Publication
Ultrasound versus elastography in the study of hepatic steatosis
| Summary: | Fatty Liver Disease, or steatosis, is a condition where excess fat builds up in the liver cells. Its prevalence is on the rise, posing a significant global health concern. If left untreated, it can progress into more serious liver diseases, such as steatohepatitis, fibrosis, liver cancer or even chirrosis. Therefore, it is essential to accurately diagnose the condition at an early stage in orderto facilitate more effective intervention and management.The aim of this study is to compare the performance of ultrasound and elastography imaging techniques for steatosis diagnosis with artificial intelligence methods, understanding if elastography could be of similar relevance as ultrasound, and also to evaluate the performance of all trained models. To this end, this study uses ultrasound and elastography images to classify liver steatosis using different classical machine learning classifiers and deep learning architectures, respectively: AdaBoost, Decision Tree, K-Nearest Neighbours, Logistic Regression, Neural Network, Random Forest, Support Vector Machine with Linear and RBF kernels, DenseNet201,InceptionResNetV2, ResNet101, ResNet50V2, VGG16 and VGG19. After careful tuning and optimisation of various parameters, the trained models were compared based on their performance. The shallow neural network showed the best performance, achieving an F1 score of 99.5% on the ultrasound set, 99.2% on the elastography set and 98.9% on the mixed set. On the other hand, despite the lower overall performance of deep learning, the results of this approach are comparable to those of machine learning, where DenseNet201 achieved an F1 score of 98.80% on the ultrasound set and ResNet50V2 achieved 98.23% and 97.58% on the elastography andmixed sets, respectively.In terms of imaging, the results obtained with the set of ultrasound images confirm its superiority as an imaging method for diagnosing steatosis, in comparison to elastography, although not by a significantly different margin.This research offers valuable insights into the domain of medical image classification and advocates the integration of advanced machine learning and deep learning technologies in diagnosing steatosis. |
|---|---|
| Main Authors: | Marques, Rodrigo Ramos |
| Subject: | Ultrasound Elastography Steatosis Machine Learning Deep Learning Ecografia Elastografia Esteatose Machine Learning Deep Learning |
| Year: | 2024 |
| Country: | Portugal |
| Document type: | master thesis |
| Access type: | open access |
| Associated institution: | Universidade de Coimbra |
| Language: | English |
| Origin: | Estudo Geral - Universidade de Coimbra |
| Summary: | Fatty Liver Disease, or steatosis, is a condition where excess fat builds up in the liver cells. Its prevalence is on the rise, posing a significant global health concern. If left untreated, it can progress into more serious liver diseases, such as steatohepatitis, fibrosis, liver cancer or even chirrosis. Therefore, it is essential to accurately diagnose the condition at an early stage in orderto facilitate more effective intervention and management.The aim of this study is to compare the performance of ultrasound and elastography imaging techniques for steatosis diagnosis with artificial intelligence methods, understanding if elastography could be of similar relevance as ultrasound, and also to evaluate the performance of all trained models. To this end, this study uses ultrasound and elastography images to classify liver steatosis using different classical machine learning classifiers and deep learning architectures, respectively: AdaBoost, Decision Tree, K-Nearest Neighbours, Logistic Regression, Neural Network, Random Forest, Support Vector Machine with Linear and RBF kernels, DenseNet201,InceptionResNetV2, ResNet101, ResNet50V2, VGG16 and VGG19. After careful tuning and optimisation of various parameters, the trained models were compared based on their performance. The shallow neural network showed the best performance, achieving an F1 score of 99.5% on the ultrasound set, 99.2% on the elastography set and 98.9% on the mixed set. On the other hand, despite the lower overall performance of deep learning, the results of this approach are comparable to those of machine learning, where DenseNet201 achieved an F1 score of 98.80% on the ultrasound set and ResNet50V2 achieved 98.23% and 97.58% on the elastography andmixed sets, respectively.In terms of imaging, the results obtained with the set of ultrasound images confirm its superiority as an imaging method for diagnosing steatosis, in comparison to elastography, although not by a significantly different margin.This research offers valuable insights into the domain of medical image classification and advocates the integration of advanced machine learning and deep learning technologies in diagnosing steatosis. |
|---|