Faculty of Electronic Engineering, Niš, Serbia

The paper is available at the following link: PDF

Abstract:

This paper presents a hybrid methodology combining circuit simulation, modelling and machine learning (ML) for optimizing X-ray detector front-end electronics. The said circuit consists of a PIN photodiode-based charge-sensitive preamplifier (CSP), and leverages LTSpice simulations to establish a baseline performance and derive a parameterized noise model. Exhaustive simulation of all and each parameter combinations is computationally prohibitive, Therefore, a Random Forest as ML guided approach was introduced. This method establishes a predictive relationship between SNR=f(Rf,Cf,Cdet,Rleak) for the said circuit, providing feature importance metrics to guide optimization. The objective is to maximize the signal-to-noise ratio(SNR) by fine tuning components whilst preserving signal integrity and bandwidth. This case study has results that correspond to approximately 20% increase in signal amplitude and samilar reduction in RMS noise. Improvement such as this could result in dose reduction in medical imaging or enhancing defect detection, demonstrative of ML's efficiency as an accelerator for analog circuit design.

Keywords:

X-ray detector, noise reduction, charge sensitive preamplifier, machine learning, Random Forest, SNR, analog circuit design.

REFERENCES:

[1] Goran S. Ristić "The Digital Flat-Panel X-ray Detectors", Applied Physics Labouratory, Faculty of Electronic Engineering, Univeristy of Nis, Serbia 2018
[2] Yuhe Zhang "Advancing X-ray imaging with deep learning", Department of Physics, Faculty of Science, Lund University, May 2024
[3] Glenn Knoll "Radiation Detection and Measurement", 3rd ed. Hoboken, NJ, USA: John Wiley & Sons, 2000
[4] Helmuth Spieler "Front-End Elctronics and Signalprocessing", Lawrence Barkeley National Laboratory, Physics Devision, Berkley, CA 94720, USA
[5] Hamamtsu Photonics "Technical Information: SDD and PIN Photodiodes for X-ray Detection",
[6] Yu-Yu Chan, Shao-Yun Fang "Optimizing Anlog Circuit Design Through a Machine Learning Assisted Evolutionary Algorithm", Department of Elctrical Engineering, National Taiwan University of Science and Technology, Taipei, Taiwan, June 2025
[7] Budak, Ahmet & Zhang, Shuhan & Liu, Mingjie & Shi, Wei & Zhu, Keren & Pan, David "Machine Learning for Analog Circuit Sizing" 10.1007/978-3-031-13074-8_12, 2022