When light is incident upon tissue, imaging contrast can be obtained from a range of interactions including absorption, scattering and fluorescence. Clinical optical imaging systems are typically optimized to report on a single contrast source, for example, using a series of filters to extract fluorescence emissions. Hyperspectral imaging has the potential to overcome the need for specialized instrumentation, by sampling spatial and spectral information simultaneously. In particular, multispectral filter arrays (MSFAs) now monolithically integrate spectral filters with CMOS image sensors to provide a robust, compact and cost-effective solution to video rate hyperspectral imaging. However, MSFAs suffer from a significant limitation: the inherent trade-off between spatial and spectral resolution. Therefore, the properties of the MSFA including the number of filters (spectral bands), their wavelength and bandwidth, needs be optimized for tissue imaging. While many approaches exist for optimizing spectral bands, none consider practical challenges such as manufacturing constraints and tolerancing. To overcome this, we have developed a framework for spectral band optimization for MSFAs that considers the constraints of our fabrication process, including establishing tolerances. Our approach shows early promise for fabricating MSFAs with appropriate spectral filters.