Ultrasound-Based Detection of Prostate Cancer Using Automatic Feature Selection with Deep Belief Networks
We propose an automatic feature selection framework for analyzing temporal ultrasound signals of prostate tissue. The framework consists of: 1) an unsupervised feature reduction step that uses Deep Belief Network (DBN) on spectral components of the tempor
- PDF / 657,428 Bytes
- 8 Pages / 439.363 x 666.131 pts Page_size
- 111 Downloads / 164 Views
The University of British Columbia, Vancouver, BC, Canada Philips Research North America, Briarcliff Manor, NY, USA 3 National Institutes of Health, Bethesda, MD, USA 4 National Cancer Institute, Bethesda, MD, USA 5 IBM Almaden Research Center, San Jose, CA, USA 6 Queen’s University, Kingston, ON, Canada
Abstract. We propose an automatic feature selection framework for analyzing temporal ultrasound signals of prostate tissue. The framework consists of: 1) an unsupervised feature reduction step that uses Deep Belief Network (DBN) on spectral components of the temporal ultrasound data; 2) a supervised fine-tuning step that uses the histopathology of the tissue samples to further optimize the DBN; 3) a Support Vector Machine (SVM) classifier that uses the activation of the DBN as input and outputs a likelihood for the cancer. In leave-one-core-out cross-validation experiments using 35 biopsy cores, an area under the curve of 0.91 is obtained for cancer prediction. Subsequently, an independent group of 36 biopsy cores was used for validation of the model. The results show that the framework can predict 22 out of 23 benign, and all of cancerous cores correctly. We conclude that temporal analysis of ultrasound data can potentially complement multi-parametric Magnetic Resonance Imaging (mp-MRI) by improving the differentiation of benign and cancerous prostate tissue. Keywords: Temporal ultrasound data, deep learning, deep belief network, cancer diagnosis, prostate cancer, feature selection, classification.
1
Introduction
The early diagnosis of prostate cancer, as the most common type of diagnosed malignancy in North American men, plays an important role in the choice and the success of treatments [8]. The definitive diagnosis of prostate cancer is histopathological analysis of biopsy tissue samples, which is typically guided under Transrectal Ultrasound (TRUS). However, conventional systematic biopsy under TRUS guidance does not have high sensitivity and specificity. Recently, c Springer International Publishing Switzerland 2015 N. Navab et al. (Eds.): MICCAI 2015, Part II, LNCS 9350, pp. 70–77, 2015. DOI: 10.1007/978-3-319-24571-3_9
Ultrasound-Based Detection of Prostate Cancer
71
fusion of multi-parametric MRI (mp-MRI) with TRUS has shown significant potential for improved cancer yield [7]. A meta-analysis of seven mp-MRI studies with 526 patients show specificity of 0.88 and sensitivity of 0.74, with negative predictive values ranging from 0.65 to 0.94 [9]. mp-MRI suffers from several limitations including low sensitivity for detection of small lesions and low grade cancer. Hence, there is a need to develop new imaging techniques that can complement mp-MRI for prostate cancer diagnosis. Over the past two decades, several ultrasound-based techniques have been proposed for characterizing cancerous tissue [8]. The clinical uptake of these methods has been slow, mainly due to the large variability of the tissue characterization results. The primary sources of such variability are the heterogeneous patient population and ca
Data Loading...
