caratterizzazione multiparametrica di tessuti biologici agli ultrasuoni con metodi spettrali e probabilistici

Microcalcifications are of great importance in breast cancer early diagnosis and sometimes they do represent the unique evidence of cancer desease. Breast carcinomas are divided into five main groups: nodular opaquenesses (26 percent), parenchimal disruptions (7.5 percent), disruptions with microcalcifications (2.5 percent), opaquensesses with microcalcifications (17 percent) and microcalcifications as unique evidence (47 percent). Thus, about the 60 percent of mammographic examinations microcalcifications are detected when a breast neoplastic process is in act. Microcalcifications can derive from active cell secretions, necrotic cell debris, inflammations, traumas and radiations. Their exact compositions is actually unknown but the main constituents are: apatite, calcite, calcium oxalate, oxalic acid. Morphological feature analysis is also of great importance to assess whether they stand for benign changes or unpalpable malignant process. The aim of microcalcification analysis is to find all of the possible parameters which might suggest the malignancy of the actual change in tissue properties: shape and size variability, non uniform spatial distribution, density, edge irregularities, number greater than 15-20, number/cm2 greater than 10, average distance less than 1mm, cluster shape and number. Actually, the best technique to reveal microcalcifications is mammographic radiology which, with a 10 micrometer resolution, can achieve the correct visualization of microcalcifications, even the benign ones and the tiny ones. Its main disadvantage, on the other hand, is that xray wavelength is comparable with the DNA’s double helix gap, thus it is a potentially neoplastic agent. Any other mean other than xray would thus be of great clinical interest, but it should give informations comparable with those by xray themselves. Standard ultrasound techniques do not reliably detect microcalcifications. Their visualization, in fact, is limited by a number of factors: speckle (the ultrasound background noise), phase aberration, system limited spatial resolution, attenuation, human perception of the image and more. In general, microcalcifications can’t be revealed when they are located in fibroglandular echogenic regions. compounding is considered, i.e., the superposition a set of Bmode images of the same object. Each single image represents the convolution between the region of interest reflectivity function and the transducer point spread function (PSF) along a scan direction performed at a certain angle ? with respect to a reference position ?=0. It is to be noted that the result of this procedure is obtaining a set of statistically independent echoes from multiple observations of the object at different spatial positions that the transducer assumes while scanning: the improvement in the image contrast (i.e. the reduction of the speckle weight with respect to the target) is known to be proportional to N1/2, N being the number of independent echoes which compound is taken from. On the other hand, the great disadvantage is the loss in spatial resolution (both lateral and axial) which is proportional to N. So far, the last result of spatial compounding is an increase in SNR and a decrease in the system resolution. The essential feature of a system for angular scan, acquisition and data compounding is that the centre of rotation (CR) of the mechanical system might be determined with an accuracy better than the desired resolution for the final image. In reflection tomography, in particular, the knowledge of the CR is essential for the correct rotation and superposition of the single images performed at different transducer’s scan angles. The transducer’s resolution can be considered to be one half of the length of the ultrasonic pulse, thus the CR must be known with a better accuracy: considering a 10 MHz linear array transducer, for example, whose axial resolution is 154 ?m, an accuracy of about 80 ?m can be considered reasonable. Considering a Bmode image of an ideal point reflector in (x,y) obtained with the transducer in position ‘A’. x1 is the distance from the point target to the surface of the transducer and y1 is its lateral position; when the transducer is rotated of 180° in position ‘B’, x2 and y2 are the coordinates of the point reflector in the new position. the coordinates of CR can be estimated from the average of (x1,y1) and (x2,y2). The accuracy of this estimate depends on (i) how much the target can be considered an ideal point target, (ii) how much accurately is the 180° rotation performed and (iii) how much accurately can the distances x1, x2, y1 and y2 be measured. As for point (i), the ideal target can be approximated by an extremely thin wire, i.e. the diameter is less than the spatial resolution of the transducer; as for point (ii) the scan is performed by a precision rotation stage (PI M-037.DG) with 30 ?rad unidirectional repeatability, which cannot be reasonably the main source of error in the estimate of CR; as for point (iii) the accuracy in the measurement of the coordinates is the axial resolution for x1 and x2 and the lateral resolution for y1 and y2. So far we expect that the accuracy in determining CR is dominated by the lateral resolution which is worse than the axial one. The actual setup cannot perform automatic tomographic reconstruction because the mechanical structure is evidently not rotating around a fixed point so no CR determination is possible. In this first part of the work it has been possible to disregard the CR determination: being the microgranule of hydroxyapatite the uniqueinclusion within the phantom, it is always recognizable because its echo is never covered by any hyperecogenic structure. At any scan angle one frame was acquired by the echo scanner. The demodulated signals were computed with Hilbert transform . After having selected a region of interest with the microgranule inside, the gated matrixes were resampled in lateral direction to operate over square pixel (30ns x 30ns) corresponding to 22 micrometers x 22 micrometers in agar. The study of the last setup involves a new structure which has been thought to be stiffer with respect to the previous setup and the very first simulations show a transducer maximum deflection in static condition of about 10 micrometers by gravitational field.Il carcinoma mammario si presenta sotto molteplici evidenze mammografiche. Nel 47 percento dei casi l’unico segno della presenza di un tumore alla mammella è la presenza di microcalcificazioni (MC). In generale in ecografia mammaria, le microcalcificazioni sono di difficile individuazione perché si tratta di inclusioni iperecogene ma dell’ordine di qualche centinaio di micrometri e dunque facilmente nascoste da un mezzo circostante molto ecogeno per la presenza di tessuto adiposo e fibroso. Per simulare la presenza di microcalcificazioni sono stati utilizzati microgranuli di idrossiapatite (HA) in fantocci di varie tipologie. Fantocci rettangolari per scansioni lineari, fantocci cilindrici per scansioni angolari, e materiali per il background, variabile da una miscela di Agar e acqua (alto contrasto) ad Agar e Latte (basso contrasto). La Tomografia tradizionale con RX, PET, ecc si basa sulla trasmissione di un pacchetto di energia (origine raggi x, positroni, ecc.) attraverso un materiale e ne registra un assorbimento da parte del tessuto. Un approccio simile si realizza quando si utilizza un sistema US con trasmettitore e ricevitore fisicamente separati, in cui viene registrata, attraverso i tempi di volo, la mappatura dell’attenuazione e della velocità sonica nel mezzo. In questo studio viene invece utilizzato un ecografo commerciale, cioè un sistema in cui trasmettitore e ricevitore coincidono:, si ottiene quindi, per ogni angolo di vista, una mappa bidimensionale degli echi di ritorno. Questo significa che tutte le tecniche standard di ricostruzione utilizzate per la tomografia tradizionale non sono applicabili, almeno direttamente, per la ricostruzione ecografica a partire da immagini bidimensionali. Scopo di questo lavoro risulta quindi anche lo studio di un protocollo di elaborazione dati che consenta di matchare un set di N immagini bidimensionali prese ad angoli ?N in una singola immagine che rappresenti l’oggetto a 360°, elaborando i dati acquisiti in termini spettrali (Faran) e probabilistici (Riciano) e disponendo di N realizzazioni tutte indipendenti tra loro