While acknowledging the work behind the updated Prostate Imaging Reporting and Data System (PI-RADS) Version 2.1 to facilitate standard improvements in prostate magnetic resonance imaging (MRI), the researchers noted: False-positive and false-negative results and important results are listed. differences between readers.1,2
However, new deep neural network models may be able to detect consistently high levels of clinically significant prostate cancer (csPCa) on biparametric MRI scans.
This study was recently reported Imaging InsightsThe researchers used a total of 2,736 biometrics, including 1,500 images from public multicenter and multivendor training databases, 1,036 in-house multicenter scans, and 200 scans from transfer learning datasets. Parametric MRI scans were reviewed.3 The study authors trained and developed a self-adaptive deep neural network (3D nnU-Net) using data from the previously mentioned public database Prostate Imaging: Cancer AI (PI-CAI).
According to this study, the researchers then investigated the ability of the 3D nnU-Net model to detect csPCa in biparametric MRI scans from in-house and in-house transfer learning datasets, as well as the ability of hidden validation and test sample images. was evaluated. PI-CAI dataset.
For the PI-CAI hidden validation and test datasets, the researchers noted an area under the receiver operating characteristic curve (AUROC) of 88.8 and 88.9 percent, respectively. According to the study, the 3D nnU-Net model provided 88.6 percent and 87 percent AUROC for biparametric MRI from in-house and transfer learning datasets, respectively.3
In patients with clinically significant prostate cancer, a deep learning model (3D nnU-Net) accurately predicted and established prostate cancer lesion boundaries. (Image courtesy of Insights into Imaging.)
“This model has been validated externally based on large multi-center and multi-vendor in-house data to provide similar performance in detecting csPCa at the scan level, demonstrating its robustness and generalizability.” …Notably, the performance of our model was much higher than the median AUC of 0.79 reported in identifying csPCa in earlier studies,” said study co-author, Faculty of Medicine, University of Tokyo. Erkan Karaarslan, a doctor in the Department of Radiology, writes: Ashbadem Mehmet Ali Aydinlal University, Istanbul, Turkey and colleagues.3,4
(Editor’s Note: For related content, see “Study Shows Benefits of AI in Prostate Cancer Detection on Multiparametric MRI” and “Explainable AI for Prostate Cancer Diagnosis on MRI and PI- Can RADS Classification Be Enhanced?”)
Regarding study limitations, the authors acknowledged that only MRI-visible clinically significant prostate cancer (csPCa) cases were utilized and the imaging sequence was omitted. They acknowledged the paucity of histopathology results in patients without csPCa. The researchers also did not compare deep learning assessments of prostate MRI with radiologist assessments.
References
1. Turkbey B, Rosenkrantz AB, Haider MA, et al. Prostate Imaging Report and Data System Version 2.1: 2019 update of Prostate Imaging Report and Data System Version 2. eurol. 2019;76(3):340-351.
2. Smith CP, Harmon SA, Barrett T, et al. Intra- and inter-reader reproducibility of PI-RADSv2: a multi-reader study. J Magn Reson Imaging. 2019;49(6):1694-1703.
3. Karagoz A, Alis D, Seker ME, et al. An anatomically guided self-adaptive deep neural network for the detection of clinically significant prostate cancer on biparametric MRI: a multicenter study. insight imaging. 2023;14(1):110. Doi: 10.1186/s13244-023-01439-0.
4. Castillo TJM, Arif M, Niesen WJ, et al. Automated Classification of Critical Prostate Cancer by MRI: A Systematic Review of the Performance of Machine Learning Applications. Cancers (Basel). 2020; 12:1606.
