Deep learning based model observer by U-Net

Model Observers (MO) are algorithms designed to evaluate and optimize the parameters of new medical imaging reconstruction methodologies by providing a measure of human accuracy for a diagnostic task. In contrast with a computer-aided diagnosis system, MOs are not designed to outperform human diagnosis but only to find a defect if a radiologist would be able to detect it. These algorithms can economize and expedite the finding of optimal reconstruction parameters by reducing the number of sessions with expert radiologists, which are costly and prolonged. Convolutional Neural Networks (CNN or ConvNet) have been successfully used in the computer vision field for image classification, segmentation and video analytics. In this paper, we propose and test several U-Net configurations as MO for a defect localization task on synthetic images with different levels of correlated noisy backgrounds. Preliminary results show that the CNN based MO has potential and its accuracy correlates well with that of the human.

[1]  H H Barrett,et al.  Addition of a channel mechanism to the ideal-observer model. , 1987, Journal of the Optical Society of America. A, Optics and image science.

[2]  L B Lusted,et al.  Signal detectability and medical decision-making. , 1971, Science.

[3]  Q. Mcnemar.  Note on the sampling error of the difference between correlated proportions or percentages , 1947, Psychometrika.

[4]  W. W. Peterson,et al.  The theory of signal detectability , 1954, Trans. IRE Prof. Group Inf. Theory.

[5]  John A. Swets,et al.  Evaluation of diagnostic systems : methods from signal detection theory , 1982 .

[6]  C. Metz.  Basic principles of ROC analysis. , 1978, Seminars in nuclear medicine.

[7]  Rongping Zeng,et al.  Classification images for localization performance in ramp-spectrum noise. , 2018, Medical physics.

[8]  Craig K. Abbey,et al.  High human-observer efficiency for forced-localization tasks in correlated noise , 2010, Medical Imaging.

[9]  H.H. Barrett,et al.  Model observers for assessment of image quality , 1993, 2002 IEEE Nuclear Science Symposium Conference Record.

[10]  Jie Yao,et al.  Predicting human performance by a channelized Hotelling observer model , 1992, Optics & Photonics.

[11]  Felix K. Kopp,et al.  Evaluation of a machine learning based model observer for x-ray CT , 2018, Medical Imaging.

[12]  Wei Zhou,et al.  A deep learning- and partial least square regression-based model observer for a low-contrast lesion detection task in CT. , 2019, Medical physics.

[13]  J. Robson,et al.  Spatial-frequency channels in human vision. , 1971, Journal of the Optical Society of America.

[14]  Jovan G. Brankov,et al.  Numerical Surrogates for Human Observers in Myocardial Motion Evaluation From SPECT Images , 2014, IEEE Transactions on Medical Imaging.

[15]  Jongduk Baek,et al.  Implementation of an ideal observer model using convolutional neural network for breast CT images , 2019, Medical Imaging.

[16]  Jovan G Brankov,et al.  Evaluation of the channelized Hotelling observer with an internal-noise model in a train-test paradigm for cardiac SPECT defect detection , 2013, Physics in medicine and biology.

[17]  Melanie Grunwald,et al.  Foundations Of Image Science , 2016 .

[18]  Mahdi M. Kalayeh,et al.  Generalization Evaluation of Machine Learning Numerical Observers for Image Quality Assessment , 2013, IEEE Transactions on Nuclear Science.

[19]  Yongyi Yang,et al.  Learning a Channelized Observer for Image Quality Assessment , 2009, IEEE Transactions on Medical Imaging.