采用自动化影像平台识别、测量和时序追踪病灶,按更新版PROMISE标准评估转移性前列腺癌患者的治疗反应
研究纳入33例去势敏感型前列腺癌(CSPC)和34例去势抵抗型前列腺癌(CRPC)患者,在治疗前和治疗完成≥2个月后行PSMA靶向分子影像,用aPROMISE平台识别病灶并计算总肿瘤体积,按RECIP 1.0标准评估治疗反应。aPROMISE与临床医生验证的一致率为89.6%(κ=0.79),将4例判为完全缓解、16例部分缓解、34例稳定、13例进展,5例分类错误。
要点
- aPROMISE在基线扫描识别1576个病灶,随访影像识别1631个病灶,其中618个(35%)为新发。
- 67例中aPROMISE判为完全缓解4例、部分缓解16例、稳定34例、进展13例,5例分类错误。
- aPROMISE与临床医生验证的一致率为89.6%(κ=0.79)。
核心结论
- aPROMISE可能作为一种新的治疗反应评估工具,整合PSMA PET/CT结果与RECIP影像标准。
- 该自动化流程的精确性和准确性应在前瞻性临床研究中验证。
研究设计与样本
- 研究纳入33例去势敏感型前列腺癌(CSPC)和34例去势抵抗型前列腺癌(CRPC)患者,在治疗前和治疗完成≥2个月后行PSMA靶向分子影像。
- 使用aPROMISE识别治疗前和治疗后PET/CT扫描中的示踪剂摄取病灶,由经验丰富的核医学医师手动确认,并计算总肿瘤体积(TTV)。
- 按RECIP 1.0标准将治疗反应评估为完全缓解(CR)、部分缓解(PR)、进展(PD)或稳定(SD)。
主要发现
- aPROMISE在基线扫描识别1576个病灶,在随访影像识别1631个病灶,其中618个(35%)为新发。
- 67例患者中,aPROMISE将4例判为完全缓解、16例判为部分缓解、34例判为稳定、13例判为进展;5例分类错误。
- aPROMISE与临床医生验证的一致率为89.6%(κ=0.79)。
局限性
- 原文在结论中提出该自动化流程的精确性和准确性应在前瞻性临床研究中验证。
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)。
原文
BACKGROUND AND OBJECTIVE: Prostate-specific membrane antigen (PSMA) molecular imaging is widely used for disease assessment in prostate cancer (PC). Artificial intelligence (AI) platforms such as automated Prostate Cancer Molecular Imaging Standardized Evaluation (aPROMISE) identify and quantify locoregional and distant disease, thereby expediting lesion identification and standardizing reporting. Our aim was to evaluate the ability of the updated aPROMISE platform to assess treatment responses based on integration of the RECIP (Response Evaluation Criteria in PSMA positron emission tomography-computed tomography [PET/CT]) 1.0 classification. METHODS: The study included 33 patients with castration-sensitive PC (CSPC) and 34 with castration-resistant PC (CRPC) who underwent PSMA-targeted molecular imaging before and ≥2 mo after completion of treatment. Tracer-avid lesions were identified using aPROMISE for pretreatment and post-treatment PET/CT scans. Detected lesions were manually approved by an experienced nuclear medicine physician, and total tumor volume (TTV) was calculated. Response was assessed according to RECIP 1.0 as CR (complete response), PR (partial response), PD (progressive disease), or SD (stable disease). KEY FINDINGS AND LIMITATIONS: aPROMISE identified 1576 lesions on baseline scans and 1631 lesions on follow-up imaging, 618 (35%) of which were new. Of the 67 patients, aPROMISE classified four as CR, 16 as PR, 34 as SD, and 13 as PD; five cases were misclassified. The agreement between aPROMISE and clinician validation was 89.6% (κ = 0.79). CONCLUSIONS AND CLINICAL IMPLICATIONS: aPROMISE may serve as a novel assessment tool for treatment response that integrates PSMA PET/CT results and RECIP imaging criteria. The precision and accuracy of this automated process should be validated in prospective clinical studies. PATIENT SUMMARY: We used an artificial intelligence (AI) tool to analyze scans for prostate cancer before and after treatment to see if we could track how cancer spots respond to treatment. We found that the AI approach was successful in tracking individual tumor changes, showing which tumors disappeared, and identifying new tumors in response to prostate cancer treatment.
引用
- Benitez CM, Sahlstedt H, Sonni I, et al. Treatment Response Assessment According to Updated PROMISE Criteria in Patients with Metastatic Prostate Cancer Using an Automated Imaging Platform for Identification, Measurement, and Temporal Tracking of Disease. Eur Urol Oncol. 2024. PMID: 39521638. doi:10.1016/j.euo.2024.10.011
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。