Effects of a deep learning network on imaging quality of low- <i>b</i> -value diffusion–weighted imaging and lesion detection in prostate cancer.

Abstract

e17163 Background: Diffusion–weighted imaging with higher b -value improves detection rate for prostate cancer lesions. However, obtaining high b -value DWI requires more advanced hardware and software configuration. Here we use a novel deep learning network, NAFNet, to generate a deep learning reconstructed (DLR 1500 ) images from 800 b -value to mimic 1500 b -value images, and to evaluate its performance and lesion detection improvements based on whole-slide images (WSI). Methods: We enrolled 303 prostate cancer patients with both 800 and 1500 b -value from Fudan University Shanghai Cancer Centre between 2017 and 2020. We assigned these patients to the training and validation set in a 2:1 ratio. The testing set included 36 prostate cancer patients from an independent institute who had only preoperative DWI at 800 b-value. Two senior radiologists and two junior radiologists read and delineated cancer lesions on DLR 1500 , original 800 and 1500 b-value DWI images. WSI were used as the ground truth to assess the lesion detection improvement of DLR 1500 images in the testing set. Results: After training and generating, within junior radiology doctors, the diagnostic AUC based on DLR 1500 images is not inferior to the that based on 1500 b -value images (0.832 (0.788-0.905) vs. 0.821 (0.747-0.899), P=0.824). The same phenomenon is also observed in senior radiology doctors. Furthermore, in the testing set, DLR 1500 images could significantly enhance junior radiology doctors’ diagnostic performance than 800 b -value images (0.848 (0.758-0.938) vs. 0.752 (0.661-0.843), P=0.043). Conclusions: DLR 1500 DWIs were comparable in quality to original 1500 b -value images within both junior and senior radiology doctors. NAFNet based DWI enhancement can significantly improve the image quality of 800 b -value DWI, and therefore promote the accuracy of prostate cancer lesion detection for junior radiology doctors. Performance comparisons between different models in prostate cancer lesions detection on testing set. Circumstances AUC on testing set (95%CI) based on WSI Delong’s Test P for comparing AUCs Sensitivity Specificity Accuracy Dice coefficient Original low- b  Junior radiology doctor 0.752 (0.661-0.843) 0.040 0.043 0.723 0.712 0.722 0.815  Senior radiology doctor 0.823 (0.745-0.901) Reference 0.129 0.872 0.813 0.821 0.865 DLR 1500  Junior radiology doctor 0.848 (0.758-0.938) 0.042 Reference 0.770 0.832 0.801 0.835  Senior radiology doctor 0.901 (0.825-0.977) Reference Reference 0.940 0.880 0.917 0.923 DLR: Deep-learning reconstructed; AUC: area under the receiver operating characteristic curve; CI: confidence interval.