Experimental method and statistical analysis to fit tumor growth model using SPECT/CT imaging: a preclinical study

Area of Science:

• Oncology
• Biophysics
• Medical Imaging

Background:

• Theoretical tumor growth models have advanced over the last decade.
• Oncology imaging utilizes modalities like CT, MRI, SPECT, and FDG-PET.
• Extracting quantitative biophysical parameters from serial tumor images can inform treatment plans.

Purpose of the Study:

• To implement and validate an inversion algorithm for fitting mathematical models to tumor imaging data.
• To extract hidden quantitative biophysical parameters from serial tumor images.
• To assess the accuracy and fit of different tumor growth models.

Main Methods:

• Implemented and applied a theoretical ordinary differential equation (ODE) compartment model and its variants in vivo.
• Developed an inversion algorithm to fit tumor growth models to simulated and experimental data.
• Acquired serial SPECT/CT scans of mice breast tumors and used SPECT data for proliferating layer segmentation.

Main Results:

• Successfully recovered 5 out of 7 parameters with <4.3% error in noisy data simulations, including a 0.07% error for tumor growth rate.
• Achieved high R-squared values (0.99 for logistic/Gompertzian, 0.96 for surface area) for P-layer volume fitting in vivo.
• Akaike Information Criterion (AIC) weights indicated Gompertzian (~0.57) and logistic (~0.43) models fit better than surface-area models (~0).

Conclusions:

• Model-fitting to in vivo mouse tumor studies confirms the feasibility of extracting quantitative features from imaging data.
• Gompertzian and logistic growth models provide a better fit for in vivo breast tumors than surface-area based models, according to AIC evaluations.