Raman signal is very weak signal ,so its very hard to use this system to detect particular component in a mixture,,,,,,,,,,and biological system is very complex

Raman-based techniques are being developed for cancer detection, especially tools that can be used in vivo such as Raman microneedle spectroscopy. Search for "Raman microneedle" or "in vivo Raman spectroscopy" along with phrases related to cancer and a whole host of review and research papers will come up. I would recommend looking at the work of Prof. Nick Stone at Exeter University who's lab developed and validated a Raman microneedle system for detecting soft tissue cancer margins and talks about the issue of sensitivity and specificity for healthy vs cancerous tissue. Many of these techniques don't look for changes in 'concentration' of markers, but rather look out for particular combinations of markers sometimes irrespective of 'concentration'. As the person above said, tissues are incredible complex chemical environments and analysing differences in concentration of trace analytes can be very difficult when autofluorescence in soft tissues can dominate the weaker Raman signals, which would make correlating changes in Raman peak intensity purely to marker relative concentration very challenging. There are numerous advanced Raman methods for getting achieving better spectral sensitivity such as CARS (coherent anti-stokes Raman spectroscopy) and while these have been used to look at cancer tissue samples in vitro, I don't think these methods have been developed enough yet to become practical in vivo tools (do a search on this to be sure..),

Regarding improvements to Raman setups, this really depends on what context you mean. In vitro i.e improving Raman as a research tool to look at cells and tissues in a dish, there are numerous options. There are specialist variations of Raman spectroscopy methods including CARS as mentioned about and Surface and Tip-enhanced Raman which aim to significantly increase the Raman scattering efficiency in a highly localised area of the sample thus giving you better lateral special resolution and spectral sensitivity. The bigger picture view is that while all this is good science, it needs to be translated to the clinic...which brings me to the second context of your question of improvements. The ultimate clinical goal is to avoid having to take samples from patients and send them to labs and be able to probe the tissue and determine the disease state there and then in surgery. This requires engineering of the Raman set ups in labs to fit and be using in theatres. Many innovations are needed (and many already achieved) here. These include engineering of the illumination-detection system into a hand held probe, commercial availability of smaller and cheaper long wavelength laser systems for probing deeper into tissues and reducing autofluorescence effects. The latest focus of research in this area is not actually on the hardware but on the software which gathers and interprets the data. Statistical analysis methods used to determine whether a patter on markers represent cancer or health tissues are being developed to run more efficiently an enable their use on basic PC hardware rather than expensive high end systems. Methods of data reduction are required to aid this process to avoid wasting time processing parts of the spectral data that might add little to the accuracy of the result. I've rather simplified the whole area here, but hopefully this gives you an insight. I'd recommend searching for some recent review papers on the use of Raman in cancer detection and clinical use to get specific details on the topics i've raised here.

@Arunkumar and Richard,...

Many thanks for your suggestions