Computer-aided diagnosis for the classification of focal liver lesions by use of contrast-enhanced ultrasonography

Abstract

The authors developed a computer-aided diagnostic (CAD) scheme for classifying focal liver lesions (FLLs) as liver metastasis, hemangioma, and three histologic differentiation types of hepatocellular carcinoma (HCC), by use of microflow imaging (MFI) of contrast-enhanced ultrasonography. One hundred and three FLLs obtained from 97 cases used in this study consisted of 26 metastases (15 hyper- and 11 hypovascularity types), 16 hemangiomas (five hyper- and 11 hypovascularity types) and 61 HCCs: 24 well differentiated (w-HCC), 28 moderately differentiated (m-HCC), and nine poorly differentiated (p-HCC). Pathologies of all cases were determined based on biopsy or surgical specimens. Locations and contours of FLLs on contrast-enhanced images were determined manually by an experienced physician. MFI was obtained with contrast-enhanced low-mechanical-index (MI) pulse subtraction imaging at a fixed plane which included a distinctive cross section of the FLL. In MFI, the inflow high signals in the plane, which were due to the vascular patterns and the contrast agent, were accumulated following flash scanning with a high-MI ultrasound exposure. In the initial step of our computerized scheme, a series of the MFI images was extracted from the original cine clip (AVI format). We applied a smoothing filter and time-sequential running average techniques in order to reduce signal noise on the single MFI image and cyclic noise on the sequential MFI images, respectively. A kidney, vessels, and a liver parenchyma region were segmented automatically by use of the last image of a series of MFI images. The authors estimated time-intensity curves for an FLL by use of a series of the temporally averaged MFI images in order to determine temporal features such as estimated replenishment times at early and delayed phases, flow rates, and peak times. In addition, they extracted morphologic and gray-level image features which were determined based on the physicians' knowledge of the diagnosis of the FLL, such as the size of lesion, vascular patterns, and the presence of hypoechoic regions. They employed a cascade of six independent artificial neural networks (ANNs) by use of extracted temporal and image features for classifying five types of liver diseases. A total of 16 temporal and image features, which were selected from 43 initially extracted features, were used for six different ANNs for making decisions at each decision in the cascade. The ANNs were trained and tested with a leave-one-lesion-out test method. The classification accuracies for the 103 FLLs were 88.5% for metastasis, 93.8% for hemangioma, and 86.9% for all HCCs. In addition, the classification accuracies for histologic differentiation types of HCCs were 79.2% for w-HCC, 50.0% for m-HCC, and 77.8% for p-HCC. The CAD scheme for classifying FLLs by use of the MFI on contrast-enhanced ultrasonography has the potential to improve the diagnostic accuracy in the histologic diagnosis of HCCs and the other liver diseases.

Figure 1

Illustration of the procedure of MFI and its sample images. Schemas on the left show sections crossing the scanning plane for the MFI. Injected contrast agent (microbubbles) was enhanced by harmonic imaging which depresses the normal structures. Diffused microbubbles in the scanning plane were destroyed by high MI in the first step of MFI, and then the microbubbles in the adjacent region were replenished into the scanning plane corresponding to the vascularity patterns of an FLL. The MFI images were recorded sequentially until the replenishment was saturated.

Figure 2

Examples of two FLLs (m-HCCs) which represent centrifugal (upper) and centripetal (lower) progressions of replenishment from early (left) to delayed (right) phases.

Figure 3

Examples of three FLLs which represent hyperechoic (upper left: m-HCC), hypoechoic (upper right: metastasis), and isoechoic (lower left: w-HCC) replenishment patterns.

Figure 4

Overall computerized scheme used in this CAD scheme for the classification of FLLs on MFI images.

Figure 5

Difference in average pixel values in sequential MFI images and replenishment times within a FLL (shown in the figure) obtained with the original and temporally averaged MFI images.

Figure 6

Explanation of the VE filtering technique for a calculation point C(i,j). P_(θ,R) is a pixel value at the location of (i+R sin θ,j+R cos θ). The VE filter output VE(i,j) was defined as the maximum value among 12 P_(θ,R) values obtained with four angles of θ and three filter sizes of R.

Figure 7

Example of (a) original MFI image including one FLL (arrow) and a kidney, (b) vessel-like pattern enhanced image obtained with the VE filter technique, (c) segmented ALP regions obtained from the original MFI image, and (d) skeleton of vessel-like pattern enhanced image for estimating the average size of vessel-like patterns on the MFI image.

Figure 8

Sample images of a FLL and illustration of a computerized scheme for dividing the FLL into central and peripheral regions. A fixed margin (M) was calculated from the effective diameter of a FLL and used for dividing the FLL. (a), (c) Original MFI image with a lesion contour which was drawn by the physician, and a contour of the central region segmented by the computer. (b), (d) Peripheral region of a FLL were divided from the central region by use of a rolling circle which has a radius M.

Figure 9

Example of the original MFI image at the delayed phase and its segmented images for hypoechoic regions at the early and delayed phases. The difference in the regions between two images at two phases was defined as delayed-enhancement region.

Figure 10

Illustration of the cascade of six ANNs used in this study. Six decisions in which alternative choices for specific groups of FLLs were determined by single ALL, leading a final diagnostic decision for five liver diseases.

Figure 11

Examples of correctly classified cases. (a) w-HCC showed relatively uniform and isoechoic replenishment patterns, and the size was relatively small (15.9 mm, pixel size: 0.29 mm). (b) m-HCC showed a relatively isoechoic replenishment pattern, but the size was large (23.4 mm, pixel size: 0.17 mm), and the replenishment pattern within a lesion was not uniform. (c) p-HCC had a relatively large size (20.1 mm, pixel size: 0.34 mm) and showed branched tumor vessels within a lesion. (d) Metastasis (hypovascularity type) had a large size (65.6 mm, pixel size: 0.29 mm) and showed a hypoechoic region at the center of the lesion at the delayed phase. And (e) hemangioma (hypovascularity type) had a relatively large size (56.6 mm, pixel size: 0.26 mm) and was replenished very slowly with centripetal progression.

Figure 12

Example of the CAD output for a FLL (hypervascularity metastasis) which was incorrectly classified as m-HCC, because the size of this lesion was relatively large (25.1 mm, pixel size: 0.24 mm), and the echogenicity of this metastasis was isoechoic to those in ALP regions.