As a science concerned with creating artificial systems, computer vision ought to emphasize the evaluation and comparison of alternative methods. It should be simple to compose large systems from components created by other researchers. That this is seldom done is partly due to the lack of a suitable, common software environment. Without a common environment, implementations and data can be exchanged onlu with great difficulty.

Moreover, even though a lot of systems or toolkits already exist to sopport image processing tasks, anly a few of them provides for the easy creation and use of arbitrary data types, such as are needed for many areas of computer vision research. The most famous of them is the software environment called Vista, developed at the University of British Columbia's Laboratory for Computational Intelligence in the early nineties.

Starting from the good ideas introduced by Vista we are designing and developing a new Computer Vision Programming Library written in JAVA and called JVista.

JVista is based on the Java Advanced Imaging API (JAI) that broadens the reach of the Java platform to allow sophisticated, high performance image processing functionality to be incorporated into Java applets and applications.

The Java Advanced Imaging API goes beyond the functionality of traditional imaging APIs to provide a high-performance, platform-independent and extensible image processing framework. For more details you can visit the JAI home page.

This project is just started and more info will be available as soon as possible. Moreover, in the Software section you will find sample programs with the source code to test preliminary versions of JVista.

GNN-CAD is a computer system for the automatic detection and localization of microcalcifications, isolated or in clusters, in digitized mammograms.

The software architecture is based on the multiresolution space-frequency approach to the description of image structure known as Gabor Transform. Specifically, a bank of Gabor elementary functions with varying spatial extent and tuned to different spatial frequencies is used for the extraction of microcalcifications characteristics. Classification is performed by an Artificial Neural Network with supervised learning.

GNN-CAD is an on-going project. The goal is to build a complete CAD system for the diagnosis of breast tumors capable to satisfy the needs of different users, e.g., radiologists, students or expert programmers, in different operating environments, e.g., Linux or Windows.

To this end, particular care has been taken in designing the user interface. The architecture is written in Java, which makes it totally independent by the user platform. It is easily extendable to include in-progress algorithms such as detection of stellate mass. The system contains a minimal set of image manipulation routines in order to meet the needs, for example, of the programmer who can implement new algorithms without leaving GNN-CAD environment. The easy intuitive interface has been designed with the user needs in mind. The system is intended to adapt to the profile of the "normal" user such as the radiologist or the surgeon, by hiding those specific functionalities which are meant for the "expert" user, e.g., the setting of parameters for the training phase or for the visualization of results.

MAGIC-5, a collaboration among Italian physicists and radiologists, starting from the results obtained from the previous projects financed by INFN from 1997 to 2003 (CALMA and GPCALMA) has the following goals:

• to connect the hospitals and the research centers by the use of the GRID technology to share data and tools;
• to develop an automated system to be used as second reader for the detection of lung cancer in CT screening images.

The project was developed between different INFN sections all over Italy: Bari, Cagliari, Lecce, Napoli, Palermo, Pisa, and Torino

Here are some results:

PROCEEDINGS MEDICON 2004 Conference, Ischia, Naples-Italy July 31-August 5, 2004
GPCALMA: a Grid-based prototype for the use of CAD in mammographic screening

PROCEEDINGS MEDICON 2004 Conference, Ischia, Naples-Italy July 31-August 5, 2004
Computer-aied detection in screening programs: mammography and lung spiral CT

The collaboration was born in 1997 with the CALMA project (Computer Assisted Library for MAmmography), that was financed by INFN. The main results obtained achieved from CALMA collaboration (1997-2001) are:

• to collect a huge database of mammographic digitized images;
• to realize a CAD (Computer-Aided Detection) system for the automatic search of microcalcifications and masses in digitized mammograms.

Images were collected in different Italian hospitals (Bari, Napoli, Palermo, Sassari, Torino, Udine) and then they were digitized in a special format, to be analyzed. For the detection of suspicious lesions were developed two different approach for masses and for microcalcifications, based on both algorithms and neural networks.

Here are some results:

Medical Imaging 2002: Image Processing
Proceedings of SPIE Vol. 4684 (2002) 1301-1310
Search of Microcalcification Clusters with the CALMA CAD Station

Nuclear Instruments and Methods in Physics Research A 518 (2004) 391–393
The CALMA system: an artificial neural network method for detecting masses and microcalcifications in digitized mammograms

Medical Imaging 2003: Image Perception, Observer Performance, and Technology Assessment
Proceedings of SPIE Vol. 5034 (2003) 51-56
Diagnostic performance of radiologists with and without different CAD systems for mammography

Results obtained from CALMA were the starting point for the GPCALMA project (GRID Platform for CALMA), financed by INFN from 2002 to 2003. The basic idea for GPCALMA was to develop a GRID configuration for the CALMA utilities, that are the database and the CAD. To this aim the results obtained from GPCALMA are:

• a distributed database of mammographic images;
• an experimental GRID connections for the CAD algorithms.

Here are some results:

Computer Assisted Radiology and Surgery (CARS), 17th International Congress and Exhibition, June 2003, 25-28, London, GB
GPCALMA, a mammographic CAD in a GRID connection

Proc. IEEE-NPSS 2003
A Computer Aided Detection system for mammographic images implemented on a GRID infrastructure

NIM A 518 (2004)394-398
GPCALMA: a GRID Approach to mammographic screening