Intelligent Imaging Informatics and Computer Aided Diagnosis and Control for Lung Cancer

A persistent difficulty in evaluating the pathologic specimens is its subjective nature, leading to high inter-observer variability. Recent literature has shown that, when morphologic classification is assisted with computer-aided analysis, objectivity and reproducibility improves significantly. Subtle changes in measurable parameters can potentially provide mechanistic and diagnostic clues which are not always apparent by human visual inspection. Lung cancer is one of the most serious and life threatening cancers in Kentucky in which successful prognosis and treatment planning varies greatly depending on consistent, reproducible, and accurate pathologic diagnosis and prognosis. Yet even differentiating among different types of lung cancer, such as adenocarcinomas and squamous cell carcinomas, could already be a challenging task for some cases because it depends on subjective assessment of morphologic variables of the digitized pathologic specimen. Until now, there is no such an objective, reproducible, and quantitative imaging informatics system exists in Kentucky, which can help doctors on computer aided diagnosis/prognosis, treatment outcome prediction, and therapy planning for lung cancer. In this study, population, informatics and clinical researchers in University of Kentucky and Markey Cancer Center (MCC) propose to collaborate together to design, develop, deploy and evaluate a quantitative imaging informatics system to score, classify and analyze digitized pathology lung tumor specimens for better personalized lung cancer treatment and control. The main objectives are: 1) Develop robust, high throughput medical image analysis component to analyze digitized lung tumors specimens; 2) Build a quantitative imaging informatics (QII) system and make it freely available to the clinical and research communities, and establish a large scale image database to link digitized pathology images with clinical metadata stored in Kentucky Cancer Registry and clinical data warehousing; 3) By discovering statistical relationships among image signatures, protein expression profiles, stage of disease, response to therapy, survival rate, etc., to achieve the ultimate goal to discover novel image markers which will help to tailor individualized treatment for better personalized lung cancer treatment and targeted therapeutics.