2145 - Measurements and Modeling of Cancer Cell Population Dynamics at Diagnosis, Treatment and Follow-Up

Purpose/Objective(s): Non-genetic phenotypic heterogeneity had lately challenged the classic view of genetic determinacy in cell growing and dying processes. Here, we simplified in-vivo cell dynamics with a logistic cancer growing and stable normal-tissue cell model pretreatment, a unified multi-activation model for the cell-killing during treatment and a growing and arresting subpopulation model at the follow-up.

Materials/Methods: Coupled ODEs of cancer cells and sugar glucose co-existing with stable normal-tissue cells results a logistic growth of cell number: N (t) = N_oK/[N_o-(K-N_o)exp(-rt)] with the initial number N_o, carry capacity of K and growth coefficient of r measurable from three times’ pretreatment images which can predict cancer cell number N_X at treatment time. During the relative short course of treatment, cell subpopulations satisfying a set of ODEs of dN/dD = -cNT and dT/dD = dcNT with T being the integral toxic concentration (e.g. free radicals) per each of total m-fractions, resulting survived cell number at EOT: N_E =N_X[n/[1+(n-1)exp(D/D_o)]^m with n and D_o estimated from measured cell survival curves. After treatment, a surviving cancer cell would either actively grow as Ng ?Ng + Ng or become an arrested cell as Ng ? Na so that coupled ODEs of dNg /dt = (g– a)Ng and dNa /dt = a*Ng would predict the total number of cancer cell at the follow-up N_F = Na+Ng = N_E*{a/(g-a) exp[g-a)t_F] - a/(g-a)} with the arrested rate, a, and growing rate, g, measured with two or more times’ follow-up data

Results: Some model values to various tumors with diagnostic PET images are listed in the table. We had found (1) fast growing GTV with high PET SUV would had N_X approaching to a stable value of K and the GTV quickly shrunk on the late treatment or follow up images; (2) Late-stage CTV had non-uniform cancer cell density, although GTVs of the H&N and rectal masses were removed prior to irradiation, the residual cancel cells grew slowly with dominant normal tissue or arrested cancer cells and response slowly to radiation; and (3) follow-up images had already shown recurrence after conventional fractionated RT with insufficient dose for N_E >1 and SBRT of the small metastatic nodule possibly due to early arrested cancer cells that are persistent to radiation.

Conclusion: Dynamic models of cell populations can reasonably explain and predict the clinical outcomes which would be further verified with AI data mining.