The Application Research of Multi-Algorithm Fusion Optimization BP Neural Network in High-Precision Measurement of Sensors

Diabetes foot, one of the most severe chronic complications of diabetes, results from the synergistic effects of neuropathy and peripheral vascular lesions induced by prolonged hyperglycemia. This pathological interplay leads to local microcirculatory disorders, which may progress to lower limb infections, ulcers, and deep tissue destruction. In severe cases, patients face the risk of amputation, significantly compromising their quality of life and physical health. The dedicated spectral therapy device for diabetic foot offers a novel non-invasive therapeutic approach by generating deep-penetrating composite spectra capable of reaching subcutaneous tissues up to 5 cm. Realtime and precise temperature monitoring of the foot is critical during spectral therapy, as excessive heat can cause thermal injury and exacerbate tissue damage. Microelectromechanical systems (MEMS) non-contact temperature sensors, favored for their rapid response and ease of use, are commonly employed for this purpose. However, their accuracy is often compromised by environmental temperature fluctuations and measurement distance variability, posing challenges for clinical precision. To address this issue, this study employed a backpropagation (BP) neural network algorithm to compensate for measurement errors in MEMS sensors. Through comparative analysis of BP, GA-BP (genetic algorithm-optimized BP), and PSO-BP (particle swarm optimization-optimized BP) algorithms, the PSO-BP approach demonstrated superior performance in enhancing detection accuracy. By mitigating the inherent limitations of traditional BP networks, such as local optima entrapment and slow convergence, the PSO-BP algorithm provides a robust technical foundation for safe and precise temperature monitoring in diabetic foot spectral therapy.