Optimizing thermal dose prediction in nanoparticle-mediated photothermal therapy using a convolutional neural network-based model

Abstract

Nanoparticle-mediated photothermal therapy (NMPTT) is an up-and-coming targeted cancer treatment. Here, nanoparticles are used to convert near-infrared light into localized heat that can kill tumour cells while sparing surrounding healthy tissue. Nevertheless, variability among tissue properties and distributions of nanoparticles and laser parameters decreases the effectiveness of optimal thermal dosages.

This study presents a CNN-based model for predicting the optimized thermal doses in NMPTT to effectively ablate tumours by relying on input parameters such as nanoparticle concentration, laser intensity, exposure time, and tissue characteristics.

For the dataset used to train the model, the mean squared error was 0.015, the root mean squared error was 0.122, and the mean absolute error was 0.098. The model showed a validation accuracy of 89.5% and a testing accuracy of 87.8%, thus having high predictive accuracy. Its R-squared level at 0.92 exhibits the model's strong generalizability.

The proposed method offers a robust predictive instrument for increasing the precision and safety of photothermal therapy. It, thus, provides practitioners with the means to tailor treatments for each patient. By providing reliable predictions of thermal dose to inform clinical decisions and improve therapeutic outcomes, it may help advance nanoparticle-mediated photothermal therapy.