An image analyzing system using an artificial neural network for evaluating the parenchymal echo pattern of cirrhotic liver and chronic hepatitis

Heterogeneous liver uptake of Tc-99m-GSA as quantified through SPECT/CT helps to evaluate the degree of liver fibrosis: A retrospective observational study

Tc-99m-galactosyl human serum albumin (GSA) scintigraphy is used to assess the hepatic functional reserve, and allows for visual assessment of the residual hepatocyte distribution on single-photon emission computed tomography/computed tomography (SPECT/CT) images. The association between heterogeneous liver uptake of Tc-99m-GSA and liver fibrosis remains to be studied in detail. We analyzed this association.Fifty-one patients with chronic hepatobiliary disease undergoing a Tc-99m-GSA scintigraphy were included in this study. The receptor (LHL15) and blood clearance (HH15) indexes (the uptake ratios of the liver and heart) were obtained from dynamic planar images. The liver uptake count maximum-to-mean ratio (LUC Max/Mean) was calculated from single-photon emission computed tomography/computed tomography (SPECT/CT) images as an indicator of the Tc-99m-GSA liver uptake heterogeneity. We assessed the relationship between these quantified values and liver fibrosis.There were 30 Child-Pugh classification grade A patients, 16 grade B patients, and 5 grade C patients. Among the 30 patients whose liver histopathology was evaluable, those with advanced liver fibrosis (F2-4) had a lower LHL15 than those with mild liver fibrosis (F0-1) (median, 0.90 vs. 0.92, P = .04), and a higher LUC Max/Mean (median, 1.80 vs. 1.70, P = .02). The multivariate analysis identified platelets (P = .04) and the LUC Max/Mean (P = .04) as contributing factors of advanced liver fibrosis.These findings suggest that Tc-99m-GSA SPECT/CT can be used not only to assess the hepatic functional reserve, but also to evaluate a degree of liver fibrosis.

A computer-aided diagnostic system to discriminate SPIO-enhanced magnetic resonance hepatocellular carcinoma by a neural network classifier

In this paper, a computer-aided diagnostic (CAD) system for the classification of rat liver lesions from MR imaging is presented. The proposed system consists of two modules: the feature extraction and the classification modules. 40 rats are used for hepatocellular carcinoma (HCC) induction with Diethylnitrosamine via drinking water. After Resovist is administrated by tail vein the animals are scanned by a 1.5-T MR scanner with T2-weighted FRFSE sequence. SPIO-enhanced images of 106 nodules (RNs(:) 24, HCCs: 82) are acquired, and 161 regions of interest (ROIs) are taken from the MR images .Six parameters of texture characteristics including Angular Second Moment, Contrast, Correlation, Inverse Difference Moment, Entropy, and Variance of 161 ROIs are calculated and assessed by gray-level co-occurrence matrices, then fed into a BP neural network (NN) classifier to classify the liver tissue into two classes: cirrhosis and HCC. Difference of each texture parameter between cirrhosis and HCC group is significant. The accuracy of classification of HCC nodules from cirrhosis is 91.67%. It indicates the ANN classifier based on texture is effective for classifying HCC nodules from cirrhosis on rat SPIO-enhanced imaging.

Quantitative assessment of liver fibrosis: a novel automated image analysis method

BACKGROUND

Semiquantitative staging of liver fibrosis is a highly subjective procedure and may lead to an uncertainty in judgment regarding the degree of severity and hence the progression of the disease.

AIM

In this work, we present an automated quantification system (AQS) for evaluating the degree of severity of fibrosis in liver biopsies based on Ishak et al.'s classification. Accordingly, liver fibrosis is classified into six classes depending on its severity and progression. The described system is of special value in accurately assessing the prognosis of chronic liver disease.

METHODS

In our method, we tried to approximate the architecture of the fibrosis in the subject sample using texture features and shape representation of the fibrosis structural expansion with an overall accuracy of about 98%.

RESULTS AND CONCLUSION

The presented AQS is considered to be a novel approach in the domain of automatic liver fibrosis quantification. It is a true quantification and intelligent approach that attempts to utilize the current semiquantitative methods of liver fibrosis assessment to turn them into real quantitative ones with significant reduction in variability and subjectivity. We propose that our method can be adopted by a panel of expert liver pathologists and software to be developed and used on a wide scale.

A pilot approach for quantitative assessment of liver fibrosis using ultrasound: preliminary results in 79 cases

BACKGROUND/AIMS

Ultrasound is noninvasive and useful to evaluate liver disease despite its operator dependency. This pilot study was conducted to quantitatively assess liver fibrosis using ultrasound.

METHODS

Fibrosis extraction ratios (FER) (fiber volume/total volume) of ultrasound and histological images of 8 autopsy specimens were compared. We also compared FER of ultrasound images from clinical patients (n=79) with histological fibrosis stages.

RESULTS

In the autopsy study, FER correlation coefficient between histological images and ultrasound images was 0.992. Regarding clinical patients, there was sufficient evidence to indicate differences in the distributions of FER for each fibrosis stage (Kruskal-Wallis test P<0.0001). With FER cut-off to distinguish > or =F2 from F0 and F1 defined as mean plus standard deviation of F0 and F1, sensitivity, specificity, positive predictive value, negative predictive value, and likelihood ratio were 62, 75, 78, 57%, and 2.47, respectively. Regarding HCV cohort (n=44), they were 55, 87, 89, 50%, and 4.14, respectively. Areas under receiver operating characteristic curves were 0.78, 0.79, 0.83 and 0.83 for > or =F1, > or =F2, > or =F3 and =F4, respectively. Regarding HCV cohort, they were 0.74, 0.71, 0.79 for > or =F2, > or =3 and =4, respectively.

CONCLUSIONS

The FER method has great potential for diagnosing liver fibrosis using ultrasound.

CLASSIFICATION OF CIRRHOSIS FROM B-SCAN IMAGES USING PYRAMID NEURAL NETWORK

This paper proposes a system applying a pyramid neural network for classifying the hepatic parenchymal diseases in ultrasonic B-scan texture. The conventional multilayer neural network emphasizing on the data carried by the last hidden layer has the drawback of not fully utilizing the information carried by the input data. A pyramid network can solve the problem successfully.

To solve the common problem of neural network, which is time-consuming in computation, FDWT (Fast Discrete Wavelet Transform) is a key technique used during preprocessing to cut down the size of patterns feed to the network. The B-scan patterns are wavelet transformed, and then the compressed data is fed into a pyramid neural network to diagnose the type of cirrhotic diseases.

The performance of the proposed system and that of a system based on the conventional multilayer network architecture is compared. The result shows that compared to the conventional 3-layer neural network, the performance of the proposed pyramid neural network is improved by effectively utilizing the lower layer of the neural network.

Impairment of change in diameter of the hepatic portion of the inferior vena cava: a sonographic sign of liver fibrosis or cirrhosis

OBJECTIVE

We propose a new sonographic technique for detecting parenchymal stiffness of the liver. This technique measures the physiologic change in the diameter of the hepatic inferior vena cava (IVC). The hepatic portion of the IVC is extensively attached to the hepatic parenchyma so that changes in diameter of the venous lumen could not occur without changes in the shape of the surrounding parenchyma. Therefore, increased parenchymal stiffness due to cirrhosis or fibrosis may result in an impaired change in venous diameter.

METHODS

Thirty patients with histologically proven cirrhosis and 30 patients with normal livers were examined. A commercial ultrasound system was used in conjunction with a convex probe at a center frequency of 4.0 MHz. The major axis of the IVC was measured during normal breathing. The patients were then requested to take a deep breath to produce negative intrathoracic pressure, and the same measurement as that during normal breathing was repeated immediately.

RESULTS

In the normal liver group, the maximal diameter of the vena cava was 2.35 +/- 0.34 cm (mean +/- SD), and this was reduced by 1.30 +/- 0.67 cm (range, 0.4-2.85 cm) during deep inspiration. In the cirrhotic patient group, the maximal diameter was 1.74 +/- 0.35 cm, and this was reduced by 0.03 +/- 0.09 cm (range, 0.0-0.4 cm) (P < .0001).

CONCLUSIONS

The physiologic change in the diameter of the hepatic portion of the IVC enhanced by deep respiration may reflect the stiffness of hepatic parenchyma.

Irregularity of parenchymal echo patterns of liver analyzed with a neural network and risk of hepatocellular carcinoma in liver cirrhosis

OBJECTIVE

In this study, we scored the hepatic parenchymal echo patterns as the coarse score (CS) analyzed with a neural network in cirrhosis patients and calculated the variations in CS as the coefficient of variation, and evaluate their usefulness as predictor of the development of hepatocellular carcinoma (HCC).

METHODS

The relationship between the degree of variation in CS and histopathological findings was assessed in 10 autopsied livers fixed in formalin. The degree of intrahepatic variation in CS was calculated as the coefficient of variation of CS (CVCS). Irregular regeneration of liver cells in autopsied livers was classified into two categories, slight and severe. A total of 56 cirrhosis patients were prospectively followed to evaluate the predictors of HCC.

RESULTS

A significant positive correlation was observed between CVCS and the coefficient of variation of the diameter of the regenerative nodules. Coefficient variation of nodule diameter and CVCS in patients in whom irregular regeneration was severe were significantly higher than those in patients in whom irregular regeneration was mild (p < 0.05). Concerning the relationship between hepatitis virus markers and CS or CVCS, CVCS was significantly higher in those who were hepatitis C virus antibody positive and those who were hepatitis B surface antigen negative (p < 0.01). Using a combination of CS and CVCS, the incidence of HCC, as determined by the Kaplan-Meier method, was significantly higher in patients whose CS was > or = 1.5 and CVCS was > or = 15%, as compared with that in patients whose CS was <1.5 and CVCS was <15% (p < 0.01) and whose CS was > or = 1.5 and CVCS was <15% (p < 0.05). Multivariate analysis of the predictors for HCC using the Cox's proportional hazards model showed a significant correlation between the risk of development of HCC and CVCS, CS and serum alpha-fetoprotein level.

CONCLUSIONS

CVCS and CS can be calculated from liver echo patterns and are useful for identifying a high-risk group for HCC.

Differentiation of intracardiac tumors and thrombi by echocardiographic tissue characterization: comparison of an artificial neural network and human observers

The feasibility of classifying ultrasound images of intracardiac tumors and thrombi with a neural network-based algorithm was compared with the performance of experienced echocardiographers. The neural network used statistical descriptors of the apparent echocardiographic texture of the masses, and the blinded echocardiographers were given photographic prints of enlarged regions of interest without clinical data. The network classified 66% of the images correctly and the echocardiographers, 83%. The network and echocardiographers agreed in 88% of the images. Human observers usually base their classification of intracardiac masses on clinical data. The echocardiographic texture of tumors is quantitatively different from that of thrombi. This difference can be recognized by a neural network and potentially be useful in assisting with the diagnosis when clinical clues are insufficient.

Parenchymal echo patterns of cirrhotic liver analysed with a neural network for risk of hepatocellular carcinoma

BACKGROUND

To objectively evaluate the parenchymal echo patterns of the liver in cirrhosis, an image analysing system in which a neural network is used has been found capable of numerically calculating coarse score (CS). Using this system, we analysed whether or not CS can serve as a predictive factor for the development of hepatocellular carcinoma (HCC).

METHODS

The risk factors for HCC were evaluated in 95 patients with liver cirrhosis with an average follow-up period of 2041 +/- 823 days. We used a three-layer feed-forward neural network and a back-propagation algorithm to calculate CS.

RESULTS

There were strong correlations between CS, alanine aminotransferase (ALT) and alpha-fetoprotein (AFP) and the average cumulative incidence rate of HCC evaluated by the Cox's proportional hazards model. The adjusted rate ratios were estimated to be 3.00, 2.80 and 2.01, respectively. The cumulative risks of HCC were significantly higher with an initial CS > or = 1.5 than with an initial CS < 1.5, with ALT > or = 80 IU/L than with initial ALT < 80 IU/L and with AFP > or = 20 ng/mL than with initial AFP < 20 ng/mL, all analysed by the log-rank test.

CONCLUSIONS

Coarse score is a useful predictor for development of HCC.