DESCRIPTION (provided by applicant): Primarily due to the lack of physical symptoms in the early stage, lung cancer remains the leading cause of cancer deaths in the United States and worldwide. Although high-resolution computed tomography (CT) has been proved to be a sensitive, non-invasive modality for visualizing small lung nodules, which could be the early manifestation of lung cancer, a considerable number of false positive detections are often resulted. Consequently, additional procedures, such as invasive biopsy / follow-up scans, are frequently needed to verify the nature of the indeterminate nodules. The negative effects associated with these over-diagnosis procedures, such as biopsy complications, exposure to additional radiation, patient anxiety, and economic cost, significantly limits the efficacy of CT screening for early diagnosis of lung cancer. In this project, we propose to develop a computer model to quantitatively assess the nature of indeterminate nodules using a longitudinal dataset. Unlike available investigations or lung cancer risk models, we will comprehensively quantify a wide variety of properties (features) of a nodule in an unprecedented detailed manner as well as their variations over time, and synergize them with patient demographic information (e.g., age, gender, smoke history) using machine learning techniques. Not only the image features of lung nodules but also their spatial relationship with respect to important lung landmarks as well as other smoke related lung abnormalities (e.g., emphysema) will be incorporated into this model. The output of this project, namely a novel computer tool, could aid clinicians to more accurately and efficiently assess the nature of indeterminate nodules, ultimately leading to the reduction of unnecessary harm and costs to patients and the healthcare system. All these will significantly improve the efficacy of CT for early lung cancer screening by maintaining its high sensitivity while reducing false positive findings. In terms of commercial potential, the developed tool could be easily integrated with the available image information systems at medical institutions by following the widely adopted Digital Imaging and Communications in Medicine (DICOM) standard. " |