DNA Quantization
ADCIS and the Francois Baclesse Cancer Centre in Caen, France, jointly developed a turn-key system to perform DNA quantization and Ploidy analysis. These two tests are performed routinely by laboratories involved in cytopathology in oncology. Starting from image processing algorithms on an existing UNIX system [1], ADCIS ported all the image processing algorithms into the Aphelion™ environment on a PC running Windows. Porting this application to the PC with Aphelion™ improved the system performance and ease-of-use, while dramatically reducing the cost of the system by using the standard PC environment.
The system includes an optical microscope equipped with an automatic stage able to move in the X, Y and Z directions, a black and white camera mounted on the microscope, and a PC running Windows with a standard frame grabber board.
The image processing techniques involved in the system are rather complex, and can be described in the following steps:
- Acquisition of a series of images from the sample placed on the automatic stage. The motion of the stage is fully controlled by the application, with a set of predefined tasks that can be called from a user friendly environment based. Wizards are used to make it easy for the operator to re-configure the system.
- Live images are calibrated using a technique based on Optical Density Analysis, to insure that the densitometry of the cells is in the same range.
- Image segmentation to generate a set of small images each containing one cell, and a spreadsheet holding cell characteristics.
- A set of 19 parameters based on the shape, intensity, and texture are computed, and the measurements are saved in a spreadsheet. In this spreadsheet, each cell is now represented as an object in an Aphelion™ ObjectSet table.
- A manual classification is performed only once for any cancer location under study to sort cells into various categories according to their nuclear morphology: normal epithelial cells, lymphocytes, stromal cells, and abnormal epithelial cells with non standard size and texture.
- A model is derived from the training set, using techniques based on principal components analysis.
- In routine use, automatic classification is performed using the same statistical techniques.
- After analyzing the whole sample, the various ploidy histograms of well identified normal or abnormal cells are displayed in the Graphical User Interface.
During the development of the software, all results were matched against the results obtained with a flow cytometry system, pointing out the main benefits of image analysis: identification and elimination of debris and unwanted stromal cells, analysis of fixed formalin and paraffin embedded samples.
The following screen captures present two of the windows in the analysis.
The custom engineering work performed by ADCIS engineers included the definition of the GUI, the implementation of the image segmentation algorithms, the full control of the stage of the microscope, and the development of the classification module, which is now available as an ActiveX component. The stand-alone application was developed using Visual Basic, calling for the Aphelion™ ActiveX components and Toolkits.
Future development on the product will include more advanced statistical analysis, such as dispersion and dynamic clustering.
This application demonstrates the power of the Aphelion™'s ActiveX components in use. The system was developed quickly and is easy to maintain and expand as needed.
ADCIS and the GRECAN at Francois Baclesse Cancer Centre are also currently working together on related biology projects such as the immuno-marker analysis, and other original tools dedicated to experimental and clinical pathology.