fMRI-Based Hierarchical SVM Model for the Classification and Grading of Liver Fibrosis

Abstract

We present a novel method for the automatic classification and grading of liver fibrosis based on hepatic hemodynamic changes measured noninvasively from functional MRI (fMRI) scans combined with hypercapnia and hyperoxia. The supervised learning method automatically creates a classification and grading model for liver fibrosis grade from training datasets. It constructs a statistical model of liver fibrosis by evaluating the signal intensity time course and local variance in T2(*)-W fMRI scans acquired during the breathing of air, air-carbon dioxide, and carbogen with a hierarchical multiclass binary-based support vector machine (SVM) classifier. Two experimental studies on 162 slices from 34 mice with the hierarchical multiclass binary-based SVM classifier yield 96.9% separation accuracy between healthy and histological-based fibrosis graded subjects, and an overall accuracy of 75.3% for healthy, fibrotic, and cirrhotic subjects. These results outperform existing image-based methods that can discriminate between healthy and mild-grade fibrosis subjects.

Quantifying hepatic fibrosis using a biexponential model of diffusion weighted imaging in ex vivo liver specimens

“…The optimal method of combining multiple complementary channels of MRI data, such as those derived from the biexponential model employed herein, is an active area of ongoing research work and various approaches including such statistical analyses as employed here as well as artificial intelligence and machine learning approaches have been explored [34][35][36][37]. In fact, a recent machine learning-based approach, combining MRI parameters derived from hepatic hemodynamic changes demonstrated strong correlations with hepatic fibrosis [38]. While the correlation of the combination of the biexponential model derived parameters and hepatic fibrosis did not exceed that achieved with the monoexponential model-derived diffusion coefficients, the multiple parameters derived with this approach may offer utility in mitigating confounding factors in deriving liver diffusion coefficients, such as the presence of hepatic steatosis, known to affect diffusion coefficients [39,40].…”

Object recognition in clutter: cortical responses depend on the type of learning

“…The goal of this analysis was to determine the extent to which the BOLD responses in a given region were diagnostic of the subject’s response during either clutter condition. For analyses of this sort where the diagnosticity of a given marker (or classifier) is of greater interest than the nature of the marker itself, ROC analysis is a principled, although by no means the only available, choice (Reddy et al, 2006; Zhang et al, 2008; Diana et al, 2010; Sela et al, 2011; Wee et al, 2012). …”

Unsupervised abnormality detection through mixed structure regularization (MSR) in deep sparse autoencoders

“…Automatic detection of abnormalities in medical images has the potential to provide objective and accurate analysis of medical images, reduce radiologists workload and overall health‐related costs . Traditionally, discriminative machine learning approaches were used to design algorithms that are capable to distinguish between normal and abnormal samples . In the past few years, discriminative deep‐learning approaches demonstrated substantial improvement over traditional machine learning approaches in different tasks including image classification and segmentation .…”