CT thermometry: will it ever become ready for use?

Over the last decades, image-guided hyperthermal tumour ablation became increasingly popular. This includes techniques such as radiofrequency (RF) ablation, laser-induced thermotherapy and microwave ablation. Common ground of these methods is to heat homogeneously a well-defined volume to more than 60 °C. At this target temperature, a nearly instantaneous protein denaturation with subsequent coagulation necrosis is achieved. As the percutaneous application of these techniques relies on a homogeneous heating of the target volume, non-invasive, image-based temperature mapping is appealing to ensure complete tumour ablation, including a safety margin and to avoid thermal damage to adjacent structures. Consequently, researchers are working on this issue for almost four decades, and several technical solutions have been proposed, including ultrasound (1,2) or magnetic-resonance (MR)-based thermometry and microwave thermal imaging (3). Particularly MR-based thermometry is considered promising, and several techniques using the proton resonance frequency, the proton density, the diffusion coefficient, T1 and T2 relaxation times and magnetisation transfer have been evaluated. Even temperature-sensitive contrast agents have been proposed in the literature (4). Some of these approaches were even applied clinically for monitoring hyperthermal ablation with MR imaging (5). However, there are several technical and logistic limitations hampering the use of MR thermometry, including the high cost of MR-guided ablation and the availability of MR-compatible instruments. More than three decades ago, the temperature dependence of computed tomography (CT) numbers was investigated. A temperature-dependent shift of CT numbers was found for different materials, but because of technical capabilities of the CT scanners at that time, this application was not further developed (6). Until recently it was almost forgotten. This perspective aims to summarise the currently available data on CT thermometry and to elucidate the potential use of this technique. The temperature-dependent change of CT numbers in water and biological tissue samples is known since the late 1970s (6). X-ray attenuation at the energy levels used in CT is largely a result of Compton scattering. In this photon energy range, if molecular interactions are ignored, the energy loss is proportional to electron density, which in turn varies with physical density for a given material. The density of water, for instance, decreases with increasing temperature. The coefficient of thermal density expansion of water near human-body temperature is −3.6 × 10−4/°C i.e. a temperature change of 1 °C causes a change in density of water molecules of 0.036%, which corresponds to a decrease in attenuation of roughly 0.4 Hounsfield units. This correlates well with the experimental data from the late 1970s and early 1980s (6,7). At that time, it had been postulated that temperature discrimination of a fraction of °C at a spatial resolution of approximately 10 mm would be achievable with contemporary equipment. However, at that time CT values were far from stable, and reproducibility of quantitative CT measurements was not guaranteed. Moreover, there was no indication of using this technique except for calibration purposes. It was not until 1997 that CT temperature measurements were considered useful in local tumour ablation (8). Still, precision of the CT scanners was not sufficient and local ablation was rarely used. Thus, CT′s potential for measuring temperature sank into oblivion. Modern CT scanners with state-of-the-art detectors provide fairly stable CT numbers, resulting in reliable, linear thermal calibration curves for a broad variety of materials (Figure 1) (9). The technique reliably works in organic tissue too (Figure 2). Since the late 1990s, image-guided thermal ablation became a mainstay of tumour therapy with roughly 100,000 hepatic RF ablation procedures per year. Given a mature technical basis and a clinical demand for monitoring therapy, the point is reached to push CT thermometry another step forward. There is a linear relationship between the temperature of a variety of substances and CT values, as measured with current state of the art CT scanners. The relationship between the temperature of diluted contrast material (1 : 128 Iopromide 370) and CT values exemplarily illustrates this feature Greyscale CT images of liver specimen heated by a bipolar RF probe. Increasing temperature is encoded by increasing brightness. At baseline (top) the temperature was 37 °C. The increase in temperature started close to the RF-probe (middle), and by convection the specimen was heated to 65 °C (bottom). For the first time, in vivo studies are on the way for evaluating the use of CT for temperature mapping during RF ablation. Now, we have to consider what this technique can do and what is needed to make it an efficient tool. For monitoring thermal ablation in clinical routine practice, CT thermometry has to meet several demands: The entire volume intended to treat needs to be covered with a single data acquisition. Dedicated software tools are needed to visualise isotherms with a precision of an estimated 5 °C/mm, providing near real-time visualisation of the temperature distribution. Metal artefacts from the devices used for thermal ablation need to be corrected sufficiently. Computed thermometry needs to become independent from contaminations such as attenuation changes caused by contrast material injection. The radiation exposure from repeated CT measurements needs to stay in reasonable ranges, limiting the additional radiation exposure that may be used for CT thermometry. Considering recent developments in CT technology, there are potential technical solutions for all of these aspects on the horizon. Broad detectors with up to 320 detector rows permit examination of volumes with a length of 4–16 cm along the patient's z-axis. Dedicated algorithms for metal artefact reduction have been described in the literature with some of these algorithms providing excellent results in the presence of limited and geometrically well-defined objects such as puncture needles. Dual energy imaging permits computation of virtual unenhanced CT images. Thereby, CT thermometry can be performed even if the contrast material is applied. Nevertheless, these techniques have to be validated with actual data from animal studies prior to clinical use. Limiting the radiation exposure caused by CT thermometry will be the most important and most challenging demand. However, with innovative filters and data reconstruction techniques such as iterative image reconstruction, there is a great toolbox providing a whole bunch of different techniques for developing CT thermometry into a clinical tool. Thus, after more than three decades CT thermometry is finally at the cusp for clinical evaluation.