelwise variance estimation. The V/Q datasets are then tested by applying a student's t-test in a voxelwise fashion seeking areas of mismatch. The resulting probability image is finally thresholded to display the detected differences. Two different datasets were used: 72 different cases of a computerised model of pulmonary embolism with emboli of varying magnitude and in different segments, plus seven clinical cases with physicians' reports. The method was shown to be robust irrespective of the location of the mismatch for large defects but greater count densities than those used in the simulation studies are needed for the detection of smaller defects. The results from the clinical datasets agree with the clinical reports but in some cases reveal unreported abnormalities.

In this paper two tests based on statistical models are presented and used to assess, quantify and provide positional information of the existence of bias and/or variations between planar images acquired at different times but under similar conditions. In the first test a linear regression model is fitted to the data in a pixelwise fashion, using three mathematical operators. In the second test a comparison using z-scoring is used based on the assumption that Poisson statistics are valid. For both tests the underlying assumptions are as simple and few as possible. The results are presented as parametric maps of either the three operators or the z-score. The z-score maps can then be thresholded to show the parts of the images which demonstrate change. Three different thresholding methods (naive, adaptive and multiple) are presented: together they cover almost all the needs for separating the signal from the background in the z-score maps. Where the expected size of the signal is known or can be estimated, a spatial correction technique (referred to as the reef correction) can be applied. These tests were applied to flood images used for the quality control of gamma camera uniformity. Simulated data were used to check the validity of the methods. Real data were acquired from four different cameras from two different institutions using a variety of acquisition parameters. The regression model found the bias in all five simulated cases and it also found patterns of unstable regions in real data where visual inspection of the flood images did not show any problems. In comparison the z-map revealed the differences in the simulated images from as low as 1.8 standard deviations from the mean, corresponding to a differential uniformity of 2.2% over the central field of view. In all cases studied, the reef correction increased significantly the sensitivity of the method and in most cases the specificity as well. The two proposed tests can be used either separately or in combination and are capable of showing trends and/or the magnitude of difference between images acquired under similar conditions with high positional and statistical precision. In addition to gamma camera quality control, they could be applied to any pair (or set) of registered planar images to detect subtle changes, e.g. a set of scintigrams or conventional radiographs of a patient before, during and after treatment.

Statistical comparisons of images under different conditions (e.g. abnormal vs. normal) require multiple images per condition in order to estimate the variance. This paper presents a new method for calculating the variance from a single scan using image subsampling to remove possible spatial correlation which would bias and reduce the variance. The method uses the mean absolute deviation of the difference image over the whole field of view as a robust estimator of the variance. This single-scan method was compared with the absolute estimator derived from a dynamic PET acquisition of the IEC phantom with different radioactive concentration ratios, acquisition times, spatial extent and threshold levels. The results were quantified by measuring the sensitivity, specificity and detectability of the spherical sources. The new method was found to behave as expected and, for low statistics, perform better than the standard approach.

Comparison of two medical images often requires image scaling as a preprocessing step. This is usually done with the scaling-to-the-mean or scaling-to-the-maximum techniques which, under certain circumstances, in quantitative applications may contribute a significant amount of bias. In this paper, we present a simple scaling method which assumes only that the most predominant values in the corresponding images belong to their background structure. The ratio of the two images to be compared is calculated and its frequency histogram is plotted. The scaling factor is given by the position of the peak in this histogram which belongs to the background structure. The method was tested against the traditional scaling-to-the-mean technique on simulated planar gamma-camera images which were compared using pixelwise statistical parametric tests. Both sensitivity and specificity for each condition were measured over a range of different contrasts and sizes of inhomogeneity for the two scaling techniques. The new method was found to preserve sensitivity in all cases while the traditional technique resulted in significant degradation of sensitivity in certain cases.