1. Introduction

Positron emission tomography (PET) is a nuclear medicine imaging technique, which utilizes positron-emitting radionuclide labeled molecules to reveal biological and physiological information of human body or animals.

Before PET scan, a short-lived positron-emitting tracer isotope is injected into the living subject. The tracer is chemically incorporated into a biologically active molecule. As the radioisotope undergoes positron emission decay, it emits a positron. After traveling up to a few millimeters the positron encounters an electron, and the two particles annihilate, producing a pair of 511 KeV gamma-ray photons moving in opposite directions. These two gamma-rays, if detected by surrounding detectors of the PET scanner within a timing window, will be recorded as a line of response (LOR). Thus, after many LORs are collected, the original distribution of the radiopharmaceutical could be obtained and corresponding biological information could be analyzed.

As a unique molecular imaging tool in cancer diagnosis, radiation therapy and small animal imaging, PET system with high spatial resolution and high sensitivity is demanded in both clinic and research applications. The most important factors affecting spatial resolution of PET system are detector pixel size and electronic readout method. Therefore, we have proposed two front-end electronics themes which could be used for individual readout for pixellated detector based PET system to achieve high spatial resolution (<2mm).

2. Detectors for high resolution PET

There are mainly three types of radiation detectors available for PET system, Gaseous detector, Semiconductor detector and Scintillation detector coupled to a photodetector.

Based on all overall characteristic we recommend to use scintillator and photodetector for high resolution PET system since the 2mm pixel size are already available in mature commercial products. Both phoswich and one to one coupling scheme with pixellated scintillator array and pixellated APD could achieve the desired spatial resolution.

3. Front-end scheme for high resolution PET

As it was concluded, the most effective method to improve spatial resolution of PET system is to reduce the detector pixel size and use individual readout method [1]. However this will make the PET system require a very large scale multi-channel readout system, whose cost, power consumption and reliability become a big challenge.

Therefore, we have proposed two front-end electronic themes for proposed PET detector. The waveform sampling based front-end electronics is able to perform PSD function [2]. Therefore it is suitable for phoswich detectors. At the meantime, time-over-threshold method is able to achieve much higher channel density, which is suitable for one to once coupling type detectors.

3.1 Waveform Multiplexing

Although all channels within multi-channel system are usually designed with same configurations, characteristics variation always exists. This can be caused by fabrication process or detector uniformity. Therefore it is feasible to use some variable design parameters for tagging waveforms and multiplex signals from different channels. The demultipelxing function can be performed by analyzing the tag information. Parameters like rise time, decay time and gain polarity can be used as tag information for this purpose.

The multiplexing method we have implemented has chosen decay time as tagging information. The multiplexed preamplifiers are charge sensitive preamplifiers, and the decay time can be varied by adjusting the feedback resistance. These preamplifiers are connected together, amplified by a gain amplifier, digitized by an ADC, and then encoded by digital circuits. The digital signals will be analyzed by FPGA to identify decay time given by

Where Tpeak is the peaking time and Tthreshold is the time of half peak in decay edge.

While this structure is simple to implement, the noise degradation will happen because noise components from multiplexed preamplifier will accumulate. Assuming there is no correlation noise, the equivalent noise charge (ENC) is give by

3.2 Pulse-width based readout method

3.2.1 Time-over-Threshold method

In time-over-threshold (ToT) readout approach, the charge signals from detectors are amplified by a charge-sensitive preamplifier and then shaped by a semi-Gaussian shaper. The shaper output, the peak amplitude, is a linear function of Q [3]. If the signal at the shaper output is sent to a comparator with a preset threshold (Vth), a pulse is generated at the output of comparator, whose width is equal to the time during which the shaped signal exceeds the threshold. ToT has fast signal processing, easy implementation and less transmission lines. The Data Acquisition (DAQ) System is easy to implement with just FGPA instead of ADC. What is more, the output signal is already digital signal, which enables more digital processing feature.

3.2.2 Dynamic Time-over-Threshold

During the study of ToT, we have noticed that the linearity of ToT is a limitation of this powerful method. Therefore we have proposed an improvement for this method.

In the new approach Dynamic ToT, once shaper output surpass the base threshold level (Vref). The monostable multivibrator will be triggered by comparator output and feedback to threshold voltage through RC circuits, which leads to a rapid increase for threshold voltage. With this method, we have realized a self triggered multi-level ToT without extra encoding or decoding because the Dynamic ToT signal could be processed same as normal ToT signal.

There are two important factors of using Dynamic ToT method. First, there must be a delay time for threshold increase, which was achieved by multivibrator. Without this delay time, the low amplitude signals cannot be recognized because the threshold voltage will immediately increase to a quite high level. Second, the shaper output width and threshold increase timer (τ=RC) must match so that the threshold increase is significant before the shaper signal decay to a low level.

4. Experimental Results

4.1 Waveform Multiplexing

We have characterized the 32-channel multiplexing preamplifier ASIC and tested it with 2mm pixel LYSO-APD detector. The two multiplexed preamplifier was set with 10 μs (fast preamp) and 30 μs decay time (slow preamp).

When used for readout LYSO-APD detector irradiated with Na-22 source, the two preampllifer output was captured by digital oscilloscope separately, and the waveform length is clear enough for pulse shape discrimination.

When testing the gain linearity with ORTEC 419 pulse generator, the two multiplexed channel has showed 0.84 mV/fC for slow preamp and 0.9 mV/fC for fast preamp and the linearity is < 0.5% for both preamplifiers.

The rise time with no input capacitance was 35 ns for preamplifier with 10 us decay time and 40 ns for preamplifier with 30 us decay time. When applied a 10 pF input capacitance, the rise time is 48 ns and 60 ns correspondingly.

The Na-22 spectrum was obtained when APD was biased at 380 V (gain~70) and shaping time was set at 0.5 μs. The energy resolution of 511 keV annihilation peak was 16.8% for fast preamplifier and 18.3% for slow preamplifier.The optimum ENC with no input capacitance was 1225 e- FWHM for preamplifier with 10 us decay time, and 1334 e- FWHM for preamplifier with 30 us decay time. It is obtained with 0.5 us shaping time.

The noise characteristics with different detector capacitance were also measure for both preamplifiers. (Fig. 10) The noise slope is 81e-/pF at 0.5 us shaping time for fast preamplifier and 120 e-/pF for slow preamplifier.

The multiplexing waveform sampling ASIC has been tested to verify the function of waveform multiplexing. Of the two multiplexed preamplifier, one is set to have an 800 ns decay time (fast channel), while the other is set to 1.6 us decay time.

By injecting same voltage step pulse, gain linearity of both slow channel and fast channel have been measured separately. The gain is 0.94 mV/fC for fast channel and 0.96 mV/fC for slow channel. Nonlinearity of both channels is 2.5%.

The digitized signals were record by logical analyzer and then reconstructed by computer program. As is shown in Fig. 11, the decay time of fast channel and slow channel can be identified and hence the channel information can be obtained.

4.2 Dynamic Time-over-Threshold

For linearity measurement, a pulser was used to inject charge through a 1 pF input capacitor. The charge signal was processed by ASIC charge sensitive preamplifier and shaped by semi-Gaussian shaper, which were set with same parameters as M-MSGC readout. The gain of ASIC preamplifier is~0.75 mV/fC, and the shaper has a gain of~25 while the shaping time was 3 uS.

According to the experiment results, the linearity of Dynamic ToT has shown significant improvement over normal ToT method in the over all dynamic range. When input charge was from 10 fC to 100 fC, the coefficient of determination (R-Squared) for DToT linearity (Vth=Dynamic) is~0.94.

We have also used our dynamic ToT method to readout a MSGC plate, which was irradiation by 5.9 keV X-ray. The reconstructed spectrum has clearly show both the photon peak and escape peak (Fig. 13).

5. Summary and Conclusion

Waveform multiplexing has been successfully implemented and demonstrated with 2-to-1 multiplexing on front-end ASIC. This method is very effective on improving front-end ASIC designed for phoswich detected based high spatial resolution PET system.

Time-over-Threshold method with digital multiplexing has been validated, which has great potential on high spatial resolution PET system.

Dynamic Time-over-Threshold method has been proved. This method is effective on improving the linearity and dynamic range. Dynamic ToT has the potential to replace ADC in front-end signal processing.

Fig.1 Principle of PET Imaging

Fig. 2 Proposed high resolution PET detector

Fig. 3 Scheme of Waveform Multiplexing

Fig. 4 Scheme of Time-over-Threshold

Fig. 5 ToT implementation with digital multiplexing

Fig. 6 Experiment result with ToT readout

Fig. 7 Scheme of Dynamic Time-over-Threshold

Fig. 8 Waveforms of multiplexed channel for reading out APD detctor

Fig. 9 Na-22 spectrum with APD biased at 380 V with both preamplifier

Fig. 10 noise characteristics of 2 preamplifier within multiplexing preamplifier ASIC

Fig.11 Reconstructed waveform from waveform multiplexing ASIC

Fig. 12 Linearity of ToT and DToT

ポジトロンエミッショントモグラフィ(PET)は、放射性薬剤を用いることで、体内の機能情報を可視化することに用いられ、近年多くの成功をおさめている。しかし、こと分解能となると、代表的なPETにおいては、数mmにとどまっており、X線CTやMRIなどの他の医用イメージング手法と比較して、劣っている。本研究では、この点を改善し、高分解能のPETを実現するために必要な要素技術として、γ線検出器からの新しい信号読み出し手法の研究を行ったものであり、特に、従来、見落とされがちであったディジタル信号に含まれる時間軸の情報の有効な利用について詳細な検討を経て、新しい信号処理システムの提案を行い、その有用を示したものである。本論文は、以下に示す7章から構成されるものである。

第一章は、序論であり、本研究の背景と研究に至るモティベーションを示したのち、PETの現状を紹介し、分解能という点でまだ不十分とはいえず、大いなる研究の必要性があることを示している。その後、高分解能化における本研究の位置づけについて、特に多重化が重要であり、それを実現するために時間領域のディジタル信号を用いることが有効であるということを示している。

第二章では、PETの一般的な説明を行っており、PETの歴史から紐解き、PETの原理、また、放射性薬剤と応用分野、さらにはPET-CTなどのマルチモダリティ化について述べている。

第三章は、高分解能PETの実現に焦点を絞り、前章で述べたPETの原理における空間分解能特性との関連について物質と放射線との相互作用、検出器の動作原理とシステム構成などの諸観点から、詳細に議論を行っている。また、装置の直径や検出器結晶のサイズなどのパラメータと空間分解能特性のマップを示したのち、視野周辺における分解能における深さ方向の位置検出(Depth of Interaction: DOI)の重要性について議論している。

第四章は、PETにおけるフロントエンドエレクトロニクスについてまとめたものであり、本研究において前提として扱っている、集積回路技術の基礎からはじめて、現状技術のサーベイを経て、新しい高分解能PETの原理とシステム構成について波形サンプリング手法と時間幅の利用という2つの可能な手法の提案までを行っている。

第五章は、前章で提案された波形サンプリング手法によるフロントエンドASIC (Application Specific Integrated Circuit) の開発について述べたものである。初めに、ASICの内部構成として、プリアンプ、VGA,高速ADCとメモリからなることを示したのち、プリアンプASICの内部回路構成と評価結果について示し、その性能が十分であることを実験的に示している。その後、波形サンプリング手法に用いられるASICの各部の性能として、十分な評価結果が得られていることを示している。さらに、この波形サンプリングシステムの高分解能化のために、プリアンプの減衰波形の時定数を各チャネル毎に変化させることで、複数のチャンネルを測定し、デコードすることが可能であることを実証している。

第六章では、時間幅情報を用いたPETフロントエンドについての議論を行っており、シンプルな回路構成ゆえに、極めて多くのチャネルを扱うことが容易である特性に注目し、特に波高値とパルス幅の関係における直線性についての議論を行っている。これらの間の関係の改善には、動的にしきい値を変化させる手法が有用であることを見出し、その特性の評価を行っている。これにより、本手法の実用化に一歩近づくことができることを示している。

第七章は、結論であり、本研究において初めて示された新しいマルチプレクシング法と動的なしきい値の変化などの手法が、高分解能化において、極めて有効に機能し、高分解能PETの実現に一歩近づけたとまとめている。

以上のように、本研究はPETの高分解能化を目指して、新しい信号処理法の開発について、示したものであり、その成果は大規模・高分解能γ線計測システムとしての新世代PET実現にむけて極めて重要な分野をカバーするものであり、工学、特に原子力工学の進展に寄与するところが少なくない。よって本論文は博士(工学)の請求論文として合格であると認められる。