scholarly journals Radiomics for Gleason Score Detection through Deep Learning

Sensors ◽  
2020 ◽  
Vol 20 (18) ◽  
pp. 5411
Author(s):  
Luca Brunese ◽  
Francesco Mercaldo ◽  
Alfonso Reginelli ◽  
Antonella Santone

Prostate cancer is classified into different stages, each stage is related to a different Gleason score. The labeling of a diagnosed prostate cancer is a task usually performed by radiologists. In this paper we propose a deep architecture, based on several convolutional layers, aimed to automatically assign the Gleason score to Magnetic Resonance Imaging (MRI) under analysis. We exploit a set of 71 radiomic features belonging to five categories: First Order, Shape, Gray Level Co-occurrence Matrix, Gray Level Run Length Matrix and Gray Level Size Zone Matrix. The radiomic features are gathered directly from segmented MRIs using two free-available dataset for research purpose obtained from different institutions. The results, obtained in terms of accuracy, are promising: they are ranging between 0.96 and 0.98 for Gleason score prediction.

2021 ◽  
Vol 15 (9) ◽  
Author(s):  
Ryan Sun ◽  
Andrew Fast ◽  
Iain Kirkpatrick ◽  
Patrick Cho ◽  
Jeffery Saranchuk

Introduction: The role of magnetic resonance imaging (MRI)-fusion biopsy (FB) remains unclear in men with prior negative prostate biopsies. This study aimed to compare the diagnostic accuracy of FB with concurrent systematic biopsy (SB) in patients requiring repeat prostate biopsies. Methods: Patients with previous negative prostate biopsies requiring repeat biopsies were included. Those without suspicious lesions (≥Prostate Imaging Reporting and Data System [PI-RADS] 3) on MRI were excluded. All patients underwent FB followed by SB. The primary outcome was the sensitivity for clinically significant prostate cancer (Gleason score ≥7). The secondary objective was identification of potential predictive factors of biopsy performance. Results: A total of 53 patients were included; 41 (77%) patients were found to have clinically significant prostate cancer. FB had a higher detection rate of significant cancer compared to SB (85% vs. 76%, respectively, p=0.20) and lower diagnosis of indolent (Gleason score 3+3=6) cancer (10% vs. 27%, respectively, p=0.05). FB alone missed six (15%) clinically significant cancers, compared to 10 (24%) with SB. SB performance was significantly impaired in patients with anterior lesions and high prostate volumes (p<0.05). There was high degree of pathological discordance between the two approaches, with concordance seen in only 34% of patients. Conclusions: In patients with prior negative biopsies and ongoing suspicion for prostate cancer, a combined approach of FB with SB is needed for optimal detection and risk classification of clinically significant disease. Anterior tumors and large prostates were significant predictors of poor SB performance and an MRI-fusion alone approach in these settings could be considered.


2021 ◽  
Vol 14 (3) ◽  
pp. 86-93
Author(s):  
R.A. Romanov ◽  
◽  
A.V. Koryakin ◽  
A.V. Sivkov ◽  
B.Ya. Alekseev ◽  
...  

Introduction. Significant improvement in the quality of visualization of the prostate using magnetic resonance imaging (MRI), as well as the development of technologies for virtual combination of MRI and ultrasound images opens new horizons in the diagnosis of prostate cancer. The introduction of the PI-RADS system has allowed the standardization of MRI findings, and the development of fusion biopsy systems seeks to make diagnostics more accurate and less operator-dependent. Materials and methods. In this literature review, we evaluate the effectiveness of various biopsy approaches and discuss the prospects for targeted biopsies. The search for publications was carried out in the databases PubMed, e-library, Web of Scince et al. For citation, 55 literature sources were selected that met the search criteria for the keywords, «prostate cancer», «biopsy», «MRI», «TRUS», «fusion». Results. Diagnosis of prostate cancer using MRI. Modern technologies for radiological diagnosis of prostate cancer using magnetic resonance imaging (MRI) are based on the standardized PI-RADS protocol, using different modes (T2, diffusion-weighted images and contrast enhancement), which provides the best visualization of tumor-suspicious nodes in the prostate gland, allowing determination of lesion localization and size for subsequent targeted biopsy. Options for performing a prostate biopsy to diagnose prostate cancer. A description of the methods and effectiveness of transrectal and transperineal biopsy under ultrasound guidance is carried out - due to the fact that ultrasound diagnostics of prostate cancer has a rather low sensitivity due to small differences in the ultrasound structure of normal and tumor tissue of the prostate, an extended template biopsy technique was proposed, which involves puncture of the prostate through a special lattice. It also describes the technology of fusion biopsy and also provides literature data comparing the diagnostic accuracy of standard TRUS and fusion prostate biopsy, as well as the importance of transrectal / transperineal access. Questions for further study. Given the desire to reduce the number of biopsies while maintaining or even increasing the accuracy of diagnosing prostate cancer, data from studies investigating the feasibility of combining polyfocal (non-targeted) and targeted (targeted) biopsies are presented. Conclusion. The existing methods of non-targeted biopsy (polyfocal, saturation, template) and targeted (fusion biopsy) have their advantages and disadvantages, which currently do not allow making certain recommendations for their use, but a significant number of authors prefer MRI-as sisted, fusion -biopsy.


2019 ◽  
Vol 18 (2) ◽  
Author(s):  
Ida Bagus Leo Mahadya Suta ◽  
Rukmi Sari Hartati ◽  
Yoga Divayana

Tumor otak menjadi salah satu penyakit yang paling mematikan, salah satu jenis yang paling banyak ditemukan adalah glioma sekitar 6 dari 100.000 pasien adalah penderita glioma. Citra digital melalui Magnetic Resonance Imaging (MRI) merupakan salah satu metode untuk membantu dokter dalam menganalisa dan mengklasifikasikan jenis tumor otak. Namun, klasifikasi secara manual membutuhkan waktu yang lama dan memiliki resiko kesalahan yang tinggi, untuk itu dibutuhkan suatu cara otomatis dan akurat dalam melakukan klasifikasi citra MRI. Convolutional Neural Network (CNN) menjadi salah satu solusi dalam melakukan klasifikasi otomatis dalam citra MRI. CNN merupakan algoritma deep learning yang memiliki kemampuan untuk belajar sendiri dari kasus kasus sebelumnya. Dan dari penelitian yang telah dilakukan, diperoleh hasil bahwa CNN mampu dalam menyelesaikan klasifikasi tumor otak dengan akurasi yang tinggi. Peningkatan akurasi diperoleh dengan mengembangkan algoritma CNN baik melalui menentukan nilai kernel dan/atau fungsi aktivasi.


Sign in / Sign up

Export Citation Format

Share Document