The human serum uric acid (UA) level is a crucial indicator for diagnosing dementia in middle-aged and elderly individuals, with decreased levels being expressed in patients. Therefore, developing a high-performance sensor for online uric acid detection is of significant research interest. Herein, a microfiber knot resonator (MKR) sensor for quantitative detection of UA levels is reported. Combining polydimethylsiloxane (PDMS) with an adiabatic MKR, the specificity of UA detection in serum is improved using the photothermal effect of Prussian blue-enhanced uricase reaction. During the sensor operation, UA generates the photothermal effect under 650 nm laser irradiation, causing the PDMS film to expand with heat absorption, thereby shifting the resonance spectrum of the PDMS-MKR sensor. Experimental results demonstrate a sensitivity of 0.0559 nm/(μmol/L) within a UA concentration range of 40–120 μmol/L and a detection limit as low as 0.3578 μmol/L. The proposed sensor shows potential applications in clinical point-of-care early diagnosis and prognosis of dementia due to its specificity, fast detection speed, robustness, and high integration.