ENGS 88 Honors Thesis (AB Students)

Degree Program


Year of Graduation


Faculty Advisor

Geoffrey Luke

Document Type

Thesis (Senior Honors)

Publication Date

Spring 6-10-2020


Photoacoustic (PA) imaging is a non-invasive diagnostic imaging technique that gives images of photoabsorbers based on their absorption of optical energy. These optical absorption properties can then be linked to important tissue properties. For the method to be quantitative, however, it is necessary to have an accurate estimation of the light fluence in the tissue. The current gold standard in addressing the fluence estimation problem, a Monte Carlo Simulation, is costly in time and computation. In this work, we developed a deep neural network to quickly and accurately estimate light fluence in arbitrary tissue types and geometries. The network was trained on light fluence estimations from Monte Carlo simulations of light propagation through pseudo-random tissue structures. The network estimated light fluence with only 0.65\% error in less than 2 ms per tissue. We hope these results bring real-time PA imaging closer to clinical relevancy.