Percutaneous thermal ablation is an effective treatment for focal tumors but can result in local tumor recurrence associated with insufficient heating of tumor and a surrounding margin of healthy tissue. Real-time, non-invasive temperature monitoring with spectral CT thermometry may reduce this local tumor recurrence rate by providing a volumetric temperature distribution to ensure complete ablation. To develop spectral CT thermometry, a spectral physical density quantification model was developed and implemented to relate physical density to temperature through the principle of thermal volumetric expansion in ex vivo bovine muscle. Liver-mimicking phantoms were also evaluated for model reproducibility and temperature tolerance at different radiation dose levels to meet clinical requirements. A metal artifact reduction algorithm was then designed specifically for the ablation probe and spectral CT data. Finally, microwave ablation was performed on an in vivo porcine liver to assess the feasibility of spectral CT thermometry for non-invasive temperature monitoring under physiological conditions. Spectral CT thermometry was validated in ex vivo bovine muscle, demonstrating thermal volumetric expansion with a Pearson’s correlation coefficient of 0.9781. Liver-mimicking phantoms similarly exhibited a strong relationship between physical density and temperature, which was reproduced in three separate phantoms with a coefficient of variation of 9.6% and 0.08% for the slope and intercept, respectively. Additionally, for slice thicknesses less than 5 mm, radiation dose necessary to meet temperature tolerance requirements was reduced from > 20 mGy to 2 mGy by applying external denoising. Implementation of the designed metal artifact reduction algorithm decreased streaks in the periphery and along the axis of the probe, reducing temperature overestimation by 8.1 ± 3.4 °C. In vivo ablation showed a strong relationship between temperature and physical density with a Pearson’s correlation coefficient of 0.8785. Overall, development and characterization of spectral CT thermometry have demonstrated the utility of its accurate temperature maps, enabling clinical translation of non-invasive temperature monitoring to ensure complete ablation of tumor and ablation margin, and thus reduce local tumor recurrence.