Deep Learning Algorithms in the Automatic Segmentation of Liver Lesions in Ultrasound Investigations

Abstract

Background: The ultrasound is one of the most used medical imaging investigations worldwide. It is non-invasive and effective in assessing liver tumors or other types of parenchymal changes. Methods: The aim of the study was to build a deep learning model for image segmentation in ultrasound video investigations. The dataset used in the study was provided by the University of Medicine and Pharmacy Craiova, Romania and contained 50 video examinations from 49 patients. The mean age of the patients in the cohort was 69.57. Regarding presence of a subjacent liver disease, 36.73% had liver cirrhosis and 16.32% had chronic viral hepatitis (5 patients: chronic hepatitis C and 3 patients: chronic hepatitis B). Frames were extracted and cropped from each examination and an expert gastroenterologist labelled the lesions in each frame. After labelling, the labels were exported as binary images. A deep learning segmentation model (U-Net) was trained with focal Tversky loss as a loss function. Two models were obtained with two different sets of parameters for the loss function. The performance metrics observed were intersection over union and recall and precision. Results: Analyzing the intersection over union metric, the first segmentation model obtained performed better compared to the second model: 0.8392 (model 1) vs. 0.7990 (model 2). The inference time for both models was between 32.15 milliseconds and 77.59 milliseconds. Conclusions: Two segmentation models were obtained in the study. The models performed similarly during training and validation. However, one model was trained to focus on hard-to-predict labels. The proposed segmentation models can represent a first step in automatically extracting time-intensity curves from CEUS examinations.

An Automated Method for Classifying Liver Lesions in Contrast-Enhanced Ultrasound Imaging Based on Deep Learning Algorithms

“…The segmentation model was trained on B-mode frames extracted from contrast-enhanced ultrasound video investigations. A more detailed description of the model can be found in our previous published study [ 10 ]. A feed-forward neural network model was then trained with data from two sources: values extracted from the time-intensity curves and clinical information of the patient.…”

“…The most dominant diagnosis was hepatocellular carcinoma (40.67%), while the least dominant diagnosis was liver adenoma (1.69%). A more detailed description of the patient cohort was presented in our previous studies [ 10 , 20 ]. Being modular, the proposed system can be enhanced to detect other liver lesions by performing transfer learning only on the feed-forward neural network model.…”

“…The labels for the dataset (masks) were automatically exported from QuPath. A more detailed presentation of this method was conducted in our previous work [ 10 ]. The demographic information of the patients is presented in Table 1 .…”

“…The U-Net [ 11 ] image segmentation model obtained from our previous work [ 10 ] was used to obtain region of interest for each frame in the video investigation file. By passing a frame from the ultrasound video investigation through the presented image segmentation model, a mask of 256 by 256 pixels is obtained.…”

“…To create the dataset and for extracting the lesion from each frame in the video investigation, boundaries were defined for each region in the video investigation. The boundaries for B-mode image were defined in our previous work [ 10 ] and were used to create the dataset but also for predicting the mask. For the contrast image section in the video investigation file, the boundaries were determined experimentally as for the B-mode image.…”