Evaluation of a large-area double-sided silicon strip detector for quality assurance in ion-beam radiotherapy

TL;DR

The paper validates the fIVI Range Monitoring System, a large-area, two-layer silicon-strip tracker designed for quality assurance in ion-beam radiotherapy. It combines CBM STS sensors with fast front-end readout and a scalable GBT-based data chain to operate at clinically relevant energies and rates, achieving rates up to $1.3\mathrm{MHz}$ with minimal pileup and a timestamp resolution of $6.25\mathrm{ns}$. Geant4 simulations and beam tests with 19 MeV protons and 206.9 MeV/u carbon ions demonstrate high tracking efficiency ($>90\%$ of interactions form unique hits contributing to tracks) and precise beam-spot reconstruction, supporting the system’s suitability for Bragg peak range monitoring at conventional dose rates and potential applicability to FLASH scenarios. The results indicate the approach is scalable and robust, with actionable insights for bandwidth optimization and threshold calibration to maximize performance in clinical settings. Overall, the study presents a viable, high-rate detector solution that could enable more precise dose delivery and reduced healthy-tissue exposure in radiotherapy.

Abstract

Designed to provide quality assurance for ion-beam radiotherapy, the prototype fIVI (filtered Interaction Vertex Imaging) Range Monitoring System is a two-layer tracker which employs double-sided strip-segmented silicon detectors. To meet the high demands of a clinical environment, a large sensitive area is required, along with a fast and compact readout. As this device utilizes sensors and readout electronics adapted from particle physics, where the expected energy and count rate differ significantly from radiotherapy, validation was necessary to ensure that these sensors would function effectively at the order 100 MeV/u energies and order MHz count rates expected during clinical irradiation. Tests were conducted using scattered subclinical 19 MeV protons at high intensity, and clinical 207 MeV/u carbon ions at low intensity to independently validate these variables. The detection system is found to operate at rates up to 1.3 MHz, with a negligible fraction of events being affected by pileup. The efficiency of hit reconstruction is high, with a timestamp resolution of 6.25 ns, and a coincidence window of 31.25 ns, as is required for clinical event rates. With these settings, over 90% of particle interactions are able to reconstruct unique hit positions and contribute to track formation. This device is the first system using large-area, high-resolution detectors which meets the demanding count rate requirements associated with clinical radiotherapy.

Figures (16)

• Figure 1: (a) Representative schematic of a sensor, showing the kite-shaped elements formed by the intersection of axial and angled segments. Two interaction positions are shown (yellow), along with the active segments (blue, red). Note the Z-strip angled segment (blue) which wraps around the lateral edge of the sensor. Due to the use of angled segments, there is no confusion between the two hits. In the case of orthogonal segments, these two interaction positions would instead produce four candidate positions, as each segment on one side intersects with every segment on the opposite side. (b) Detail view of a Z-strip segment, with the two disjoint sensitive regions (red) joined by an electrical interconnect (blue). This connection allows the entire sensor to be read out along the bottom edge, while maintaining position sensitivity across the entire sensor area.
• Figure 2: Complete [product-units=power]60x120 module. The sensor (right) was affixed to the backing at its four corners. Only the 1 margin of the sensor guard ring was directly supported by the backing; the portion corresponding to the sensitive area was removed. The sensor and backing were mounted 35 in front of the rear frame. A shielded bundle of analog microcables (center, yellow) connected the sensor and the front end electronics (left). These cables were held in place by two clamps (white). The front frame was separated from the rear frame by 50.
• Figure 3: Both modules installed in the tracker. (a) Covering panels removed, showing the modules mounted to the base plate, with the sensors separated by 12. (b) Covering panels installed, with the exception of the entrance window. For data collection, this window region was covered with a 16 thick aluminum foil. The region of the base plate exterior to the enclosure was used to affix the tracker to the experimental setup.
• Figure 4: Front-end board used for readout of one side of a silicon sensor. Eight SMX ASICs are mounted in two staggered rows on the center of the board. Communication occurs through a ribbon cable inserted into the socket on the left, which provides a common clock distributed to all ASICs, a common control interface which can address commands to an individual ASIC or to all ASICs simultaneously, and 16 independent data uplinks (two per ASIC). The eight-pin header on the upper right provides low-voltage power for the analog and digital electronics on each ASIC, as well as a bias voltage for the connected sensor.
• Figure 5: GBTX Emulator board. Connections to front-end electronics are made via ribbon cables connected to the three sockets on the left. Only the upper two sockets are active during data collection, with the uppermost socket supporting 16 simultaneous uplinks, and the middle socket being restricted to 12. The Artix 7 FPGA is mounted to the underside of the board, in the center. Power is provided through terminals on the lower right edge of the board. Upstream interfaces are visible on the right, with the active optical link in the lower cage. The second optical link and the ethernet interface for slow control are not used in this setup.