MAPSSIC, a communicating MAPS-based intracerebral positrons probe for deep brain imaging in awake and freely-moving rats

— Radioisotope imaging is a powerful tool to understand the biological mechanisms in-vivo, especially in the brain of small animals, providing a significant model to study the human brain. In this context, we have developed and built a pixelated intracerebral positron probe to be embedded on awake and freely moving small animals, typically rats. This pixelated probe will represent a key instrument for neuroscientists to study neural mechanisms and correlate them to behavioral experiments. We describe in this paper the simulations carried out to design the intracerebral sensor, its architecture, and the detection of positrons in a volume with a couple of sensors assembled back-to-back. We also depict the architecture of the wireless acquisition system. Finally, we present the first measurements performed in real-time by this miniaturized probe with sealed radioactive sources and a 18 F solution.


MAPSSIC, a communicating MAPS-based intracerebral
positrons probe for deep brain imaging in awake and freely-moving rats In the context of understanding the brain function, which is altered by diseases such as Parkinson, Alzeimer or by addictions, the correlation of neural activity measurements with behavioral experiments is of major interest for biologists. Small animals constitute preclinical models for this study. Therefore, we have started the development of a miniaturized and real-time intracerebral positrons probe for awake and freely moving small animals, typically rats. Taking advantage of the availability of new technologies and of the expertise we have acquired from former developments [1], [2], we propose a novel, robust, wireless, autonomous and lightweight active pixel positron probe.
We present in the following section the main results of the simulations carried out to validate the intracerebral sensor design with an emphasis on the modeling of the measurements with radioactive solutions (in particular 18 F).
Then we describe the architecture of the sensor and detection of positrons in a volume with the whole probe made of a couple of sensors assembled back-to-back. The wireless acquisition system is also described.
Finally, in the last section, we present the first measurements performed in real-time by this miniaturized probe and compare them to Monte Carlo simulations.

II. MONTE CARLO SIMULATIONS OF THE SENSOR DESIGN
In order to estimate the sensor performance regarding positron detection (referred to as direct detection in Fig 1) and rejection of annihilation photons and Compton electrons (referred to as indirect detection in Fig 1), we have performed Monte Carlo simulations using Gate [3] and studied the sensor performances including efficiency and sensitivity, and the deposited energy.
The pixel sensor circuit is developed in 180 nm CMOS technology, with an 18 µm thick high resistivity epitaxial layer sensitive to charged particles.
The sensor dimensions for the probe must take into account the biological, the electronics and the mechanical constraints. It results in a 12000  610  200 µm 3 sensor with a sensitive volume of 6400  480  190 µm 3 . As shown in Fig 1, this model was placed at the center of a cylindrical and homogeneous aqueous solution of 18 F, 15 O, or 11 C. The simulations were carried out, without accounting for signal processing, nor charge diffusion. For more details about the Monte Carlo model of the sensor and the associated analysis, please refer to [4].

A. Sensor efficiency
The efficiency of the probe given by the count rate of the probe immersed in a 18 F solution is illustrated in Fig 2, where isocontours at 10% and 1% are drawn. It appears that the efficiency of the probe to the detection of positrons is far greater than for the detection of 511 keV annihilation photons MAPSSIC, a communicating MAPS-based intracerebral positrons probe for deep brain imaging in awake and freely-moving rats and Compton electrons. This is what we expect from the sensor, which is designed to determine the local distribution of positron emitters in the brain. The volume bounded by the 1% efficiency isosurface is (17.34 ± 0.05) mm 3 . This is in accordance to the volumes of interest for the rat brain structures, hippocampus, or caudate putamen, which are in the order of 40 and 30 mm 3 , respectively.

B. Probe sensitivity
The probe sensitivities to direct detection Sdirect with 11 C, 15 O and 18 F are summarized in Table 1. We observed a greater sensitivity to direct detection for 11 C and 15 O than for 18 F solutions. This difference arises from the fact that positrons have a larger range in water for 11 C and 15 O than for 18 F. Nevertheless 18 F is the most used isotope and has the best definition in terms of brain structure due to the lower range of its positrons. In addition sensitivity with 18 F has been proven sufficient with the previous probe version [2]. Indirect detection sensitivity Sindirect continuously increases because of the high penetration of 511 keV annihilation γ-rays in water.
With a 18 F source of 5 mm radius, we get a Sdirect over Sindirect ratio of 8,28 ± 0,01, which also make this isotope the most interesting for our application.

C. Deposited energy spectra
The CMOS monolithic active pixel sensor (MAPS) features 16  128 pixels of 30  50 µm². The spectra of deposited energies by positrons, electrons and photons are presented in Fig 3. Deposited energies are maximum for 6.9 keV and 7.9 keV for positrons and electrons respectively, while 511 keV γ-rays involve relatively lower energy depositions. Table 2 summarizes the deposited energy distribution for direct detection. This confirms that the MAPS with a detection threshold in the order of few keV can be suitable for direct detection of positrons.

III. MAPSSIC PROBE DESIGN AND SETUP
At this point the positron sensor has been specified and modeled by simulations. We had then to build a wireless, lightweight, robust, autonomous, biocompatible and reliable acquisition system. For this, we realized the MAPS CMOS sensor circuit, then assembled a couple of these sensors backto-back to constitute the probe, and designed a hybrid flex-rigid printed circuit board (PCB) to interface the probe with the embedded control and communication electronics. Finally, we coated the probe to make it biocompatible. The following sections detail all the design and manufacturing steps of the setup. Fig 4 shows the whole acquisition system.

A. MAPS CMOS circuit
Our positron sensor has been designed in a 0.18 µm CMOS technology on 18 μm thick high resistivity (> 1 kΩ•cm) epitaxial layer. It is composed of 30  50 µm² pixels arranged in a 16  128 matrix and has a power consumption lower than 55 nW/pixel [5]. The pixel readout is a rolling shutter scheme with an asynchronous memorization of the hit in the pixel (1 bit). Each pixel is reset right after it has been readout. The control of the pixel analog biasing and configuration (DACs values) is performed via SPI (Serial Peripheral Interface) protocol.

B. 3D positron sensing and biocompatible coating
The positrons are detected mainly on the epitaxial layer side of the sensor. In order to build a probe able to detect the positrons on both sides while increasing the sensitivity of the probe as well as its robustness, we decided to assemble a couple of sensors back-to-back, as shown on Fig 6. Back-to-back assembly of a couple of MAPS probes wirebonded to a thin hybrid flex-rigid PCB resulted in 3D devices featuring various thicknesses of 400, 500 and 560 µm.
An additional 6 µm biocompatible coating of parylene was then deposited over the whole probe surface to allow for its implantation in the brains (Fig 7).

C. Wireless embedded electronics
As described in Fig 4, the control of the probe is carried out by a computer, which communicates with an embedded electronics through the Zigbee 2.4 GHz wireless protocol. This radio communication is performed via2two radiofrequency Xbee S2C modules from Digi-International. These modules are very well suited for our application since each module embeds two micro-controllers. One micro-controller is programmable and therefore set up to control the probe and readout the pixels, the other one is dedicated to the RF communication.
A battery and a power management is also embedded on the animal to provide energy. All the embedded electronics weights less than 20 g.
The micro-controller program is written in C for the implementation of low level instructions. The computer software has been developed in Python for the Human Machine Interface in order to control the probe and manage data handling.

IV. MEASUREMENTS RESULTS
To validate experimentally the probe, we have measured the counting rate of the probe with different sealed radioactive sources and its sensitivity in a radioactive solution.

A.
Radioactive setup Fig 10 bellow describes the setup for measurements with the probe immersed in a 18 F water solution. The probe flex connector is plugged in the corresponding male part with an interface board connected to the micro-controller PCB.
In the case of measurements with sealed radioactive sources, we used a trusquin to place the sources at a given and stable distance from the probe in its transportation tool (Fig 9).  With a pixel integration time of 224 ms, we measured a dark noise of 5.5  10 -3 cps/mm² with a few noisy pixels removed. The mean pixel cluster size (typical for beta particles interactions) was 3,87 pixels.
We have also immersed the probe in a 18 F water solution and measured a source decay half-life of (110.7 ± 0.8) min., thus assessing the probe linearity (Fig 12). However we experienced important instabilities leading to strong disparities in the count rate from one probe to the other.
These two surprising results are possibly explained by a problem on the sensor digital inputs/outputs pads, which suffer from a very important leakage current (tens of mA). This resulted in increasing the temperature of the sensor up to 60 °C, which affected the performances of the chip. A new version of the MAPS CMOS circuit is currently under design to correct this problem.

V. CONCLUSION
Behavioral neuro-imaging is required to address the limits of anesthesia model and to correlate neuro-imaging with behavioral studies. In that context we developed MAPSSIC, the first CMOS active pixel sensor device for real-time positron imaging of freely moving small animals.
Monte Carlo simulation studies were performed to design the probe, which was then fabricated together with the associated wireless control and acquisition system.
Measurements carried out with sealed sources and also with the probe immersed in a radioactive solution have demonstrated that MAPSSIC can provide a robust tool for invivo imaging.
The short terms actions are the design of a new CMOS circuit to correct for high current on digital pads problem encountered with the first version of the probe that resulted in a significant probe temperature elevation.