A sonographic software program, Fluctuational Imaging, for diagnosis of hepatic hemangioma

Hepatic hemangioma is the most common benign solid lesion of the liver. Contrast-enhanced computed tomography or magnetic resonance imaging is recommended for definitive diagnosis of hepatic hemangioma. However, these modalities have drawbacks in terms of radiation exposure, invasiveness, and high cost for examination. “Fluttering sign” is one of the candidate findings considered specific for hepatic hemangioma that can be useful for diagnosis of hepatic hemangioma using grayscale US alone. However, the assessment is subjective and the findings are weak and likely to be overlooked in some cases. We developed a software program, Fluctuational Imaging, for objective detection and depiction of “fluttering sign”. Here, we evaluated the ability of Fluctuational Imaging software to depict “fluttering sign” in hepatic hemangioma. Presence or absence of “fluttering sign” was evaluated in the grayscale US videos and Fluctuational Imaging software analysis results of patients with hepatic hemangioma. The Cohen’s kappa test showed very good agreement (0.95). Fluctuational Imaging software can detect and depict the phenomenon of “fluttering sign” well and may be a useful tool for diagnosis of hepatic hemangioma.


Methods
This study involving retrospective interpretation of prospectively acquired data was approved by the Kawasaki Medical School ethics committee. The study was performed in accordance with the Declaration of Helsinki. Informed consent was obtained from all participants.
US examinations. The US equipment used was a TUS-AI900 system with 4-MHz convex or 7-MHz linear probes (Canon Medical Systems Corporation). Patients were placed in the supine position and asked to hold their breath, rather than performing the Valsalva maneuver, while the grayscale US video was acquired. The focus of the US beam was set at the lowest (deepest) margin of the lesion, and software that may influence the results of Fluctuational Imaging software analysis, such as enhancement of structure boundaries or multidirectional beaming, was turned off during data acquisition. The grayscale US video of each lesion was recorded for about 5 s as clip data on an inbuilt hard disk. All US examinations were performed by a single doctor who was an expert in abdominal US (35 years of experience).
Patients. Grayscale  Fluctuational imaging software. Fluctuational Imaging is a software program for detection and depiction of "fluttering sign" in grayscale images. The software was developed with consideration of the lesion type using the program language of Visual C + + 2010. Because "fluttering sign" is a phenomenon in which the signal intensity changes on the time axis, a cross-correlation coefficient was evaluated. The frame rate was 20 frames/s and 70 frames (3.5 s) were used for Fluctuational Imaging software analysis. The time required for Fluctuational Imaging software analysis was approximately 10-20 s. The procedure is shown below.
1. Manual determination of the region of interest according to the size of the lesion 2. Elimination of motion effects of the heart or arterial pulsation using "motion tracking" technology that can track and cancel overall movement in the region of interest to reduce the effects of body movements arising from pulsation and breathing 3. Calculation of "R", a cross-correlation coefficient, for each point in the region of interest using data within a range of few millimeters (approximately 2 mm) I : luminosity value; n: frame number; i, j: coordinates of the point. 4. Calculation of degree of coincidence of the cross-correlation coefficient around the analysis point.
Degree of fluctuation: change in local signal strength along the time axis. Large degree of fluctuation: degree of coincidence was small.

Assignment of "R" 0-1 to 255 gradation colors
Large degree of fluctuation: yellow to red. Small degree of fluctuation: blue.
Because the analysis was qualitative rather than quantitative, there was no specific cutoff value for "R" and no degree of coincidence value above or below which a specific classification of the lesion was made. Procedure numbers 2 to 4 interacted for analysis. Based on the results of preliminary research, 255 gradation colors were assigned to emphasize "fluttering sign".
The procedure for Fluctuation Imaging software analysis. Agreement between Fluctuational Imaging software and grayscale US video results. The agreement between Fluctuational Imaging software analysis results and grayscale US video results regarding the presence or absence of "fluttering sign" was evaluated by Cohen's kappa coefficient. Statistical analysis was performed with EZR (Jichi Medical University, Japan), which is a graphical user interface for R 16 . Four cases with different results in the grayscale US videos and Fluctuational Imaging software analysis were negative on grayscale US videos and positive on Fluctuational Imaging software analysis. The different results were considered to arise from the effects of pulsation. Figure 4 shows a colored area thought to arise from an effect of the heartbeat rather than "fluttering sign" (Supplementary video V4). www.nature.com/scientificreports/

Discussion
This initial and preliminary report demonstrates that the sonographic software program Fluctuational Imaging was able to detect and depict "fluttering sign" in hepatic hemangioma. "Fluttering sign" is considered a specific finding for hepatic hemangiomas and its appearance rate is about 50-70% 14,15 . The appearance rate of "fluttering sign" in the present study was similar. Even though "fluttering sign" is only observed in about 50-70% of hepatic hemangiomas, it can be a useful finding for diagnosis of hepatic  www.nature.com/scientificreports/ hemangioma using grayscale US videos alone when it is present. Similarly, Fluctuation Imaging software is only useful for cases in which "fluttering sign" is observed. Because there is no standard for judging the presence or absence of "fluttering sign", we determined its presence or absence by repeatedly reviewing playback of the grayscale US videos in the present study. We then evaluated the performance of the Fluctuational Imaging software analysis based on the degree of matching with the grayscale US video results. The fact that the grayscale US video and Fluctuational Imaging software analysis results were almost the same does not mean that Fluctuational Imaging software analysis is unnecessary, because repeated playback of grayscale US videos is not possible during actual US examinations. There were cases that demonstrated the usefulness of Fluctuational Imaging software analysis results because the "fluttering sign" was weak and likely to be overlooked. Meanwhile, a few cases were considered to be false-positives for Fluctuational Imaging software analysis, because the internal echogenicity of the hepatic hemangioma did not change continuously or seem to be moving. It may be better to use Fluctuational Imaging software for areas without continuous motion, because the effect of pulsation was considered the cause of the false-positive cases. Further studies are warranted to determine how to optimize Fluctuational Imaging software analysis.
As hepatic hemangiomas are common benign lesions and biopsy involves a risk of bleeding, the diagnosis of hepatic hemangioma was made based on a previous report 17 and guidelines 7,18 . There are several subtypes of hepatic hemangioma, including cavernous hemangioma and capillary hemangioma. Because the differences in "fluttering sign" and Fluctuational Imaging software analysis for subtypes of hepatic hemangioma were not evaluated, further studies are required to validate the software analysis for the subtypes of hepatic hemangioma.
Because a saved grayscale US video is analyzed, Fluctuational Imaging software is not invasive and has no associated cost, but real-time evaluation is not possible. The analysis using Fluctuational Imaging software is easy and the time required is about 1-2 min. The analysis requires surrounding tissues of approximately the same size as the hepatic hemangioma for comparison. Therefore, the analysis is theoretically considered possible for a hepatic hemangioma in which an ROI of twice its area can be set. However, the size criteria for a hepatic hemangioma for use of Fluctuational Imaging software analysis have not been determined, and further studies are needed. Two different transducers were used in the present study, but the effect of the differing frequencies  www.nature.com/scientificreports/ was not examined. Although the spatial resolutions of the images collected with the 4-MHz convex or 7-MHz linear transducers were different, the same algorithm was used, and thus it is possible that the colored parts differed slightly between the two transducers. Evaluation of the effect of the different transducers is necessary. Because this is an initial and preliminary report involving a small number of cases, the study has several limitations. Specifically, because the statistical agreement was based on the observations of only one reviewer, the inter-rater or intra-rater agreement was not assessed, some biases may be present, and the reproducibility was not assessed. Studies with larger numbers of hepatic lesions are needed to evaluate the reproducibility, the effect of lesion position in the liver, the differential diagnosis ability for focal liver lesions, and the clinical usefulness of Fluctuational Imaging software.
In conclusion, we have developed a software program named Fluctuational Imaging. Fluctuational Imaging software can detect and depict the phenomenon of "fluttering sign" well and may be a useful tool for diagnosis of hepatic hemangioma.

Data availability
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.