文献速递泌尿生殖腹部MR

AI计算机辅助诊断在前列腺MRI癌症分类中的附加价值

The added value of AI-based computer-aided diagnosis in classification of cancer at prostate MRI.
PubMed 文献 · Eur Radiol · 2023-02-01

本研究为多中心回顾性研究,开发基于卷积神经网络的前列腺分割与前列腺癌检测AI模型,并在内部和外部测试集中比较AI-CAD(并发阅读与AI优先阅读)与常规PI-RADS评估的诊断性能、一致性和效率。结果显示AI-CAD方法与放射科医师诊断准确性相当,并发阅读在外部测试中优于AI优先阅读且不劣于/优于放射科医师,AI方法显著提高一致性并缩短阅读时间,并发阅读可提升经验不足医师在不熟悉场景下的诊断能力。

要点

  • AI-CAD诊断准确性与放射科医师相当,并发阅读在外部测试中表现更优。
  • AI方法(AI优先/并发阅读)较放射科医师显著提高诊断一致性(κ更高)。
  • AI优先阅读速度远快于医师阅读,并发阅读居中,可提升效率。

核心结论

  • 基于AI的CAD可提高准确诊断的一致性和效率;并发阅读方法可增强经验不足放射科医师在不熟悉场景下的诊断能力。

研究设计与样本

  • 多中心回顾性研究,纳入接受前列腺穿刺活检和多参数MRI的患者。
  • 训练/验证/内部测试集来自一个中心共650例(400/100/150),外部测试集来自三个中心共100例(25/25/50)。
  • 训练并验证基于卷积神经网络的AI模型,在内部/外部队列测试不同CAD方法(并发阅读和AI优先阅读)的可靠性,比较放射科医师与AI-CAD的诊断性能、一致性和效率。

关键影像表现/征象

  • 前列腺癌分割方面,多小Vnet性能优于小Vnet和Vnet(DSCmsvnet vs. DSCsvnet, p=0.021;DSCmsvnet vs. DSCvnet, p<0.001)。
  • 前列腺腺体分割方面,精细分割和粗糙分割的平均/中位DSC分别为0.91/0.91和0.88/0.89,精细分割优于粗糙分割(p<0.001)。

诊断标准与关键数值

  • 内部测试诊断准确性:AI-CAD方法与放射科医师无显著差异且等效(读者1:127/150 vs. 130/150 vs. 125/150;读者2:127/150 vs. 132/150 vs. 131/150;所有p>0.05)。
  • 外部测试:并发阅读优于/等于AI优先阅读(读者2:87/100 vs. 71/100,p<0.001;读者1:79/100 vs. 69/100,p=0.076),且优于/等于放射科医师(读者1:79/100 vs. 72/100,p=0.039;读者2:87/100 vs. 86/100,p=1.000)。
  • 一致性:AI优先阅读/并发阅读在内部测试(κ=1.000, 0.830)和外部测试(κ=0.958, 0.713)中较放射科医师(κ=0.747, 0.600)提高一致性。
  • 效率:AI优先阅读(8.54 s/7.66 s)快于放射科医师(92.72 s/89.54 s)和并发阅读(29.15 s/28.92 s)。

临床应用与阅片要点

  • 在日常阅片中可采用并发阅读模式,即医师参考AI结果进行判断,以提高诊断一致性并辅助经验不足医师。
  • AI优先阅读可大幅缩短阅读时间,适合高效率筛查场景,但外部测试中其准确性可能低于并发阅读。

局限性

  • 回顾性研究设计,可能存在选择偏倚。
  • 外部测试样本量较小(100例),需更大规模验证。

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)。

英文原文(摘要)· 点击展开

OBJECTIVES: To develop an artificial intelligence (AI) model for prostate segmentation and prostate cancer (PCa) detection, and explore the added value of AI-based computer-aided diagnosis (CAD) compared to conventional PI-RADS assessment.

METHODS: A retrospective study was performed on multi-centers and included patients who underwent prostate biopsies and multiparametric MRI. A convolutional-neural-network-based AI model was trained and validated; the reliability of different CAD methods (concurrent read and AI-first read) were tested in an internal/external cohort. The diagnostic performance, consistency and efficiency of radiologists and AI-based CAD were compared.

RESULTS: The training/validation/internal test sets included 650 (400/100/150) cases from one center; the external test included 100 cases (25/25/50) from three centers. For diagnosis accuracy, AI-based CAD methods showed no significant differences and were equivalent to the radiologists in the internal test (127/150 vs. 130/150 vs. 125/150 for reader 1; 127/150 vs.132/150 vs. 131/150 for reader 2; all p > 0.05), whereas in the external test, concurrent-read methods were superior/equal to AI-first read (87/100 vs. 71/100, p < 0.001, for reader 2; 79/100 vs. 69/100, p = 0.076, for reader 1) and better than/equal to radiologists (79/100 vs. 72/100, p = 0.039, for reader 1; 87/100 vs. 86/100, p = 1.000, for reader 2). Moreover, AI-first read/concurrent read improved consistency in both internal test (κ = 1.000, 0.830) and external test (κ = 0.958, 0.713) compared to radiologists (κ = 0.747, 0.600); AI-first read method (8.54 s/7.66 s) was faster than readers (92.72 s/89.54 s) and concurrent-read method (29.15 s/28.92 s), respectively.

CONCLUSION: AI-based CAD could improve the consistency and efficiency for accurate diagnosis; the concurrent-read method could enhance the diagnostic capabilities of an inexperienced radiologist in unfamiliar situations.

KEY POINTS: • For prostate cancer segmentation, the performance of multi-small Vnet displays optimal compared to small Vnet and Vnet (DSCmsvnet vs. DSCsvnet, p = 0.021; DSCmsvnet vs. DSCvnet, p < 0.001). • For prostate gland segmentation, the mean/median DSCs for fine and coarse segmentation were 0.91/0.91 and 0.88/0.89, respectively. Fine segmentation displays superior performance compared to coarse (DSCcoarse vs. DSCfine, p < 0.001). • For PCa diagnosis, AI-based CAD methods improve consistency in internal (κ = 1.000; 0.830) and external (κ = 0.958; 0.713) tests compared to radiologists (κ = 0.747; 0.600); the AI-first read (8.54 s/7.66 s) was faster than the readers (92.72 s/89.54 s) and the concurrent-read method (29.15 s/28.92 s).

来源:Eur Radiol,查看原文。版权归原作者所有。

引用

  1. Liu, Pan, Zhao, et al. The added value of AI-based computer-aided diagnosis in classification of cancer at prostate MRI. Eur Radiol. 2023. PMID: 36725719. doi:10.1007/s00330-023-09433-2

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。