I and my senior PhD student Sarfaraz Hussein have published a paper in IPMI 2017, which is a premier conference in medical image computing field, and occur every two years. Information Processing in Medical Imaging (IPMI) is the latest in a series where novel developments in the acquisition, formation, analysis and display of medical images are presented, discussed, dissected, and extended. Over the lat three decades, IPMI has evolved with the medical imaging community it serves.
Here are the paper’s details, please click on the title to download the PDF:
Sarfaraz Hussein, Kunlin Cao, Qi Song, and Ulas Bagci, Risk Stratification of Lung Nodules Using 3D CNN-Based Multi-task Learning, Information Processing in Medical Imaging 2017 (IPMI 2017), Boone, NC, June 25-30, 2017
Figure shows the fffect of fine-tuning on 3D CNN features. t-SNE visualization for features obtained from (a) pre-trained network and (b) network after fine-tuning. Separation between features belonging to two classes, i.e. benign nodules (represented in blue) and malignant nodules (shown in red) can be readily observed in (b).