de Roode Lotte M, de Boer Lisanne L, Da Silva Guimaraes Marcos, van Leeuwen Pim J, van der Poel Henk G, Dashtbozorg Behdad, Ruers Theo J M
Department of Nanobiophysics, University of Twente, Enschede, The Netherlands.
Image-Guided Surgery, Department of Surgery, Netherlands Cancer Institute-Antoni van Leeuwenhoek Hospital, Amsterdam, The Netherlands.
Eur Urol Open Sci. 2024 Aug 10;67:62-68. doi: 10.1016/j.euros.2024.07.112. eCollection 2024 Sep.
A positive surgical margin (PSM) occurs in up to 32% of patients undergoing robot-assisted radical prostatectomy (RARP). Diffuse reflectance spectroscopy (DRS), which measures tissue composition according to its optical properties, can potentially be used for real-time PSM detection during RARP. Our objective was to assess the feasibility of DRS in distinguishing prostate cancer from benign tissue in RARP specimens.
In a single-center prospective study, DRS measurements were taken ex vivo for RARP specimens from 59 patients with biopsy-proven prostate carcinoma. Discriminating features from the DRS spectra were used to create a machine learning-based classification algorithm. The data were split patient-wise into training (70%) and testing (30%) sets, with ten iterations to ensure algorithm robustness. The average sensitivity, specificity, accuracy, and area under the receiver operating characteristic curve (AUC) from ten classification iterations were calculated.
We collected 542 DRS measurements, of which 53% were tumor and 47% were healthy-tissue measurements. Twenty discriminating features from the DRS spectra were used as the input for a support vector machine model. This model achieved average sensitivity of 89%, specificity of 82%, accuracy of 85%, and AUC of 0.91 for the test set. Limitations include the binary label input for classification.
DRS can potentially discriminate prostate cancer from benign tissue. Before implementing the technique in clinical practice, further research is needed to assess its performance on heterogeneous tissue volumes and measurements from the prostate surface.
We looked at the ability of a technique called diffuse reflectance spectroscopy to guide surgeons in discriminating prostate cancer tissue from benign prostate tissue in real time during prostate cancer surgery. Our study showed promising results in an experimental setting. Future research will focus on bringing this technique to clinical practice.
在接受机器人辅助根治性前列腺切除术(RARP)的患者中,高达32%会出现手术切缘阳性(PSM)。漫反射光谱(DRS)可根据组织的光学特性测量其成分,有可能用于RARP术中实时PSM检测。我们的目的是评估DRS在区分RARP标本中前列腺癌与良性组织方面的可行性。
在一项单中心前瞻性研究中,对59例经活检证实为前列腺癌患者的RARP标本进行离体DRS测量。利用DRS光谱的鉴别特征创建基于机器学习的分类算法。数据按患者分为训练集(70%)和测试集(30%),进行十次迭代以确保算法的稳健性。计算十次分类迭代的平均灵敏度、特异性、准确性和受试者操作特征曲线下面积(AUC)。
我们收集了542次DRS测量数据,其中53%为肿瘤组织测量,47%为健康组织测量。从DRS光谱中提取的20个鉴别特征用作支持向量机模型的输入。该模型对测试集的平均灵敏度为89%,特异性为82%,准确性为85%,AUC为0.91。局限性包括分类的二元标签输入。
DRS有可能区分前列腺癌与良性组织。在将该技术应用于临床实践之前,需要进一步研究以评估其在异质组织体积和前列腺表面测量中的性能。
我们研究了一种名为漫反射光谱的技术在前列腺癌手术中实时指导外科医生区分前列腺癌组织与良性前列腺组织的能力。我们的研究在实验环境中显示出了有前景的结果。未来的研究将集中于将该技术应用于临床实践。