Unsupervised tumor detection in Dynamic PET/CT imaging of the prostate

Early detection and localization of prostate tumors pose a challenge to the medical community. Several imaging techniques, including PET, have shown some success. But no robust and accurate solution has yet been reached. This work aims to detect prostate cancer foci in Dynamic PET images using an unsupervised learning approach. The proposed method extracts three feature classes from 4D imaging data that include statistical, kinetic biological and deep features that are learned by a deep stacked convolutional autoencoder. Anomalies, which are classified as tumors, are detected in feature space using density estimation. The proposed algorithm generates promising results for sufficiently large cancer foci in real PET scans imaging where the foci is not viewed by the tomographic devices used for detection.

[1]  J. Pruim,et al.  Visualization of prostate cancer with 11C-choline positron emission tomography. , 2002, European urology.

[2]  Martin J. Wainwright,et al.  Noisy matrix decomposition via convex relaxation: Optimal rates in high dimensions , 2011, ICML.

[3]  Leyun Pan,et al.  Quantitative approaches of dynamic FDG-PET and PET/CT studies (dPET/CT) for the evaluation of oncological patients , 2012, Cancer imaging : the official publication of the International Cancer Imaging Society.

[4]  Keith Godfrey,et al.  Compartmental Models and Their Application , 1983 .

[5]  Ernest J. Feleppa,et al.  Prostate-cancer imaging using machine-learning classifiers: Potential value for guiding biopsies, targeting therapy, and monitoring treatment , 2009, 2009 IEEE International Ultrasonics Symposium.

[6]  F. Turkheimer,et al.  Kinetic modeling in positron emission tomography. , 2002, The quarterly journal of nuclear medicine : official publication of the Italian Association of Nuclear Medicine (AIMN) [and] the International Association of Radiopharmacology.

[7]  Nathalie Japkowicz,et al.  A Novelty Detection Approach to Classification , 1995, IJCAI.

[8]  M. Conti.  New prospects for PET in prostate cancer imaging: a physicist's viewpoint , 2014, EJNMMI Physics.

[9]  Yaozong Gao,et al.  Deformable MR Prostate Segmentation via Deep Feature Learning and Sparse Patch Matching , 2016, IEEE Transactions on Medical Imaging.

[10]  Radford M. Neal.  Pattern Recognition and Machine Learning , 2007, Technometrics.

[11]  D. W. Scott,et al.  Multidimensional Density Estimation , 2005 .

[12]  Mario Marengo,et al.  Detection and localization of prostate cancer: correlation of (11)C-choline PET/CT with histopathologic step-section analysis. , 2005, Journal of nuclear medicine : official publication, Society of Nuclear Medicine.

[13]  Jürgen Schmidhuber,et al.  Stacked Convolutional Auto-Encoders for Hierarchical Feature Extraction , 2011, ICANN.

[14]  Simon J. Doran,et al.  Autoencoder in Time-Series Analysis for Unsupervised Tissues Characterisation in a Large Unlabelled Medical Image Dataset , 2011, 2011 10th International Conference on Machine Learning and Applications and Workshops.

[15]  Geoffrey E. Hinton,et al.  Deep Learning , 2015, Nature.

[16]  L. Bachmann,et al.  The role of 11C-choline and 18F-fluorocholine positron emission tomography (PET) and PET/CT in prostate cancer: a systematic review and meta-analysis. , 2013, European urology.

[17]  Malik Yousef,et al.  One-class document classification via Neural Networks , 2007, Neurocomputing.

[18]  Peter L Choyke,et al.  Imaging prostate cancer: a multidisciplinary perspective. , 2007, Radiology.

[19]  J. Kurhanewicz,et al.  Prostate cancer: sextant localization with MR imaging, MR spectroscopy, and 11C-choline PET/CT. , 2007, Radiology.

[20]  H. Malcolm Hudson,et al.  Accelerated image reconstruction using ordered subsets of projection data , 1994, IEEE Trans. Medical Imaging.

[21]  R H Huesman,et al.  Simulation of compartmental models for kinetic data from a positron emission tomograph. , 1992, Computer methods and programs in biomedicine.

[22]  V. Bettinardi,et al.  [11C]Choline uptake with PET/CT for the initial diagnosis of prostate cancer: relation to PSA levels, tumour stage and anti-androgenic therapy , 2008, European Journal of Nuclear Medicine and Molecular Imaging.

[23]  Cauligi S. Raghavendra,et al.  Autoencoder-derived Features as Inputs to Classification Algorithms for Predicting Well Failures , 2015 .

[24]  Geoffrey E. Hinton,et al.  On the importance of initialization and momentum in deep learning , 2013, ICML.

[25]  Rongrong Guo,et al.  Automatic segmentation of the prostate on CT images using deep learning and multi-atlas fusion , 2017, Medical Imaging.

[26]  M. Schwaiger,et al.  [(11)C]Choline positron emission tomography/computed tomography for staging and restaging of patients with advanced prostate cancer. , 2008, Nuclear medicine and biology.

[27]  Robert A. Lordo,et al.  Nonparametric and Semiparametric Models , 2005, Technometrics.

[28]  Wei Li,et al.  Learning Image Context for Segmentation of Prostate in CT-Guided Radiotherapy , 2011, MICCAI.

[29]  B. Krause,et al.  Choline PET and PET/CT in Primary Diagnosis and Staging of Prostate Cancer , 2012, Theranostics.

[30]  Bram van Ginneken,et al.  A survey on deep learning in medical image analysis , 2017, Medical Image Anal..

[31]  P. Waldenberger,et al.  18F choline PET/CT in the preoperative staging of prostate cancer in patients with intermediate or high risk of extracapsular disease: a prospective study of 130 patients. , 2010, Radiology.

[32]  Hyunjin Park,et al.  Detection of Aggressive Primary Prostate Cancer with 11C-Choline PET/CT Using Multimodality Fusion Techniques , 2009, Journal of Nuclear Medicine.

[33]  Nitish Srivastava,et al.  Dropout: a simple way to prevent neural networks from overfitting , 2014, J. Mach. Learn. Res..

[34]  VARUN CHANDOLA,et al.  Anomaly detection: A survey , 2009, CSUR.