Multiple instance learning combined with label invariant synthetic data for guiding systematic prostate biopsy: a feasibility study
Autor: | Mohammad H. Jafari, Golara Javadi, Antonio Hurtado, Silvia D. Chang, Claudia Kesch, Peter C. Black, Sharareh Bayat, Samira Sojoudi, Purang Abolmaesumi, Mehran Pesteie, Samareh Samadi, Parvin Mousavi |
---|---|
Rok vydání: | 2019 |
Předmět: |
Image-Guided Biopsy
Male Prostate biopsy Computer science 0206 medical engineering Biomedical Engineering Health Informatics 02 engineering and technology Sensitivity and Specificity Synthetic data 030218 nuclear medicine & medical imaging 03 medical and health sciences Prostate cancer 0302 clinical medicine Prostate Biopsy medicine Humans Radiology Nuclear Medicine and imaging Sensitivity (control systems) Invariant (computer science) Ultrasonography Interventional medicine.diagnostic_test business.industry Prostatic Neoplasms Pattern recognition General Medicine Pathology Report medicine.disease 020601 biomedical engineering Computer Graphics and Computer-Aided Design 3. Good health Computer Science Applications medicine.anatomical_structure Feasibility Studies Surgery Computer Vision and Pattern Recognition Artificial intelligence Neural Networks Computer business |
Zdroj: | International journal of computer assisted radiology and surgery. 15(6) |
ISSN: | 1861-6429 |
Popis: | Ultrasound imaging is routinely used in prostate biopsy, which involves obtaining prostate tissue samples using a systematic, yet, non-targeted approach. This approach is blinded to individual patient intraprostatic pathology, and unfortunately, has a high rate of false negatives. In this paper, we propose a deep network for improved detection of prostate cancer in systematic biopsy. We address several challenges associated with training such network: (1) Statistical labels: Since biopsy core’s pathology report only represents a statistical distribution of cancer within the core, we use multiple instance learning (MIL) networks to enable learning from ultrasound image regions associated with those data; (2) Limited labels: The number of biopsy cores are limited to at most 12 per patient. As a result, the number of samples available for training a deep network is limited. We alleviate this issue by effectively combining Independent Conditional Variational Auto Encoders (ICVAE) with MIL. We train ICVAE to learn label-invariant features of RF data, which is subsequently used to generate synthetic data for improved training of the MIL network. Our in vivo study includes data from 339 prostate biopsy cores of 70 patients. We achieve an area under the curve, sensitivity, specificity, and balanced accuracy of 0.68, 0.77, 0.55 and 0.66, respectively. The proposed approach is generic and can be applied to several other scenarios where unlabeled data and noisy labels in training samples are present. |
Databáze: | OpenAIRE |
Externí odkaz: |