Key points are not available for this paper at this time.
Lung cancer is the most prevalent cancer and is the leading cause of cancer deaths worldwide. The overall survival rate of lung cancer patients is only 14%. Lives of cancer patients can be saved if the cancer is detected in the initial stages. Positron Emission Tomography / Computed Tomography (PET/CT) is the preferred imaging modality in cancer detection with improved diagnostic accuracy due to the integration of functional (PET) and anatomical (CT) information into a single scan. Although PET/CT is advantageous over other modalities, visual inspection of these images may be an error prone task, as it is difficult to distinguish between background tissues and lung nodules and subject to inter and intra observer variability. Therefore, computational systems are essential to assist radiologists in the elucidation of images and accurate diagnosis. This paper aims at developing a methodology for automatic detection of lung cancer from PET/CT images. Image pre-processing methods such as Contrast Limited Adaptive Histogram Equalization (CLAHE) and Wiener filtering were performed to remove the artifacts due to contrast variations and noise. Lung region of interest (ROI) were extracted from images using morphological operators. Haralick statistical texture features were preferred as they extract more texture information from the cancer regions than the visual assessment. Fuzzy C means (FCM) clustering was used to classify the regions as normal or abnormal. The proposed method was carried out using PET/CT images of lung cancer patients and implemented using MATLAB. The performance of the proposed methodology was evaluated using Receiver Operating Characteristics (ROC) curve. The proposed method provides better classification and cancer detection with an overall accuracy of 92.67%.
Punithavathy et al. (Sun,) studied this question.