Speckle x-ray imaging: detecting early changes in lung microstructure

Abstract. In the United States alone, the number of proton therapy centers has increased to 41 sites, with many more currently under construction or in planning stage. While the investment for such centers is in the hundreds of millions of US dollars, research is ongoing to determine whether proton therapy improves treatment outcomes.

A sensitive diagnostic tool for the evaluation of alveoli architecture in this active research area would not only enable early targeted treatment to slow down progression of radiation-induced lung fibrosis but also significantly benefit the ongoing preclinical evaluation. The imaging tools currently in use have a poor to

moderate sensitivity that is insufficient for detecting early changes in the lungs and/or are proving impractical with respect to radiation dose and logistical complexity for longitudinal preclinical studies. To address this critical need, we introduce an imaging tool for early detection of lung microstructural changes by advancing the

emerging field of x-ray darkfield imaging. In conventional x-ray, image contrast is formed by attenuation based on the interpretation of x-rays as particles. If sensing x-rays as electromagnetic waves, additional x-ray contrast mechanisms such as diffraction, phase-shift and small-angle scattering can be accessed. X-ray scattering on

healthy, gas-filled pulmonary alveoli generates a strong darkfield signal, and the signal decreases when the integrity of the alveoli is affected. Preliminary in-vivo small animal experiments successfully demonstrated an on average ten-weeks-earlier detection of early onset of radiation-induced lung fibrosis from

conventional photon therapy. A number of methods for acquiring x-ray darkfield images have been investigated in recent years. However, current solutions require complicated, shock-sensitive and expensive hardware implementations. A more practical method involves the use of filters consisting of random structures (so-called

diffusers) to generate near-field interference speckle patterns for acquiring darkfield images. Our long-term goal is translating x-ray dark-field imaging from physics research laboratories into the preclinical imaging arena to provide the needed tool for longitudinal lung assessment. Our solution includes the design of novel deep-

learning based speckle tracking in combination with a diffuser design based on nanoparticles which is inexpensive to fabricate compared to gratings. The following specific aims will be pursued: (1) to develop a software infrastructure for in-vivo small animal x-ray darkfield imaging, (2) to implement an x-ray darkfield

prototype for detection of early pulmonary toxicity from radiotherapy, and (3) to evaluate x-ray darkfield prototype performance in phantoms and in-vivo longitudinal animal studies. This proposal will advance the field of speckle- based x-ray dark-field imaging by deepening the basic understanding and by translating it from physics

research laboratories into the preclinical arena. Toward this end, we anticipate that our x-ray dark-field imaging concept will serve as a low-dose tool for longitudinal in-vivo small animal studies. The proposed solutions have the potential to drive the translation of x-ray dark-field imaging forward into the clinical routine.