In the clinical context microscopy still plays a pivotal role in analysing cells with a range of downstream analyses available. The choice of which is usually dependent on the information required, such as the level of protein expression, genetic marker identification or cell population estimation. Many cell analyses are optical in nature, but use either bulk properties that are unspecific or where measurement is based on pre-labelling the sample. Alternatively, Raman spectro-microscopic approaches are highly attractive for cell diagnostics as they provide a molecular fingerprint of a cell that is very sensitive to the cell micro-environment, is label-free and can be non-destructive. There is however common understanding that the rate of signal generation is too low for use in high-throughput scenarios, such as for clinical screening and diagnosis. Broadband CARS (BCARS) is a spectroscopic technique that probes the same molecular vibrations as Raman spectroscopy using coherent excitation by employing focused ultrafast laser pulses. The resulting signal obtained can be orders of magnitude stronger than conventional Raman scattering because it is a nonlinear effect. In this thesis, BCARS is investigated as a tool for single cell imaging, with the ultimate goal of label-free single-cell classification and high-throughput imaging. Exogenous fluorochromes were not used in any of the samples studied in this thesis. In this work, single cell imaging consisted of preparing an adequate cell sample, acquiring a hyperspectral dataset, where each point in the image corresponds to a Broadband CARS spectrum, and finally, interpreting the molecular information. In order to reach this capability, a highly optimised BCARS opto-electronic system was constructed, consisting of a commercial ultrafast laser, a modified microscope and several optical elements. The microscope was rigorously tested on non-biological samples such as pure solvents and microplastics, which enabled the tuning of the optical parameters of the system such as its resolution. After system optimisation much work was done on developing a sample targeting procedure and an automated software program was developed to enable scanning of images. Finally, bespoke data analysis procedures were developed and implemented in several single-cell image studies that were of relevance to clinical diagnostics.