一种透明、轻量且可持续的绿色学习AI模型用于MRI前列腺癌检测
本研究开发了绿色学习(GL)模型,用于双参数MRI上自动前列腺分割(PS)和临床显著前列腺癌(csPCa)检测,并与PI-RADS标准和U-Net深度学习模型比较。602例MRI随机分为训练集483例和测试集119例,其中224例活检证实csPCa。GL分割Dice系数高于U-Net(0.91 vs 0.88,P<0.001),csPCa检测AUC与PI-RADS(0.75 vs 0.76,P=0.8)及U-Net(0.74,P=0.3)相似,GL联合PI-RADS的AUC达0.81,高于单用PI-RADS(P=0.02)。
要点
- GL模型全自动分割前列腺的Dice系数优于U-Net(0.91 vs 0.88)。
- GL检测csPCa的AUC与PI-RADS和U-Net相当。
- GL联合PI-RADS可将csPCa检测AUC提升至0.81。
核心结论
- 新型GL模型可在双参数MRI上全自动检测csPCa,性能与PI-RADS和深度学习模型相当。
- GL与PI-RADS联合可显著提高csPCa检测性能。
研究设计与样本
- 回顾性识别接受3-T MRI和前列腺活检的男性,MRI按PI-RADS v2或v2.1采集和解读。
- 共602例MRI随机分为训练集483例和测试集119例,其中224例活检证实csPCa。
- 开发GL模型自动化前列腺分割和双参数MRI上的csPCa检测,并与放射科医生PI-RADS及U-Net深度学习模型对比。
主要发现
- 前列腺分割中位Dice系数GL高于U-Net(0.91 vs 0.88,P<0.001)。
- 患者水平csPCa检测AUC:GL为0.75,PI-RADS为0.76(P=0.8),U-Net为0.74(P=0.3)。
- GL联合PI-RADS的AUC为0.81,高于单用PI-RADS(P=0.02)。
- GL参数量为1.21×10^6,小于U-Net的177×10^6;计算量为9.8×10^9 FLOPs,少于U-Net的1027×10^9 FLOPs。
临床怎么用
- GL可作为轻量级AI工具辅助双参数MRI上前列腺分割和csPCa检测。
- GL与PI-RADS联合判读可能提高csPCa检出准确度。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
原文
OBJECTIVES: To develop a novel transparent and lightweight machine learning model, the Green Learning (GL), for automated prostate segmentation (PS) and clinically significant prostate cancer (csPCa) detection on magnetic resonance imaging (MRI). PATIENTS AND METHODS: Men who underwent 3-T MRI and prostate biopsy (PBx) were identified. MRI was acquired and interpreted according to the Prostate Imaging-Reporting and Data System (PI-RADS), version 2 or 2.1. The GL was created to automate PS and csPCa detection on biparametric MRI. The performance was compared to the standard-of-care radiologists using PI-RADS, and a conventional deep learning (DL) U-Net model as benchmarking. The PS performance was evaluated by the Dice similarity coefficient (DSC). The area under the curve (AUC) for patient-level csPCa detection was assessed. Model size and computational workload, measured by floating point operations (FLOPs), were reported. RESULTS: A total of 602 MRIs were randomly divided for training (N = 483) and testing (N = 119). Overall, 224 patients had csPCa on PBx. The median DSC for PS was higher for GL than U-Net (0.91 vs 0.88, P < 0.001). The AUC for csPCa detection of GL was similar to PI-RADS (0.75 vs 0.76, P = 0.8) and U-Net (vs 0.74, P = 0.3). A combination of GL and PI-RADS showed a higher AUC of 0.81 than PI-RADS alone (P = 0.02). Compared with U-Net, the GL had smaller magnitude parameters (1.21× 106 vs 177× 106) and less computational workload (9.8× 109 vs 1027× 109 FLOPs). CONCLUSION: A novel GL model fully automatically detects csPCa on prostate biparametric MRI with comparable performance to PI-RADS and DL. Combined with PI-RADS, GL significantly improves csPCa detection.
引用
- Kaneko M, Yang J, Magoulianitis V, et al. A transparent, lightweight and sustainable Green Learning AI model for prostate cancer detection on MRI. BJU Int. 2026. PMID: 41749494. doi:10.1111/bju.70203
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。