Combining Convolutional Neural Network and Long Short-Term Memory to Classify Sinusitis

Abstract

As one of the common health problems, sinusitis is inflammation of the mucous membranes lining one or more of the paranasal sinuses. Improvement of detection tools to classify acute or chronic sinusitis is required because of its impact on the patient's treatment. In some of the previous research, deep learning has demonstrated good accuracy to classify disease. Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) are now popularly used for deep learning tasks. This research applied one dimensional (1D) CNN and its advanced modification with LSTM called 1D CNN-LSTM to classify the type of sinusitis. Data sinusitis patients are received from Cipto Mangunkusumo Hospital, Jakarta, Indonesia. This dataset consists of 200 data with four features, such as Gender, Age, Hounsfield Unit (HU), and Air Cavity. The result is 1D CNN-LSTM has higher accuracy than 1D CNN with 98,33% of accuracy.