According to World Health Organization figures (GLOBOCAN 2008), 1.6 million people were diagnosed worldwide with lung cancer in 2008, and 1.3 million died - one in every five cancer deaths. In the United States, lung cancer is taking as many lives each year as the combined total of lives lost to colon, breast, prostate and pancreatic cancers, the next four most lethal cancers. The five year survival rate is still only 15%. This is due in large part to the fact that only 16% of cases are being diagnosed at an early curable stage.
The research effort to better understand, detect, and effectively treat lung cancer is increasingly relying upon the acquisition and quantitative interpretation of radiological imaging studies such as Computed Tomography (CT) and Positron Emission Tomography (PET).
Advancement in this field depends heavily on the analysis and study of image databases containing large collections of patient data. Although numerous medical imaging conferences and workshops have made the recommendation to create a large and freely available image database resource and several attempts have been made to create one, none have achieved an open collection with the size and quality needed.
The Give A Scan® program was designed to answer this need. The program also responds to the keen desire of lung cancer patients to play a direct role in increasing the pace of research and encouraging more researchers around the world to address the world's leading cancer killer.
A large and open imaging database has the potential to accelerate several important areas of lung cancer research including lung cancer screening, computer aided detection and diagnosis, and the development of quantitative methods for drug therapy assessment. All rely heavily on the ability to extract statistically meaningful insights and observations from real-world clinical data.
The study of a large collection of data helps researchers avoid the common pitfall of investing valuable research time and effort on the study of a biased dataset. A recent review of 60 quantitative imaging papers covering 20 years of research on MRI bias field correction found that the median number of datasets used to support claims in publications has risen to 15. This is far from the numbers of cases needed to make claims on the diversity of data available in a clinical setting and leaves the conclusions of many scientific investigations in question. The field is also in great need of an open image database in order to objectively compare the performance of developed quantitative methods.
Currently most observations and results are reported on proprietary databases and, as a result, it is extremely difficult to determine the strengths and weaknesses of competing methods. A recent article from the Editor-in-Chief of IEEE Transactions on Image Processing highlighted the critical need for the field to move more toward reproducible research and the availability of open datasets.
Finally, the utilization of a large and open image database, if adopted widely, will help build a higher level of scientific consensus and in so doing help change the clinical standard of care for lung cancer.