Biography writer

Lung cancer is one of the most common diseases, and it has the highest morbidity and mortality rates of any cancer worldwide. As reported by global cancer statistics [1], there were 2.1 million new cases and 1.8 million deaths from lung cancer in 2018. Approximately one in five cases (18.4%) of lung cancer leads to death. In oncology, medical imaging plays a vital role in reducing the mortality rate. Imaging is a non-invasive and painless procedure with the least harmful side eects for patients. It can provide detailed anatomical information about the disease by generating visual representations of tumors. Unlike other invasive methods such as surgeries and biopsies which extract and study a small portion of tumor tissue, imaging can give a more comprehensive view and analysis of the entire tumor Appl. Sci. 2019, 9, 2329; doi:10.3390/app9112329 [login to view URL] Appl. Sci. 2019, 9, 2329 2 of 26 region. Moreover, it is preferable for clinical routines that require an iterative analysis of the tumor during treatment [2]. Computed tomography (CT) is a standard imaging modality for lung cancer diagnosis. The National Lung Screening Trial (NLST) reported that the lung cancer mortality rate of 15–20% could be reduced by performing low-dose CT screening [3]. CTs display tumors on cross-sectional images of the body called “slices”. These slices can be stacked together to reconstruct a three-dimensional structure which makes possible more comprehensive analysis of tumors. However, the manual interpretation of the CT scans is prohibitively expensive in terms of time and eort. Also, it is subject to the skill and clinical practices of the interpreter; hence, the diagnosis results may suer from inter- and intra-observer variation [4]. Due to these inconveniences, any real-world application is necessary for automated interpretations of the CT scans. This fact has strongly motivated computer-aided diagnosis systems (CADs) to become an extensive research area in the field of biomedical engineering. Scientists have proposed a vast number of CADs for lung cancer diagnosis using modern image processing and machine learning techniques. Nonetheless, most of these CADs have focused only on the detection of pulmonary nodules. Nodules can evoke the possibility of lung cancer and appear as round or oval-shaped opacities on the CT scans [5]. Nonetheless, the detection of nodules does not provide sucient information about the severity of the disease to make proper treatment decisions. Generally, half of the patients who undergo CT screening present more than one nodule, but not all of these nodules are cancerous [6]. Physicians usually use the terms “nodule” and “tumor” interchangeably when they are uncertain about the severity of the disease. Indeed, identifying the severity of a tumor nodule is a principal issue for lung cancer diagnosis. In clinical practice, the severity of lung cancer can be expressed as dierent stages using the Tumor, Node, and Metastasis (TNM) system [7]. As the name describes, this system assesses the severity of the disease based on three descriptors. The first descriptor (T) assesses the characteristics of the primary tumor such as its size, local invasion, and the presence of satellite tumor nodules. The CT modality performs T-staging well, because information about the tumor can be easily obtained from the CT scans. The second descriptor (N) assesses the involvement of the regional lymph nodes. On the CT scans, lymph nodes appear as small opacities with unclear silhouettes; thus, the CT modality performs weakly for N-staging [8,9].