Federated Learning in Renal Tumor Model
Research on the Application of Federated Learning in the Construction of Deep Learning Models for Renal Tumor Imaging
This is a multicenter, retrospective and prospective diagnostic clinical trial evaluating the effectiveness and safety of CascadeDiagnose Renal Tumor CT, an AI-assisted detection system for renal tumors based on contrast-enhanced multiphase CT imaging. The system employs a cascaded deep learning architecture to perform fully automated analysis of multiphase CT images, covering image quality review, lesion detection and segmentation, benign-malignant differentiation, and risk stratification, with traceable evidence chains and interpretable outputs. The study is conducted across six tertiary hospitals in Guangxi, China, utilizing a distributed "data stays on-site, computation moves across centers" federated learning network, which ensures that original patient data remain within each hospital while encrypted intermediate features are shared for cross-center collaborative analysis. A total of at least 3000 patients with renal tumors will be enrolled (approximately 2600 in the retrospective phase and 400 in the prospective phase). The primary effectiveness outcomes include area under the receiver operating characteristic curve (AUC), sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV). The primary safety outcomes include false negative rate, false positive rate, and adverse events. The study also incorporates a multi-reader, multi-case (MRMC) design to compare the diagnostic performance of the AI system with radiologists of varying seniority, and to evaluate the system's utility in assisting junior radiologists. The findings of this study are expected to provide high-quality clinical evidence for the regulatory approval of this AI-assisted diagnostic system as a Class III medical device.
調査の概要
状態
条件
研究の種類
入学 (推定)
連絡先と場所
研究連絡先
- 名前：Hai-qi Liang, MD
- 電話番号：86+18176618137
- メール：liahiki@163.com
研究連絡先のバックアップ
- 名前：Nai-kai Liao, MD
- 電話番号：86+18776891703
- メール：2546419718@qq.com
研究場所
-
-
Guangxi
-
Guigang、Guangxi、中国、537100
- Guigang People's Hospital
-
コンタクト：
- Jiale Feng, MD
- 電話番号：86+17695684196
- メール：15498419548@qq.com
-
Guilin、Guangxi、中国、541002
- Guilin People's Hospital
-
コンタクト：
- Ke Qin, MD
- 電話番号：86+15954268457
- メール：QINKE541002@163.com
-
Liuzhou、Guangxi、中国、545006
- Liuzhou People's Hospital
-
コンタクト：
- Jianwen Yang, MD
- 電話番号：86+13675215625
- メール：JWyang235@163.com
-
Nanning、Guangxi、中国、530021
- The First Affiliated Hospital of Guangxi Medical University
-
コンタクト：
- Ji-wen Cheng, PhD
- 電話番号：86+18776891703
- メール：chengjiwen@stu.gxmu.edu.cn
-
コンタクト：
- Haiqi Liang, MD
- 電話番号：86+18176618137
- メール：liahiki@163.com
-
副調査官：
- Shenghua Deng, MD
-
Nanning、Guangxi、中国、530021
- Nanning Second People's Hospital
-
コンタクト：
- Yang Xie, MD
- 電話番号：86+18346826751
- メール：XieYNan@163.com
-
Wuzhou、Guangxi、中国、543002
- Wuzhou Red Cross Hospital
-
コンタクト：
- Deyan Ou, MD
- 電話番号：86+18675651254
- メール：OudYan1254@163.com
-
-
参加基準
適格基準
就学可能な年齢
健康ボランティアの受け入れ
受講資格のある性別
サンプリング方法
調査対象母集団
The study population consists of approximately 3,000 patients with pathologically confirmed renal tumors (benign or malignant) who underwent contrast-enhanced CT at ten Guangxi hospitals between January 2019 and June 2026. Cases are consecutively enrolled using a unified time window and strict inclusion/exclusion criteria. Eligible patients must have complete pre-operative contrast-enhanced CT, good image quality, and traceable clinical-pathological data. Exclusions include non-contrast CT only, severe artifacts, prior nephrectomy (except recurrence/residual), and incomplete data. Enrollment targets are 600 cases at the lead site and 300 at each of the nine participating centers.
説明
Inclusion Criteria:
- Patients pathologically confirmed as having renal tumors (either benign or malignant) by surgery or biopsy;
- Underwent contrastenhanced renal CT before surgery or treatment;
- CT images are of good quality and clearly depict the renal lesion contour;
- Complete and traceable clinical and pathological data.
Exclusion Criteria:
- Patients who did not undergo contrastenhanced CT before surgery, or only had noncontrast CT;
- CT images with significant motion artifacts, metal artifacts, or excessive noise that impair lesion assessment;
- Lesions too small (maximum diameter <1 mm) or not identifiable on imaging;
- Patients who previously underwent partial or radical nephrectomy for renal tumors (except for recurrent/residual lesions);
- Incomplete clinical data or pathological results.
研究計画
研究はどのように設計されていますか？
デザインの詳細
- 観測モデル：CASE_CONTROL
- 時間の展望：OTHER
コホートと介入
グループ/コホート |
|---|
|
Malignant renal tumor
|
|
Benign renal tumor
|
この研究は何を測定していますか？
主要な結果の測定
結果測定 |
メジャーの説明 |
時間枠 |
|---|---|---|
|
Dice Similarity Coefficient for Renal Tumor Segmentation
時間枠：The primary outcome will be measured in October 2027, upon completion of model training, final parameter aggregation, and ensemble model establishment, using the test set from the multi-center retrospective cohort.
|
The primary outcome measure is the Dice Similarity Coefficient for renal tumor segmentation on contrast-enhanced CT, comparing the federated learning model against a centralized training model with a non-inferiority margin of Δ = -0.05. This outcome will be assessed at the end of the model training phase using an independent multi-center test set. |
The primary outcome will be measured in October 2027, upon completion of model training, final parameter aggregation, and ensemble model establishment, using the test set from the multi-center retrospective cohort.
|
二次結果の測定
結果測定 |
メジャーの説明 |
時間枠 |
|---|---|---|
|
Classification Performance Metrics and Heterogeneity Impact Assessment
時間枠：Assessed at the completion of the model training phase, approximately 18 months after study initiation, following final aggregation and ensemble construction, using the independent multi-center test set, with center-specific and subtype-specific analyses
|
The secondary outcomes are: (1) classification performance metrics-AUC, accuracy, sensitivity, and specificity-for benign versus malignant renal tumor discrimination, with pathological diagnosis as the reference; and (2) the impact of data heterogeneity on model performance, assessed by comparing Dice coefficients and AUC values across centers and pathological subtypes. All secondary outcomes will be evaluated at the completion of the model training phase, approximately 18 months after study initiation, after the final federated aggregation and ensemble model construction, using the independent multi-center test set, with center-specific and subtype-specific analyses performed thereafter. |
Assessed at the completion of the model training phase, approximately 18 months after study initiation, following final aggregation and ensemble construction, using the independent multi-center test set, with center-specific and subtype-specific analyses
|
協力者と研究者
捜査官
- スタディディレクター：Min Qin, MD、First Affiliated Hospital of Guangxi Medical University
出版物と役立つリンク
一般刊行物
- Sung H, Ferlay J, Siegel RL, Laversanne M, Soerjomataram I, Jemal A, Bray F. Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. CA Cancer J Clin. 2021 May;71(3):209-249. doi: 10.3322/caac.21660. Epub 2021 Feb 4.
- Bex A, Ghanem YA, Albiges L, Bonn S, Campi R, Capitanio U, Dabestani S, Hora M, Klatte T, Kuusk T, Lund L, Marconi L, Palumbo C, Pignot G, Powles T, Schouten N, Tran M, Volpe A, Bedke J. European Association of Urology Guidelines on Renal Cell Carcinoma: The 2025 Update. Eur Urol. 2025 Jun;87(6):683-696. doi: 10.1016/j.eururo.2025.02.020. Epub 2025 Mar 20.
- Xi IL, Zhao Y, Wang R, Chang M, Purkayastha S, Chang K, Huang RY, Silva AC, Vallieres M, Habibollahi P, Fan Y, Zou B, Gade TP, Zhang PJ, Soulen MC, Zhang Z, Bai HX, Stavropoulos SW. Deep Learning to Distinguish Benign from Malignant Renal Lesions Based on Routine MR Imaging. Clin Cancer Res. 2020 Apr 15;26(8):1944-1952. doi: 10.1158/1078-0432.CCR-19-0374. Epub 2020 Jan 14.
研究記録日
主要日程の研究
研究開始 (推定)
一次修了 (推定)
研究の完了 (推定)
試験登録日
最初に提出
QC基準を満たした最初の提出物
最初の投稿 (実際)
学習記録の更新
投稿された最後の更新 (実際)
QC基準を満たした最後の更新が送信されました
最終確認日
詳しくは
本研究に関する用語
追加の関連 MeSH 用語
その他の研究ID番号
- 2026-S0835
個々の参加者データ (IPD) の計画
個々の参加者データ (IPD) を共有する予定はありますか?
IPD プランの説明
Individual Participant Data (IPD) will not be shared in this study. The primary reasons are threefold: (1) strict legal restrictions under China's Data Security Law and Personal Information Protection Law prohibit cross-institutional transfer of raw medical data without explicit consent; (2) as a retrospective study spanning several years, obtaining renewed consent from all participants is practically infeasible; and (3) the core federated learning framework inherently eliminates the need for IPD sharing-only encrypted model parameters are transmitted, while raw CT images and clinical data remain securely stored at each participating center. This design fully protects patient privacy, complies with all regulatory requirements, and ensures the study's collaborative objectives are achieved without ever exposing individual-level data.
医薬品およびデバイス情報、研究文書
米国FDA規制医薬品の研究
米国FDA規制機器製品の研究
この情報は、Web サイト clinicaltrials.gov から変更なしで直接取得したものです。研究の詳細を変更、削除、または更新するリクエストがある場合は、register@clinicaltrials.gov。 までご連絡ください。 clinicaltrials.gov に変更が加えられるとすぐに、ウェブサイトでも自動的に更新されます。