Ultrashort-Pulsed Laser-Tissue Interactions for Precision Oncology: A Multiphysics Modeling Approach

Current surgical solutions for skin cancer lesions are limited to excision on histopathological evidenced tissue. This is a highly challenging and imprecise process leading to functional and cosmetic sequelae for the patients. Recently, a new approach to tackle this problem based on the use of high-precision ultrashort-pulsed laser irradiation of tissue has been reported. Together with a multi-physics model to study the interactions, simulators have been utilized to analyze the propagation, absorption and scattering of laser pulses in heterogeneous multilayer geometry tissues, and a reworked fast finite element scheme has been used for modeling energy deposition and biophysical effects. First a simpler simulation for normal tissue to obtain a feedback mechanism to adjust wavelength, and irradiance for the respective integuments depth was simulated.

Given that patients will have lesions of different size, an optimization study with the same multi-physical simulator has been developed to optimize parameters such as power and maximum lesions size in lasering skin cancer lesions. Using Monte-Carlo and fast finite element simulator, optimal treatment conditions to obtain a clinically application were associated with recurrent variables. A finite element multi-domain model solves laser irradiation, energy deposition and biophysical tissue response simultaneously. Temperature, density and mechanical stress were computed, emphasizing that pulse duration and model domain size are critical. Post-processing of computed fields provides stress gradients, plastic spots, and ejected tissue shapes. Evaluation of successful treatments requires integration of in-vivo histological expert domain knowledge.

A second mutiphysics simulator analyzes laser parameters and coupling to the first pre-simulated lesion. With Monte-Carlo and finite volumes methods, laser propagation inside a heterogeneous 3D lesions geometrical model was simulated and coupled with rigid fluid mechanics and laser energy deposition inside the tissues. Innovative accelerated multi-parametric simulations were conducted to extensively characterize problem space. A simplicity and computational burden trade-off was found in size comparison. A database of optimal treatment responses was produced, where lesions metrics are clinically relevant variables. A neural network surrogate to rank prior treatments was trained to effectively select candidate treatments.