Self-supervised hybrid neural network to achieve quantitative bioluminescence tomography for cancer research

Bioluminescence imaging (BLI) has been widely used to visualize biological events in animal models due to its superior soft tissue contrast. Despite its advantages, the conventional surface BLI is limited in topographically localizing internal sources, like bioluminescent tumors, and is not able to provide quantitative information such as tumor volume for treatment assessment. Bioluminescence tomography (BLT) improves upon commonly-used 2D bioluminescence imaging by reconstructing 3D distributions of bioluminescence activity within biological tissue, allowing tumor localization and volume estimation—critical for cancer therapy development. However, conventional model-based BLT is computationally challenging due to the ill-posed nature of the problem and data noise. We introduce Self-supervised Hybrid Neural Network (SHyNN) that integrates the strengths of both conventional model-based methods and machine learning (ML) techniques to address these challenges. The network structure and converging path of SHyNN are designed to mitigate the effects of ill-posedness for achieving accurate and robust solutions. Through simulated and in vivo data on different disease sites, it is demonstrated to outperform the conventional reconstruction approach, particularly under high noise, in tumor localization, volume estimation, and multi-tumor differentiation, highlighting the potential towards quantitative BLT, making it a reliable and promising imaging modality for pre-clinical cancer research.

Speaker bio: Beichuan Deng received his Ph.D. degree in mathematics from Wayne State University in 2018. He subsequently held postdoctoral scholar positions at Purdue University and Worcester Polytechnic Institute. He has been working as a postdoctoral researcher at Department of Radiation Oncology, UT southwestern, since 2023. 

Beichuan’s research specializes in numerical methods and machine learning methods for solving partial differential equations, inverse problems and their related applications. His current work focuses on developing advanced machine learning algorithms for accurate reconstruction in quantitative bioluminescence tomography (QBLT), aimed at enhancing pre-clinical cancer research.