Tomographic detection of photon pairs produced from high-energy X-rays for the monitoring of radiotherapy dosing

Abstract

Measuring the radiation dose reaching a patient's body is difficult. Here we report a technique for the tomographic reconstruction of the location of photon pairs originating from the annihilation of positron-electron pairs produced by high-energy X-rays travelling through tissue. We used Monte Carlo simulations on pre-recorded data from tissue-mimicking phantoms and from a patient with a brain tumour to show the feasibility of this imaging modality, which we named 'pair-production tomography', for the monitoring of radiotherapy dosing. We simulated three image-reconstruction methods, one applicable to a pencil X-ray beam scanning through a region of interest, and two applicable to the excitation of tissue volumes via broad beams (with temporal resolution sufficient to identify coincident photon pairs via filtered back projection, or with higher temporal resolution sufficient for the estimation of a photon's time-of-flight). In addition to the monitoring of radiotherapy dosing, we show that image contrast resulting from pair-production tomography is highly proportional to the material's atomic number. The technique may thus also allow for element mapping and for soft-tissue differentiation.

Fig. 1. Illustration of the principles of pair-production tomography imaging.

a, Illustration of the photoelectric effect, Compton scatter and pair-production interaction. b, Illustration of the formation process of pair-production tomography imaging (P2T). c, The energy distribution of detected photons ranging from 0 to 1 MeV. d, The energy distribution of detected photons after applying a ±10% energy window filter. e, The energy distribution of detected photons after applying the ±10% energy and 1 ns coincidence time filters. f, Comparison of VE vs SPBE, and full-view imaging (top) vs partial-view imaging (bottom). VE excites the entire imaging field simultaneously at each view angle. In SPBE, the imaging field is excited sequentially. The partial-view imaging only irradiates the ROI without exposing the majority of the imaging object.

Fig. 2. Phantom study on P2T linearity to atomic number.

a, The CT image of a nanoparticle phantom with 10 inserts, among which 7 inserts were made up of water and 5% of high-Z elements, including iodine, barium, gadolinium, ytterbium, tantalum, gold and bismuth. b, The relative increase in contrast to water was evaluated for the 7 inserts for the CT image and all P2T images. Linear regression of the increased contrast on the atomic number Z was performed for all images. c, Comparison of P2T GT image, P2T image from FBP reconstruction, P2T image from SPB-based reconstruction and P2T image from TOF reconstruction. All images were normalized such that the intensity of the water insert is 1. In the specific case, the maximal imaging dose was around 3.6 cGy, assuming 100% detector efficiency.

Fig. 3. Phantom study of standard materials.

a, The standard phantom with 10 inserts, for air, lung inhale and lung exhale, and for adipose tissue, breast tissue, water, muscle tissue, liver tissue, trabecular bone and dense bone. b, The relative increase in contrast to water was evaluated for all materials except water. Each bar in the bar plots is the mean over n = 25 independent samples. Data are presented as mean ± s.d. The dashed lines show the increments in ρZ_eff of each material relative to water. c, Comparison of P2T GT image, P2T image from FBP reconstruction, P2T image from SPB-based reconstruction, P2T image from TOF reconstruction and CT image. All images were normalized such that the intensity of the water insert is 1.

Fig. 4. P2T allows partial-view and sparse-view imaging.

a, Comparison of 20-beam full-view P2T images (top) and 2-beam partial-view P2T images (bottom), including GT image, P2T image from FBP reconstruction, P2T image from SPB-based reconstruction and P2T image from TOF reconstruction. All images were normalized such that the intensity of the water insert is 1. The three inserts (white spots) in the partial-view from left to right are iodine, ytterbium and bismuth. b, The image intensity of the 3 inserts normalized by their average values for both full-view P2T images and partial-view P2T images.

Fig. 5. Dose monitoring of the radiotherapy treatment for a patient with glioblastoma multiforme, using 10 MV X-ray beams and IMRT.

a–c, All dose (a), TERMA (b) and P2T (c) images are displayed as iso-intensity colour-wash images superimposed on the CT image. The target and the normal tissues are contoured with different colours. All colour-wash images were normalized by the mean intensity value within the target. d, The cumulative intensity volume histograms (cIVHs) of dose, TERMA and P2T GT images. The cIVH lines indicate the volume percentage of a structure receiving intensity values higher than a threshold.