08 April 2021
14:00 Master's Defense Fully distance
Theme
Multi-label Classification of Chest X-rays using Deep Machine Learning
Student
Vinicius Teixeira de Melo
Advisor / Teacher
Zanoni Dias (Advisor) / Hélio Pedrini (Co-advisor)
Brief summary
"Chest radiography is one of the most accessible radiological exams for screening and diagnosing possible diseases in the lung and heart. In addition, this type of exam is used to identify whether devices such as pacemakers, venous catheters and tubes are positioned correctly. In recent years, much attention and efforts have been devoted to improving Computer Aided Diagnostic systems, with the classification of medical images being one of the main problems addressed. Deep Learning Techniques have been increasingly used to provide predictions of detection and classification of pathologies and injuries in chest X-ray images Considering this information, we propose a method for the classification of chest X-ray images, called DuaLAnet, using deep learning techniques, such as convolutional neural networks and attention modules. aim to explore the complementarity between neural networks with volitional, with the use of attention modules to direct learning about classes, showing that the combination of complementary information extracted from chest X-ray images has a better prediction rate than when using only a neural network. To validate our method, we use the ChestX-ray14 and CheXpert databases, which have a wide variety of chest X-ray images with 14 classes each. We carried out experiments to verify the best way to initialize the weights of the neural networks, considering the initialization from ImageNet and from the radiography database that is not being used in the training. In addition, we experimented with four types of architectures and their variations to see which neural networks we should use as feature extractors. Then, we checked which attention module was best suited to each feature extractor chosen previously, among the following attention module options: Class Activation Mapping (CAM), Soft Activation Mapping (SAM) and Feature Pyramid Attention (FPA). Finally, we carried out the experiments with the DuaLAnet method, after choosing the settings that best fit each database. The results obtained show that our method has a competitive AUROC hit rate, compared to the state of the art methods in the ChestX-ray database14, and several ways we can follow to improve the hit rate in the CheXpert database . "
Examination Board
Headlines:
| Zanoni Dias | IC / UNICAMP |
| Levy Boccato | FEEC / UNICAMP |
| Alexandre Mello Ferreira | IC / UNICAMP |
Substitutes:
| Sandra Eliza Fontes de Avila | IC / UNICAMP |
| Moacir Antonelli Ponti | ICMC / USP |