Signal Analysis of Raman Spectrogram for Bacteria Detection

Gierad Laput, David Sant, Greg Auner (2008)

As a research volunteer at Wayne, I had an opportunity to support researchers in the engineering and medical fields. The researchers at the Smart Sensors and Integrated Microsystems (SSIM) group at Wayne needed a faster and more efficient way to analyze hundreds of bacteria samples digitized using Raman spectroscopy. Given that the tools available were fragmented and difficult to use, the researchers wanted a tool that had the analytical capabilities of Matlab and the familiarity of a desktop application (MATLAB had a built-in GUI framework but it was inadequate to support the group's needs).

BacLearner was a developed to directly address this need. By using C# as the user interface framework and calling MATLAB functions to perform analysis via Dynamic-Link Libraries (DLL's), BacLearner was able to incorporate the best of both worlds.

A glimpse of the process and the developed UI is shown below.

A sketch of how BacLearner was designed at a functional level
Lo-fi Prototype
BacLearner Mid-fi Prototype
Raw data plot after being loaded from BacLearner.
After pre-processing is applied, data becomes smooth, normalized (uniform), and ready for further processing (via SVM, DFA, etc.)
This is how a Raw data file looks like. Wavelength vs. Intensity.
BacLearner can apply pre-processing to the raw data.
Sets of raman data can be grouped accordingly.
A session can be saved into a workspace, and results can be exported into accessible formats.

BacLearner introduced features such as loading/saving workspaces, grouping, and exporting results in multiple formats. Biostatisticians, medical doctors, and research scientists at SSIM are using BacLearner today. Future versions of BacLearner will incorporate features that will ultimately support real-time bacteria classification.

ยป Related Links and Keywords