The basic homogeneous SEIR (Susceptible-Exposed-Infected-Removed) model is commonly used for analyzing infectious diseases. The SEIR model lacks individual-level information like location and distance between susceptible and infected individuals. To address this limitation, a Geographically Dependent Individual-Level Model (GD-ILM) within an SEIR framework was previously developed. We expand the SEIR GD-ILM to accommodate infectious diseases involving loss of immunity, like Tuberculosis, resulting in SEIRS (Susceptible-Exposed-Infected-Removed-Susceptible) GD-ILM for when both regional and individual-level spatial data are available. To overcome computational complexity, we introduce a new approach based on the Monte Carlo Expectation Conditional Maximization algorithm for efficient parameter estimation. Focusing on Manitoba's health regions, we analyze individual-level Tuberculosis data (2011-2018) to predict infection rates over time and identify high-risk geographical areas.