The main goal of this project is to generate novel intelligent automated classification algorithms for the cervical Papa smear sample sets for the implementation of a pilot intelligent system for automated diagnosis aid which will be used in medical clinics and laboratories.

This goal will be achieved by checking through the following specific objectives:

  1. Obtaining a marked image database and a test database
          The main goal of this project can only be achieved after obtaining and processing an image database which is to be used in the following segmentation and classification stages. This database is the starting point for all the research activity. An important issue here is the the quality of images which will have a high impact on the other stages. We will have a strict set of guidelines for obtaining the images. The second challenge related to this database is the number of the analyzed medical cases. We estimate at least 2000 cases will be required.
  2. Developing and implementing new segmentation and feature extraction algorithms
    This is also a very important step in achieving this project's ultimate goal. A possible issue here is identifying the relevant features as close as they are found in the manual diagnosis process. A thorough and accurate feature extraction is required.
  3. Implementing appropriate cytological images classification methods
    This is an important stage in obtaining the final integrated system. A possible challenge is dealing with real-life use-cases and obtaining good detection rates and low false positive rates. Novel approaches will be researched, so that error rates will be minimum.
  4. Testing and validation, optimizing the real-time diagnosis tools
    This specific objective is a key element. The main challenge related to this specific objective is the optimisation of all the components so that the samples can be classified in real-time. The balance between harnessing the available processing power and the classification accuracy will be very important.
  5. Publishing results and transferring the proposed system in common medical practice
    This objective is important because the successful implementation of such a system must be well-known so that a high number of clinics and laboratories will be able to use it. The success of the real-life implementation in clinics and laboratories will be also validated first in the scientific community after publishing our research results.

Project Outcomes:

  • The main result of this project is the intelligent automated diagnosis aid system which will allow for a much more efficient processing of Pap smear samples and a higher accuracy of the overall diagnosis. These also depend on the quality of the marked image database, the feature extraction, validation of the final system, but also on the number of level of competence of all the researchers involved.
  • A marked database - by trained physicians, which will further serve as a training base for the expert system and for the validation database.
  • Extracting a set of features which is able to separate as accurate as possible the samples into classes. The main contribution to this desired result will be the adaptation of existing feature to our specific use-case or the development of novel features.
  • Accurate Classification - in the first stages, we will use two classes (normal/abnormal) (as referred to the current Bethesda criteria), but a multi-class separation will also be tested in the case of the set of samples which had been previously classified as pathological.
  • Dissemination of project results through the publication of articles in professional journals, by creating a dedicated website and participation at conferences and specialized workshops. The contribution of the project results is an essential one, their promotion and exposure being necessary both for the medical scientific community and for the information technology community.
  • The employment of 17 researchers (of which 2 post doctorate researcher) on a period between 4 to 20 months.