SMART TECHNIQUES FOR FAST MEDICAL IMAGE ANALYSIS AND PROCESSING

Medical Imaging has become an important transversal applications and re- search field that embraces a great variety of sciences. Imaging is the central science of measurement in diagnosis and treating diseases. The effort of the technological progress has made possible human imaging starting from a single molecule to the whole body. The open challenge is to treat the huge amount of medical informations with the use of smart and fast techniques that allows clinical and images data analysis and processing. In this ph.D. Thesis, many issues have been addressed and a certain amount of improvement in various fields have been produced, such as biom- etry, organs and tissues segmentation, MRI thermometry, medical reports retrieval and classification. The topic prefixed at the beginning of this ph.D. route was to analyze, understand, and give a step over to various kind of problematics related to Medical Images and Data analysis, working closely to radiologist physicians, with specific equipments, and following the common denominator of fast and smart methodologies applied to the medical imaging issue. A series of contribution have been carried out in fields such as: • proposing two different kind of multimodal biometric authentication systems that investigates fingerprint and iris fusion and processing; • applying expert systems to the issue of data validation, comparing and validating data to two different methodologies that assess liver iron overload in thalassemic patients;• addressing and improving non-invasive referenceless thermometry by using Radial Basis Function as interpolator; • applying the multi-seed region growing method to the segmentation of CT liver dataset; • proposing a novel unsupervised voxel-based morphology method for MRI brain segmentation by using k-means clustering and neural net- work classification; • proposing a novel ontology-based algorithm for information retrieval from mammographic text reports. The above work has been developed with the cooperation of the medical staff of the “Dipartimento di Biopatologia e Biotecnologie Mediche e Forensi” and the “Scuola di Specializzazione in Radiodiagnostica" of the Università degli Studi di Palermo. All the proposed contributions show good performance using the stan- dard metrics. Most of them have produced scientific publications in com- puter science venues as well as in radiological venues. In addition, some specific frameworks, such as OsiriX, have been used to improve usability and easiness of the developed systems.