Nanostructured porous silicon (PSi) has been widely studied for the past two decades as an optical transducer for the detection of various molecules, with advantages of simple fabrication, high internal surface, and unique optical properties. Despite these significant advantages, the clinical implementation of label-free PSi-based biosensors has been impaired by their insufficient sensitivity, usually in the micromolar range for protein and DNA targets. In this work, we investigate the limiting factors of PSi-based optical biosensors and design methods for their improvement. As a model system, we study PSi Fabry-Pérot thin films, functionalized with DNA aptamers, and utilize the reflective interferometric Fourier transform spectroscopy method for real-time and label-free detection of different target proteins. We derive a comprehensive mathematical model, which considers all mass transport and reaction kinetics phenomena in these biosensors. We demonstrate that the model successfully captures target binding rate in these biosensors, contrary to the conventional model used in the literature. The model is used to elucidate the impact of diffusion rate in these biosensors and to develop rules of thumb for their optimization. To enhance the performance of PSi-based biosensors, we study methods for mass transfer acceleration. These include application of isotachophoresis (ITP) for on-chip protein concentration, target mixing on top of the biosensor or microfluidic integration, with up to 1000-fold enhancement in sensitivity. For the latter, we present for the first-time integration of a PSi-based biosensor in 3D-printed microfluidic devices with improved performance, compared to conventional polydimethylsiloxane microfluidics. Finally, we develop a PSi-based biosensor for detection of a relevant protein cancer biomarker and present its selective detection in a highly complex fluid of pancreatic juice.