Proton radiography and tomography with application to proton therapy

Proton radiography and tomography have long promised benefit for proton therapy. Their first suggestion was in the early 1960s and the first published proton radiographs and CT images appeared in the late 1960s and 1970s, respectively. More than just providing anatomical images, proton transmission imaging provides the potential for the more accurate estimation of stopping-power ratio inside a patient and hence improved treatment planning and verification. With the recent explosion in growth of clinical proton therapy facilities, the time is perhaps ripe for the imaging modality to come to the fore. Yet many technical challenges remain to be solved before proton CT scanners become commonplace in the clinic. Research and development in this field is currently more active than at any time with several prototype designs emerging. This review introduces the principles of proton radiography and tomography, their historical developments, the raft of modern prototype systems and the primary design issues.Despite a history going back over 50 years,¹ proton radiography (pRG) and tomography have been slow to reach the clinic.² Few manufacturers currently offer a clinical imaging system suitable for pRG and none for proton tomography. In fact, it turns out that the use of protons instead of X-rays for transmission imaging has some disadvantages. These include the need for large expensive equipment to produce proton beams (e.g. a cyclotron or synchrotron) and the limitations on image quality arising from the multiple scattering of protons.Proton sources of sufficient energy do, however, exist for several purposes, one application being for proton therapy. The multiple scattering effects remain a fundamental difficulty: protons do not move through a medium in straight lines. So why should we even attempt proton transmission imaging? The prime motivation is with application to proton therapy planning. It was Cormack¹ who was the first to realize the possibilities of proton CT (pCT). In a seminal article of the 1960s on tomographic reconstruction, the Nobel Laureate wrote:

The next application of the solution for CT] … concerns the recent use of the peak in the Bragg curve for the ionization caused by protons, to produce small regions of high ionization in tissue. The radiotherapist is confronted with the problem of determining the energy of the incident protons necessary to produce the high ionization at just the right place, and this requires knowing the variable-specific ionization of the tissue through which the protons must pass.

This is still a fair assessment of the problem facing any proton therapy team today. Cormack went on to propose that the energy loss of protons passing through a patient can tell us about proton stopping power inside the patient—something that X-rays can never give us directly.Typically, in both photon and proton external beam therapy, prior to treatment, an X-ray CT scan is acquired for treatment planning purposes. This is used for outlining structures, but also provides a map of electron density that is used to calculate dose deposition. In proton therapy, the translation of electron density to proton stopping power provides an extra and appreciable source of error. The most advanced X-ray CT calibration method in common usage is probably the stoichiometric method.³ The resulting overall uncertainty (1σ) in stopping-power ratio (SPR) for protons in different tissue types has been estimated as 1.6% (soft tissue), 2.4% (bone) and 5.0% (lung).⁴ As an illustration, note that the estimate of 1.6% for soft tissue includes contributions for (added in quadrature): stoichiometric parameterization (0.8%), human tissue composition variation (1.2%) and mean excitation energy (0.2%) and other sources (0.6%). None of the first three sources of errors contribute in a calibration in pCT and the ambition with this type of imaging should be to reduce the uncertainty in SPR substantially (to <1%). Reduced uncertainties offer the possibility of smaller planning margins and additional beam directions, potentially leading to superior patient outcomes. The surge in the number of operational and planned proton therapy centres in recent years therefore makes the exploitation of this modality timely.⁵Before proceeding further, some clarification of topic coverage should be made. pRG and pCT, in the context of this review, mean the imaging of an object using the transmission of protons through it. The energy loss of the transmitted protons is the primary mechanism for image contrast. The greatest emphasis will be given to proton-tracking systems: as will be seen, these are best able to cope with the difficulties imposed by proton multiple scattering. Some requirements for a practical pCT scanner for proton therapy are summarized in 10 For comparison, note that a typical head scan using a diagnostic X-ray CT scanner or X-ray cone beam CT (CBCT) might deliver 40 mGy.¹¹

Table 1.

Requirements for a practical (proton-tracking) CT scanner for proton therapy

Category	Parameter	Value
Proton beam	Energy	≥200 MeV (head)
		≥250 MeV (body)
	Fluxa	≥3000 protons cm−2 s−2
Imaging dose	Maximum absorbed doseb	<20 mGy
Image quality	Spatial resolution, σ	≈1 mm
	Relative stopping-power accuracy	<1%
Time	Data acquisition time	<10 min
	Reconstruction time	<10 min

Open in a separate window^aQuoted figure based on the scenario of 1-mm voxels and 180 projections, a target of 100 protons passing through a voxel per projection⁶ and a 10-min acquisition.^bQuoted figure based on a crude calculation of comparable stochastic risk to typical X-ray CT head scans (≈40 mGy^7,8), assuming a proton radiation weighting factor twice that of photons.⁹We will not be concerned here with other forms of imaging using proton beams, such as nuclear scattering tomography¹² that relies on wide-angle scattering, γ interaction vertex imaging¹³ (GIVI) using prompt γ emission or positron emission tomography¹⁴ (PET) of induced β emission. The latter two (GIVI and PET) primarily promise benefit for in vivo range verification (inferring the depths that protons penetrated).¹⁵ Finally, we emphasize that our interest in this review is with protons. Reference to heavy-ion radiography and tomography will be made only where comparison with imaging with protons is apt, and we refer the reader to other sources¹⁶ for this related topic.