Single cell analysis is critical for understanding cellular activities, diagnosing clinicaldiseases, and designing personalized treatments. However, the detection of single cells with high sensitivity has been challenging, especially for clinical samples, as targets of detection are often immersed in extremely complex background. Due to the lack of single cell sensitivity, current mainstream approaches isolate the cells and increase cell numbers by culturing, which is time consuming and often leads to the change of cellular population composition and the loss of native characteristics. In addition, the ensembled detection approaches provide only averaged information of the cell population, thereby missing vital cellular heterogeneity information. The applied probes during detection can also alter the native structures and influence the reliability of the results. In this dissertation, novel label- free optical imaging methods for single cell analysis of raw clinical samples are developed and described to address these challenges. First, a large volume imaging platform is developed for rapid diagnostics of clinical samples of critically low bacterial concentrations without enrichment. Both dual channel and multiplexed versions of the platform are introduced for continuous, detailed monitoring and high throughput minimum inhibitory concentration determination, respectively. With these platforms, the susceptibility of the pathogenic microorganisms in raw urine and blood samples are rapidly quantified within 90 minutes and 240 minutes, respectively, significantly improving the diagnostic time. Second, the large volume imaging platform is adapted for rapid drug susceptibility testing of multidrug-resistant mycobacterial species. Using this method, the susceptibility profile of Mycobacterium tuberculosis and nontuberculous mycobacteria are ascertained within a short timeframe - less than two proliferation cycles. By coupling with single particle tracking, the presence of resistant subpopulations at the therapeutic failure limit of 1% can be detected within one day. Last, by sensitively tracking the emergence of precipitation in various polymer solutions with an upper critical solution temperature upon heating using plasmonic scattering microscopy, precise temperature control over the highly localized plasmonic field is achieved. TRPV1 channel is accurately activated without the aid of an external temperature controlling platform, highlighting the capability of the method for single cell manipulation and in-depth analysis.