1
|
Klontzas ME, Kalarakis G, Koltsakis E, Papathomas T, Karantanas AH, Tzortzakakis A. Convolutional neural networks for the differentiation between benign and malignant renal tumors with a multicenter international computed tomography dataset. Insights Imaging 2024; 15:26. [PMID: 38270726 PMCID: PMC10811309 DOI: 10.1186/s13244-023-01601-8] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [Grants] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 07/22/2023] [Accepted: 12/17/2023] [Indexed: 01/26/2024] Open
Abstract
OBJECTIVES To use convolutional neural networks (CNNs) for the differentiation between benign and malignant renal tumors using contrast-enhanced CT images of a multi-institutional, multi-vendor, and multicenter CT dataset. METHODS A total of 264 histologically confirmed renal tumors were included, from US and Swedish centers. Images were augmented and divided randomly 70%:30% for algorithm training and testing. Three CNNs (InceptionV3, Inception-ResNetV2, VGG-16) were pretrained with transfer learning and fine-tuned with our dataset to distinguish between malignant and benign tumors. The ensemble consensus decision of the three networks was also recorded. Performance of each network was assessed with receiver operating characteristics (ROC) curves and their area under the curve (AUC-ROC). Saliency maps were created to demonstrate the attention of the highest performing CNN. RESULTS Inception-ResNetV2 achieved the highest AUC of 0.918 (95% CI 0.873-0.963), whereas VGG-16 achieved an AUC of 0.813 (95% CI 0.752-0.874). InceptionV3 and ensemble achieved the same performance with an AUC of 0.894 (95% CI 0.844-0.943). Saliency maps indicated that Inception-ResNetV2 decisions are based on the characteristics of the tumor while in most tumors considering the characteristics of the interface between the tumor and the surrounding renal parenchyma. CONCLUSION Deep learning based on a diverse multicenter international dataset can enable accurate differentiation between benign and malignant renal tumors. CRITICAL RELEVANCE STATEMENT Convolutional neural networks trained on a diverse CT dataset can accurately differentiate between benign and malignant renal tumors. KEY POINTS • Differentiation between benign and malignant tumors based on CT is extremely challenging. • Inception-ResNetV2 trained on a diverse dataset achieved excellent differentiation between tumor types. • Deep learning can be used to distinguish between benign and malignant renal tumors.
Collapse
Affiliation(s)
- Michail E Klontzas
- Department of Medical Imaging, University Hospital of Heraklion, Heraklion, Crete, Greece
- Computational BioMedicine Laboratory, Institute of Computer Science, Foundation for Research and Technology (FORTH), Heraklion, Crete, Greece
- Department of Radiology, School of Medicine, University of Crete, Voutes Campus, Heraklion, Greece
| | - Georgios Kalarakis
- Department of Diagnostic Radiology, Karolinska University Hospital, Stockholm, Sweden
- Division of Radiology, Department for Clinical Science, Intervention and Technology (CLINTEC), Karolinska Institutet, Stockholm, Sweden
| | - Emmanouil Koltsakis
- Department of Diagnostic Radiology, Karolinska University Hospital, Stockholm, Sweden
| | - Thomas Papathomas
- Institute of Metabolism and Systems Research, University of Birmingham, Birmingham, UK
- Department of Clinical Pathology, Vestre Viken Hospital Trust, Drammen, Norway
| | - Apostolos H Karantanas
- Department of Medical Imaging, University Hospital of Heraklion, Heraklion, Crete, Greece
- Computational BioMedicine Laboratory, Institute of Computer Science, Foundation for Research and Technology (FORTH), Heraklion, Crete, Greece
- Department of Radiology, School of Medicine, University of Crete, Voutes Campus, Heraklion, Greece
| | - Antonios Tzortzakakis
- Division of Radiology, Department for Clinical Science, Intervention and Technology (CLINTEC), Karolinska Institutet, Stockholm, Sweden.
- Medical Radiation Physics and Nuclear Medicine, Section for Nuclear Medicine, Karolinska University Hospital, 14 186, Huddinge, Stockholm, Sweden.
| |
Collapse
|
2
|
Basile G, Fallara G, Verri P, Uleri A, Chiti A, Gianolli L, Pepe G, Tedde A, Algaba F, Territo A, Sanguedolce F, Larcher A, Gallioli A, Palou J, Montorsi F, Capitanio U, Breda A. The Role of 99mTc-Sestamibi Single-photon Emission Computed Tomography/Computed Tomography in the Diagnostic Pathway for Renal Masses: A Systematic Review and Meta-analysis. Eur Urol 2024; 85:63-71. [PMID: 37673752 DOI: 10.1016/j.eururo.2023.07.013] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 12/20/2022] [Revised: 06/06/2023] [Accepted: 07/18/2023] [Indexed: 09/08/2023]
Abstract
CONTEXT The diagnostic accuracy of current imaging techniques in differentiating benign from malignant neoplasms in the case of indeterminate renal masses is still suboptimal. OBJECTIVE To evaluate the diagnostic accuracy of 99mTc-sestamibi (SestaMIBI) single-photon emission tomography computed tomography (SPECT)/CT in characterizing indeterminate renal masses by differentiating renal oncocytoma and hybrid oncocytic/chromophobe tumor (HOCT) from (1) all other renal lesions and (2) all malignant renal lesions. Secondary outcomes were: (1) benign versus malignant; (2) renal oncocytoma and HOCT versus clear cell (ccRCC) and papillary (pRCC) renal cell carcinoma; and (3) renal oncocytoma and HOCT versus chromophobe renal cell carcinoma (chRCC). EVIDENCE ACQUISITION A literature search was conducted up to November 2022 using the PubMed/MEDLINE, Embase, and Web of Science databases. Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines were followed to identify eligible studies. Studies included were prospective and retrospective cross-sectional studies in which SestaMIBI SPECT/CT findings were compared to histology after renal mass biopsy or surgery. EVIDENCE SYNTHESIS Overall, eight studies involving 489 patients with 501 renal masses met our inclusion criteria. The sensitivity and specificity of SestaMIBI SPECT/CT for renal oncocytoma and HOCT versus all other renal lesions were 89% (95% confidence interval [CI] 70-97%) and 89% (95% CI 86-92%), respectively. Notably, for renal oncocytoma and HOCT versus ccRCC and pRCC, SestaMIBI SPECT/CT showed specificity of 98% (95% CI 91-100%) and similar sensitivity. Owing to the relatively high risk of bias and the presence of heterogeneity among the studies included, the level of evidence is still low. CONCLUSIONS SestaMIBI SPECT/CT has good sensitivity and specificity in differentiating renal oncocytoma and HOCT from all other renal lesions, and in particular from those with more aggressive oncological behavior. Although these results are promising, further studies are needed to support the use of SestaMIBI SPECT/CT outside research trials. PATIENT SUMMARY A scan method called SestaMIBI SPECT/CT has promise for diagnosing whether kidney tumors are malignant or not. However, it should still be limited to research trials because the level of evidence from our review is low.
Collapse
Affiliation(s)
- Giuseppe Basile
- Department of Urology, Urological Research Institute, San Raffaele Scientific Institute, Milan, Italy; Department of Urology, Fundació Puigvert, Autonoma University of Barcelona, Barcelona, Spain.
| | - Giuseppe Fallara
- Department of Urology, IRCCS European Institute of Oncology, IEO, Milan, Italy
| | - Paolo Verri
- Department of Urology, Fundació Puigvert, Autonoma University of Barcelona, Barcelona, Spain
| | - Alessandro Uleri
- Department of Urology, Fundació Puigvert, Autonoma University of Barcelona, Barcelona, Spain
| | - Arturo Chiti
- Department of Nuclear Medicine, San Raffaele Scientific Institute, Milan, Italy
| | - Luigi Gianolli
- Department of Nuclear Medicine, San Raffaele Scientific Institute, Milan, Italy
| | - Gino Pepe
- Department of Nuclear Medicine, San Raffaele Scientific Institute, Milan, Italy
| | - Alessandro Tedde
- Department of Urology, Fundació Puigvert, Autonoma University of Barcelona, Barcelona, Spain; Department of Medicine, Surgery and Pharmacy, Università degli Studi di Sassari, Sassari, Italy
| | - Ferran Algaba
- Department of Pathology, Fundació Puigvert, Autonoma University of Barcelona, Barcelona, Spain
| | - Angelo Territo
- Department of Urology, Fundació Puigvert, Autonoma University of Barcelona, Barcelona, Spain
| | - Francesco Sanguedolce
- Department of Urology, Fundació Puigvert, Autonoma University of Barcelona, Barcelona, Spain; Department of Medicine, Surgery and Pharmacy, Università degli Studi di Sassari, Sassari, Italy
| | - Alessandro Larcher
- Department of Urology, Urological Research Institute, San Raffaele Scientific Institute, Milan, Italy
| | - Andrea Gallioli
- Department of Urology, Fundació Puigvert, Autonoma University of Barcelona, Barcelona, Spain
| | - Joan Palou
- Department of Urology, Fundació Puigvert, Autonoma University of Barcelona, Barcelona, Spain
| | - Francesco Montorsi
- Department of Urology, Urological Research Institute, San Raffaele Scientific Institute, Milan, Italy
| | - Umberto Capitanio
- Department of Urology, Urological Research Institute, San Raffaele Scientific Institute, Milan, Italy
| | - Alberto Breda
- Department of Urology, Fundació Puigvert, Autonoma University of Barcelona, Barcelona, Spain
| |
Collapse
|
3
|
Klontzas ME, Koltsakis E, Kalarakis G, Trpkov K, Papathomas T, Karantanas AH, Tzortzakakis A. Machine Learning Integrating 99mTc Sestamibi SPECT/CT and Radiomics Data Achieves Optimal Characterization of Renal Oncocytic Tumors. Cancers (Basel) 2023; 15:3553. [PMID: 37509214 PMCID: PMC10377512 DOI: 10.3390/cancers15143553] [Citation(s) in RCA: 3] [Impact Index Per Article: 3.0] [Reference Citation Analysis] [Abstract] [Key Words] [Grants] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 06/20/2023] [Revised: 07/04/2023] [Accepted: 07/06/2023] [Indexed: 07/30/2023] Open
Abstract
The increasing evidence of oncocytic renal tumors positive in 99mTc Sestamibi Single Photon Emission Tomography/Computed Tomography (SPECT/CT) examination calls for the development of diagnostic tools to differentiate these tumors from more aggressive forms. This study combined radiomics analysis with the uptake of 99mTc Sestamibi on SPECT/CT to differentiate benign renal oncocytic neoplasms from renal cell carcinoma. A total of 57 renal tumors were prospectively collected. Histopathological analysis and radiomics data extraction were performed. XGBoost classifiers were trained using the radiomics features alone and combined with the results from the visual evaluation of 99mTc Sestamibi SPECT/CT examination. The combined SPECT/radiomics model achieved higher accuracy (95%) with an area under the curve (AUC) of 98.3% (95% CI 93.7-100%) than the radiomics-only model (71.67%) with an AUC of 75% (95% CI 49.7-100%) and visual evaluation of 99mTc Sestamibi SPECT/CT alone (90.8%) with an AUC of 90.8% (95%CI 82.5-99.1%). The positive predictive values of SPECT/radiomics, radiomics-only, and 99mTc Sestamibi SPECT/CT-only models were 100%, 85.71%, and 85%, respectively, whereas the negative predictive values were 85.71%, 55.56%, and 94.6%, respectively. Feature importance analysis revealed that 99mTc Sestamibi uptake was the most influential attribute in the combined model. This study highlights the potential of combining radiomics analysis with 99mTc Sestamibi SPECT/CT to improve the preoperative characterization of benign renal oncocytic neoplasms. The proposed SPECT/radiomics classifier outperformed the visual evaluation of 99mTc Sestamibii SPECT/CT and the radiomics-only model, demonstrating that the integration of 99mTc Sestamibi SPECT/CT and radiomics data provides improved diagnostic performance, with minimal false positive and false negative results.
Collapse
Affiliation(s)
- Michail E Klontzas
- Department of Medical Imaging, University Hospital of Heraklion, Heraklion 71110, Greece
- Computational BioMedicine Laboratory, Institute of Computer Science, Foundation for Research and Technology (FORTH), Heraklion 70013, Greece
- Department of Radiology, School of Medicine, University of Crete, Voutes Campus, Heraklion 71110, Greece
| | - Emmanouil Koltsakis
- Department of Diagnostic Radiology, Karolinska University Hospital, Stockholm 17177, Sweden
| | - Georgios Kalarakis
- Department of Diagnostic Radiology, Karolinska University Hospital, Stockholm 17177, Sweden
- Division of Radiology, Department for Clinical Science, Intervention and Technology (CLINTEC), Karolinska Institutet, Stockholm 14152, Sweden
| | - Kiril Trpkov
- Alberta Precision Labs, Department of Pathology and Laboratory Medicine, Cumming School of Medicine, University of Calgary, Calgary, AB T2L 2K5, Canada
| | - Thomas Papathomas
- Institute of Metabolism and Systems Research, University of Birmingham, Birmingham B15 2TT, UK
- Department of Clinical Pathology, Vestre Viken Hospital Trust, Drammen 3004, Norway
| | - Apostolos H Karantanas
- Department of Medical Imaging, University Hospital of Heraklion, Heraklion 71110, Greece
- Computational BioMedicine Laboratory, Institute of Computer Science, Foundation for Research and Technology (FORTH), Heraklion 70013, Greece
- Department of Radiology, School of Medicine, University of Crete, Voutes Campus, Heraklion 71110, Greece
| | - Antonios Tzortzakakis
- Division of Radiology, Department for Clinical Science, Intervention and Technology (CLINTEC), Karolinska Institutet, Stockholm 14152, Sweden
- Medical Radiation Physics and Nuclear Medicine, Section for Nuclear Medicine, Karolinska University Hospital, Huddinge, Stockholm 14186, Sweden
| |
Collapse
|
4
|
Qu J, Zhang Q, Song X, Jiang H, Ma H, Li W, Wang X. CT differentiation of the oncocytoma and renal cell carcinoma based on peripheral tumor parenchyma and central hypodense area characterisation. BMC Med Imaging 2023; 23:16. [PMID: 36707788 PMCID: PMC9881251 DOI: 10.1186/s12880-023-00972-0] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 09/21/2022] [Accepted: 01/18/2023] [Indexed: 01/28/2023] Open
Abstract
BACKGROUND Although the central scar is an essential imaging characteristic of renal oncocytoma (RO), its utility in distinguishing RO from renal cell carcinoma (RCC) has not been well explored. The study aimed to evaluate whether the combination of CT characteristics of the peripheral tumor parenchyma (PTP) and central hypodense area (CHA) can differentiate typical RO with CHA from RCC. METHODS A total of 132 tumors on the initial dataset were retrospectively evaluated using four-phase CT. The excretory phases were performed more than 20 min after the contrast injection. In corticomedullary phase (CMP) images, all tumors had CHAs. These tumors were categorized into RO (n = 23), clear cell RCC (ccRCC) (n = 85), and non-ccRCC (n = 24) groups. The differences in these qualitative and quantitative CT features of CHA and PTP between ROs and ccRCCs/non-ccRCCs were statistically examined. Logistic regression filters the main factors for separating ROs from ccRCCs/non-ccRCCs. The prediction models omitting and incorporating CHA features were constructed and evaluated, respectively. The effectiveness of the prediction models including CHA characteristics was then confirmed through a validation dataset (8 ROs, 35 ccRCCs, and 10 non-ccRCCs). RESULTS The findings indicate that for differentiating ROs from ccRCCs and non-ccRCCs, prediction models with CHA characteristics surpassed models without CHA, with the corresponding areas under the curve (AUC) being 0.962 and 0.914 versus 0.952 and 0.839 respectively. In the prediction models that included CHA parameters, the relative enhancement ratio (RER) in CMP and enhancement inversion, as well as RER in nephrographic phase and enhancement inversion were the primary drivers for differentiating ROs from ccRCCs and non-ccRCCs, respectively. The prediction models with CHA characteristics had the comparable diagnostic ability on the validation dataset, with respective AUC values of 0.936 and 0.938 for differentiating ROs from ccRCCs and non-ccRCCs. CONCLUSION The prediction models with CHA characteristics can help better differentiate typical ROs from RCCs. When a mass with CHA is discovered, particularly if RO is suspected, EP images with longer delay scanning periods should be acquired to evaluate the enhancement inversion characteristics of CHA.
Collapse
Affiliation(s)
- Jianyi Qu
- grid.410645.20000 0001 0455 0905Yuhuangding Hospital, Qingdao University School of Medicine, Shandong Yantai, China
| | - Qianqian Zhang
- grid.410645.20000 0001 0455 0905Yuhuangding Hospital, Qingdao University School of Medicine, Shandong Yantai, China
| | - Xinhong Song
- grid.410645.20000 0001 0455 0905Yuhuangding Hospital, Qingdao University School of Medicine, Shandong Yantai, China
| | - Hong Jiang
- grid.410645.20000 0001 0455 0905Yuhuangding Hospital, Qingdao University School of Medicine, Shandong Yantai, China
| | - Heng Ma
- grid.410645.20000 0001 0455 0905Yuhuangding Hospital, Qingdao University School of Medicine, Shandong Yantai, China
| | - Wenhua Li
- grid.16821.3c0000 0004 0368 8293Xinhua Hospital, Shanghai Jiaotong University School of Medicine, Shanghai, China
| | - Xiaofei Wang
- grid.440653.00000 0000 9588 091XYantaishan Hospital, Binzhou Medical University, Shandong Yantai, China
| |
Collapse
|