A PRF-Based Multibaseline fitting model for temperature prediction

Abstract

MRI is capable of constructing in vivo temperature profiles to provide precise thermal guidance for benign and malignant diseases and minimally invasive, rapid recovery and highly reproducible treatments for tumors and benign nodules. Despite the real-time thermal monitoring capabilities of proton resonance frequency (PRF) thermometry, this important magnetic resonance-based temperature measurement technique faces challenges in oncology energy therapy because it is susceptible to image alignment errors and magnetic induction field variations, resulting in unreliable temperature tracking. To address these issues, this study proposes a novel multibaseline fitting temperature prediction model based on proton resonance frequency. The model processes k-space data through phase reconstruction and denoising to remove background phase noise. Multiple phase difference maps are then dynamically generated, a weighted least squares fit to the background phase outside the ROI is performed to correct for single-frame phase distortions, and a signal amplification factor (k) is used to compensate for cryogenic proton signal attenuation. The new model and temperature estimation algorithm were tested by varying the water content of the phantom, heating the phantom at high temperature and freezing the phantom at low temperature experiments. Significant improvements are demonstrated by comparative analysis: the root mean square error (RMSE) reduces from 14.62 °C/2.99 °C to 2.19 °C, maximum error decreases from 32.29 °C/5.74 °C to 3.64 °C, and the goodness-of-fit reaches 0.96, outperforming conventional single- and multi-baseline models. Notably, this study validates PRF thermometry's efficacy for the first time in sub-zero environments (down to −9.2 °C), providing a robust solution for precise temperature monitoring in cryoablation therapies. These advancements address key technical challenges in low-temperature precision control, offering significant implications for tumor cryoablation.