Revolutionizing cervical cancer care: the synergistic effects of hyperthermia and machine learning

This study aimed to examine the impact of hyperthermia in conjunction with concurrent chemoradiotherapy (CCRT) on peripheral immune markers in patients with locally advanced cervical cancer (LACC). Additionally, we sought to predict the associations between these validated immune markers and the efficacy of hyperthermia by employing a developed and optimized machine learning model.

Sixty LACC patients were enrolled, with 30 in the observation group receiving CCRT plus hyperthermia and 30 in the control group receiving CCRT alone. Data on complete blood counts; T lymphocyte subsets (CD3⁺T, CD4⁺T, and CD8⁺T); and IL-1β, IL-6, IL-8, and TNF-α levels were collected before and after treatment. Chi-square tests were employed for significant variable selection, followed by machine learning algorithms for model construction.

The study showed that there was no statistical difference in 2-years PFS (p = 0.241) and 2-years OS (p = 0.435) between the combined hyperthermia group and the chemoradiotherapy group alone, which was considered related to the shorter follow-up time. In the observation group, the neutrophil-to-lymphocyte ratio (NLR), monocyte-to-lymphocyte ratio (MLR), platelet-to-lymphocyte ratio (PLR), and IL-8; and CD8⁺T levels were significantly lower than those in the control group (p < 0.05). Conversely, the TNF-α, IL-1β, IL-6, CD3⁺T, CD4⁺T, CD3⁺T/CD8⁺T, and CD4⁺T/CD8⁺T ratios increased significantly (p < 0.05). Sixteen reliable features (p < 0.05) were selected via chi-square tests to construct machine learning models. Among the nine evaluated algorithms, random forest (RF) and multilayer perceptron (MLP) models demonstrated optimal performance in predicting the efficacy of hyperthermia combined with CCRT for LACC.

The combination of hyperthermia with CCRT alleviates the suppression of peripheral immune indicators, modulates immune function, and improves the systemic immune microenvironment in LACC patients. The RF and MLP models, which are based on dynamic peripheral immune indicators, exhibit robust performance in predicting hyperthermia efficacy.