Quantitative Study Of Ultrasound Hepatic Perfusion Based On Factor Analysis

Liver cancer is a major disease that affects the physical health of Chinese people, andits early detection and treatment is the key to improve the cure rate and to reduce themortality. Contrast-enhanced ultrasound (CEUS) imaging is a novel imaging techniquethat dynamically and clearly depicts the micro-vasculature, especially the vasculature ofthe tumor by injecting microbubble contrast agents. The rich blood supply and perfusioninformation of the tissue can be acquired by using this technique, which is importantsignificant to the clinical diagnosis of the tissue diseases. Compared withcontrast-enhanced CT and MRT, it is real-time, safety and low-cost. Now the technique iswidely applied to the differential diagnosis of focal liver lesions (FLLs). Generally, thediagnosis by using CEUS imaging mainly relies on the ultrasonographists’observationsand judgments of the enhancement level and change in different vascular phases. This wayis vulnerable to the impact of the ultrasonographists’subjectivity, and there is differencebetween the observers. To overcome it, the diagnostic rate of FLLs could be improved byusing the quantitative analysis methods for the CEUS data from different types of FLLs.This article aims to provide a practical quantitative analysis method in CEUS imagingwith high accuracy for ultrasonographists. To solve the problems in the factor analysis ofdynamic structures (FADS）methods, a novel and effective FADS method is investigatedto automatically extract time-intensity curve（sTICs）from the CEUS data of different typesof FLLs. Then, the parametric imaging is performed by utilizing the characteristicparameters of the extracted TICs. These maps reflect the perfusion characteristics ofdifferent types of FLLs, which is helpful for ultrasonographists to differentiate FLLs.The investigated clinical case data was collected during patients breathed freely.Therefore, an image-based motion correction strategy is proposed to quickly correct therespiratory motion in the image sequence. Usually, the commercial ultrasound machine can display contrast and tissue images simultaneously. Because of the low gray-levelvariation in the tissue images, the tissue images are registered by using template matchingwith sum of absolute differences metric. Then, the similar images are selected by adouble-selection method which requires global and local threshold setting. Finally, thecontrast images are determined by utilizing the relationship of position mapping betweenthe tissue and contrast windows. Simple manual operation is only needed, such as theselection of the template image and search space. It is independent of the tumor size anduser-friendly. Moreover, this strategy is suitable for most clinical cases affected by adversefactors of sampling. Due to the merit of the saving time, this strategy can be widelyapplied to clinical practice, and the diagnostic efficiency of FLLs will be improved.To solve the problems in the current FADS methods, a novel replace-approximation(RA) is proposed. The main idea of the algorithm is that the true factor curve can beapproximately replaced by the optimal TIC of the pixel that is found in the un-overlappedphysiological regions. In this way, the TIC extracted by the RA method will not lose mostof the original information. This method always starts to seek the apexes fromone-dimensional space first, instead of doing it usually from q－1dimensional space. Itdoes not need the estimate of any specific parameter, a priori knowledge or factor rotation.The algorithm is tested on the phantoms, and compared with the two other FADS methods.RA algorithm is further validated on the free-breathing clinical cases. Because thenumber of factors in the FADS is ambiguous, two-and three-factor analyses are performedon the clinical cases respectively. Experimental results show that the RA method couldextract physiological factor curves and factor images efficiently. However, all the clinicalcases are not suitable for the three-factor analysis. In contrast, the two-factor analysisperforms more stably.The wash-in time ratio indexes are calculated from the characteristic parameters ofthe two factor curves extracted by the two-factor analysis performed on the clinical cases.Then, these values generate the corresponding parametric images. These parametric maps not only represent the perfusion characteristics of different types of FLLs, but also providethe relative speed information of blood flow in the perfusion region. It indicates thattwo-factor analysis has potential to the quantitative analysis of hepatic perfusion, whichwould be helpful to the differential diagnosis of FLLs.