Long Thorax Disease Classification Using Convolutional Long Short Term Memory

According to the National Heart, Lung, and Blood Institute (NHLBI), lung thorax diseases like lung nodule, enema, mass, fibrosis among others are one of the most common causes of death globally, the British Thoracic Society (BTS) reported that for each year respiratory diseases are responsible for one in five deaths, of the reported 580,000 deaths, lung cancer was responsible for 35,000 deaths, pneumonia for 34,000, and deaths from COPD was 27,000, the same organisation reported that over 845,000 cases were reported each year. Developing better diagnosis processes for treatment using deep learning methods like Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) have proven to work quite well together in a hybridized architectural model for computer-aided diagnosis systems in the medical field. In this paper, we develop a model that makes use of hybridisation of conventional neural networks and long short term memory network in the multi-class labelling of lung thorax diseases. We explore the architecture of convolutional long short-term memory (ConvLSTM) in classification of thorax diseases using a Xray dataset from the National Institute of Health (NIH) as the data source containing 112,120 images with fourteen(14) diseases label of thorax diseases to help develop an improved hybridized model to help in the diagnosis of multiple lung thorax diseases.