Recognizing Focal Liver Lesions in CEUS With Dynamically Trained Latent Structured Models

Abstract

This work investigates how to automatically classify Focal Liver Lesions (FLLs) into three specific benign or malignant types in Contrast-Enhanced Ultrasound (CEUS) videos, and aims at providing a computational framework to assist clinicians in FLL diagnosis. The main challenge for this task is that FLLs in CEUS videos often show diverse enhancement patterns at different temporal phases. To handle these diverse patterns, we propose a novel structured model, which detects a number of discriminative Regions of Interest (ROIs) for the FLL and recognize the FLL based on these ROIs. Our model incorporates an ensemble of local classifiers in the attempt to identify different enhancement patterns of ROIs, and in particular, we make the model reconfigurable by introducing switch variables to adaptively select appropriate classifiers during inference. We formulate the model learning as a non-convex optimization problem, and present a principled optimization method to solve it in a dynamic manner: the latent structures (e.g. the selections of local classifiers, and the sizes and locations of ROIs) are iteratively determined along with the parameter learning. Given the updated model parameters in each step, the data-driven inference is also proposed to efficiently determine the latent structures by using the sequential pruning and dynamic programming method. In the experiments, we demonstrate superior performances over the state-of-the-art approaches. We also release hundreds of CEUS FLLs videos used to quantitatively evaluate this work, which to the best of our knowledge forms the largest dataset in the literature. Please find more information at "http://vision.sysu.edu.cn/projects/fllrecog/".

Focal Liver Lesions: Computer-aided Diagnosis by Using Contrast-enhanced US Cine Recordings

“…When the CAD classifier agreed with the reader, the accuracy of the experienced and designed for computational analysis. When assessing the CAD classifiable FLLs, accuracy was 81.1% (77 of 95) and AUC was 0.883 (95% CI: 0.793, 0.940), which are less than those for other published systems, which range in accuracy from 86.4% to 92.7% when classifying lesions as benign or malignant (12,15,20) and from 84.8% to 88.3% when classifying FLLs by type (11,(13)(14)(15). This is likely because our study included 11 FLL types and attempted to broadly classify them as either benign or malignant, while the published reports were limited to three to five lesion types.…”

“…Although CAD systems are designed to minimize operator intervention, this system required an operator to draw ROIs to identify the target lesion and liver parenchyma. Recent advancements in automated segmentation algorithms could eliminate this requirement (12,15,(26)(27)(28).…”

“…Computer-aided Diagnosis of Focal Liver Lesions with Contrast-enhanced US Ta et al (11)(12)(13)(14). Additional features explored included relative enhancement between the lesion and the parenchyma (11)(12)(13)(14)(15), lesion rim versus lesion center (11,(13)(14)(15), enhancement homogeneity (15), lesion morphology (11)(12)(13), and vessel morphology (13). Their performance, however, has been tested on a preselected set of three to five liver lesion types.…”

“…Cross-validation testing is a standard method used to estimate CAD performance. Cines are randomly divided into 10 partitions, and the CAD is trained on all cines except for one partition that is used for testing (11)(12)(13)(14)(15)19,20). The process is repeated 10 times by using a different partition for ULTRASONOGRAPHY: Computer-aided Diagnosis of Focal Liver Lesions with Contrast-enhanced US Ta et al were 8.7 cm 2 (0.4-70 cm 2 ) and 3.1 cm 2 (0.5-9.2 cm 2 ), respectively.…”

Three-way decision support for diagnosis on focal liver lesions

“…They also provided a computational framework to assist doctors for FLL diagnosis [35]. In this framework, a structured model is proposed to detect a number of discriminative Regions of Interest (ROIs) for the FLL recognition.…”