Preliminary study of a clinical electron beam using highly pixelated CMOS Image Sensors

Highlights

• CIS devices could be used as spatially very precise ionizing radiation detectors.

• Non-linear calibration could be used to keep uncertainty at the 1% level.

• The intrinsic homogeneous response of CIS to ionizing radiation over the whole surface open the way for uncertainty disentangling procedures.

Abstract

The characterization of high intensity charged particle beams at medical accelerators poses several challenges. In this work, we investigate the use of a highly segmented CMOS Image Sensors (CIS), an MT9T031 from Aptina, as a way to study the spatial homogeneity and the time stability of the beam. The approach relies on the possibility to define many adjacent small regions that perform the radiation flux measurement with sufficient precision (below 1%) to extract the spatial structure of the beam. The device has been exposed to a 10 MeV therapeutic electron beam at Santa Maria Hospital (Terni, Italy) to measure the electron flux at a distance of 140 cm. The whole sensor, using a non-linear calibration, has measured the value of the flux, then the segmentation approach has been applied to study the spatial structure of the beam. Concerning the variation in time, the current limitation of a rolling shutter CIS limits the capability to disentangle the time structure of the beam. However, in light of the possibility to obtain some CIS with the new global shutter architecture, having integration time of the order of a few tens of microseconds, the measurement procedure has been implemented and tested using $\sim 1Hz$ frequency frame-rate, to study its limits. Uncertainty of the order of 0.5% has been reached for measurement of both spatial and time beam homogeneity.

Introduction

The precise and accurate determination of the flux of intense ionizing radiation beams is of paramount importance especially when they are used to treat tumours. Many efforts are put into measuring with $\sim$ 1% uncertainty the energy deposition both in the tumour and the healthy tissues during a radiotherapy session. Many different types of detectors have been used up to now for such measurements. The increasing complexity of Treatment Plans and the reduction of the ionizing beam size makes very difficult the measurement task.

Thanks to the developments of the Complementary Metal–Oxide Semiconductor (CMOS) technology, the single pixels of CIS devices are reducing in size and increasing in complexity of the on-pixel and peripheral electronic. CIS devices are optimized to convert visible light into a digital image to record video or photo, hence this implies a very small pixel size, to increase the picture definition remaining within the standard form factors of the recording devices (usually below 1 cm²). Recently some investigation on the capability of CIS to detect ionizing radiation rather than visible light has demonstrated their efficiency in this task, especially for charged particles like electrons (Turchetta et al., 2001, Battaglia et al., 2008, Faruqi et al., 2005, Evans et al., 2005, Battaglia et al., 2009, Servoli et al., 2008, Servoli et al., 2010, Servoli et al., 2011, Castoldi et al., 2015, Perez et al., 2016). This is due to the extremely small Equivalent Charge Noise (ECN) that can be as low as few e${}^{-}$ for the standard 33 ms integration time, while the signal released by a minimum ionizing particle crossing the sensor is in the range of hundreds e$-$, even for the smallest thickness of the epitaxial region, namely 2-$3\mathrm{\mu }\mathrm{m}$ (Servoli et al., 2013). Hence a Signal/Noise exceeding 30 could be easily obtained and flux measurement based on the identification of each electron on the matrix could be performed. The limitation of this method lies in the possible superposition of signals when the flux increases. There are two types of effects in this case: a pixel may have a superposition of signals from different electrons, resulting in a signal over the threshold (fake electron); two electrons may hit two adjacent pixels resulting in just one cluster with one relative maximum (reduced detection). Overall effect is a sublinear behaviour that requires a non-linear calibration. A possible method attaining a precision better than 1% in the flux determination, even in the presence of intense fluxes of charged particles at therapeutic accelerators, has been proposed (Servoli and Tucceri, 0000, Servoli and Tucceri, 2016). In this work we will explore the capability of such highly segmented devices to characterize precisely ionizing radiation beams, extracting the information on their variability in time and homogeneity in space.