Skip to main content

Open Access 23.02.2024 | Imaging Informatics and Artificial Intelligence

Artificial intelligence support in MR imaging of incidental renal masses: an early health technology assessment

verfasst von: Alexander W. Marka, Johanna Luitjens, Florian T. Gassert, Lisa Steinhelfer, Egon Burian, Johannes Rübenthaler, Vincent Schwarze, Matthias F. Froelich, Marcus R. Makowski, Felix G. Gassert

Erschienen in: European Radiology

Abstract

Objective

This study analyzes the potential cost-effectiveness of integrating an artificial intelligence (AI)–assisted system into the differentiation of incidental renal lesions as benign or malignant on MR images during follow-up.

Materials and methods

For estimation of quality-adjusted life years (QALYs) and lifetime costs, a decision model was created, including the MRI strategy and MRI + AI strategy. Model input parameters were derived from recent literature. Willingness to pay (WTP) was set to $100,000/QALY. Costs of $0 for the AI were assumed in the base-case scenario. Model uncertainty and costs of the AI system were assessed using deterministic and probabilistic sensitivity analysis.

Results

Average total costs were at $8054 for the MRI strategy and $7939 for additional use of an AI-based algorithm. The model yielded a cumulative effectiveness of 8.76 QALYs for the MRI strategy and of 8.77 for the MRI + AI strategy. The economically dominant strategy was MRI + AI. Deterministic and probabilistic sensitivity analysis showed high robustness of the model with the incremental cost-effectiveness ratio (ICER), which represents the incremental cost associated with one additional QALY gained, remaining below the WTP for variation of the input parameters. If increasing costs for the algorithm, the ICER of $0/QALY was exceeded at $115, and the defined WTP was exceeded at $667 for the use of the AI.

Conclusions

This analysis, rooted in assumptions, suggests that the additional use of an AI-based algorithm may be a potentially cost-effective alternative in the differentiation of incidental renal lesions using MRI and needs to be confirmed in the future.

Clinical relevance statement

These results hint at AI’s the potential impact on diagnosing renal masses. While the current study urges careful interpretation, ongoing research is essential to confirm and seamlessly integrate AI into clinical practice, ensuring its efficacy in routine diagnostics.

Key Points

This is a model-based study using data from literature where AI has been applied in the diagnostic workup of incidental renal lesions.
MRI + AI has the potential to be a cost-effective alternative in the differentiation of incidental renal lesions.
The additional use of AI can reduce costs in the diagnostic workup of incidental renal lesions.
Hinweise

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Abkürzungen
AI
Artificial Intelligence
CEA
Cost-effectiveness analysis
CEUS
Contrast-enhanced ultrasound
CNN
Convolutional neural networks
CT
Computed tomography
HEE
Health economic evaluation
ICER
Incremental cost-effectiveness ratio
MRI
Magnetic resonance imaging
QALYs
Quality-adjusted life years
QoL
Quality of life
US
United States
WTP
Willingness to pay

Introduction

The number of computed tomography (CT) and ultrasound examinations has nearly doubled within the last 10 years [1]. In about 13–27% of individuals undergoing abdominal imaging for other purposes, incidental renal lesions are found, and it is thought that more than 50% of patients over the age of 50 have a renal lesion [2]. With an increasing number of imaging procedures expected in the future and thus a higher number of incidental renal lesions, accurate and cost-effective differentiation between benign and malignant lesions is crucial.
Diagnostic pathways after detection of incidental renal lesions depend on the presence of cystic and solid components. The malignancy of cystic renal masses can be graded using the Bosniak Classification after contrast-enhanced magnetic resonance imaging (MRI) or CT [3]. Solid renal masses are most commonly further characterized using contrast-enhanced MRI, but ultrasound has also been proposed as accurate diagnostic means [4]. For histopathological proof of malignancy, suspicious lesions seen in imaging are either biopsied or directly surgically removed. Total nephrectomy can be considered for malignant solid renal masses bigger than 7 cm, and organ-preserving partial nephrectomy can be performed for minor lesions [5]. Follow-up is recommended, due to the risk of local recurrence, especially in the first year after treatment [6].
Convolutional neural networks (CNNs) have shown positive results in detection and differentiation of benign and malignant tumors, and are on par or even outperform sub-specialized experts in diagnostic performance [7, 8]. Xi et al developed a CNN and showed that it can effectively distinguish between benign and malignant renal lesions when using routine MR imaging with high accuracy, sensitivity, and specificity [8].
Although artificial intelligence (AI) has yielded promising outcomes, the implementation in clinical routine remains challenging. High investment costs are necessary for installation of necessary hardware and software [9]. The demonstration of both patient benefit and economic advantages could accelerate the translation into routine clinical practice. Ziegelmayer et al and van Leeuwen et al conducted cost-effectiveness analyses on AI support for lung cancer screening and early vessel occlusion detection, respectively [10, 11]. In addition to this, two literature reviews on health economic evaluations (HEEs) for AI-based health interventions underscore the increasing acceptance of AI in healthcare [12, 13]. Both reviews highlight the potential for enhanced outcomes and signify the escalating significance of AI in the field of radiology. So far, no study has been conducted comparing the use of neural networks in identifying malignant renal lesions to stand-alone MR imaging from an economic point of view. The objective of our analysis was to explore the possible cost-effectiveness of an AI-based system in distinguishing incidentally found renal lesions as benign or malignant. This study, while not a definitive technology assessment, provides insights into approaching these considerations based on a single publication. Additionally, it aims to define a cost margin for potential clinical integration from the healthcare sector perspective.

Materials and methods

Model structure

Based on the clinical scenario of an incidentally found renal mass, an economic model was created, including an MRI strategy and a MRI + AI strategy as diagnostic pathways, concluding in the different clinical scenarios (timely treatment, delayed treatment, no treatment, unnecessary diagnostics) which were simulated through a Markov model. The economic model is shown in Fig. 1.
The Markov model was designed using a specific decision-analytic software (TreeAge Pro Version 19.1.1, Williamstown, MA, USA). Based on the economic model and the respective accuracy of the diagnostic pathways, the starting point for the initial iteration was defined for every patient.
The Markov model was used for running iterations of possible disease outcomes. During each iterations, patients in each state are attributed certain costs and a certain quality of life (QoL) as well as transition probabilities to other states during the next iteration.
The Markov model dedicated to renal carcinoma included the following states and is displayed in Fig. 2:
  • Benign, correctly identified (patients without malignant renal mass = true negative)
  • Benign, identified as malignant (patients without malignant renal mass, but suspicious = false positive)
  • Malignant lesion, identified as benign (patients with undetected malignant renal mass = false negative)
  • Malignant, correctly identified (patients with malignant renal mass and will be resected = true positive)
  • Metastasized/non-resectable (patients with a malignant renal mass which is unresectable/palliative)
  • Recurrence (Patients who show tumor recurrence after resection)
  • Dead
The state “Recurrence” was created, as sometimes a malignant renal mass reoccurs after resection. During each iteration, a patient is in one of the abovementioned states. Between each iteration, the patient can transition to another state or remain within the same state with a certain probability. The possible transitions are shown as arrows between the states in Fig. 2. For example, a patient has a malignant renal lesion, which is detected as malignant (Malignant, correctly identified). This malignant renal lesion will be resected. In the next iteration, the patient can transition to the states “Recurrence,” “Metastasized/Non-resectable,” “Dead,” or not transition at all. According to Margulis et al, there is a probability of recurrence of 1.8% [14] and according to Manikandan et al a probability of occurrence of metastasis after resection of 0.18% [15] and an annual risk of death with a localized tumor according to the age-adjusted US lifetables [16]. These are examples of transition probabilities and further parameters are shown in Table 1.
Table 1
Input parameters
Variable
Estimate
Distribution
Source
Pre-test probability for malignancy
0.19
β
O’Connor et al [17]
Age at diagnostic procedure
63 years
β
Karakiewicz et al [18]
Assumed WTP
$100,000/QALY
β
Sanders et al [19]
Discount rate
3%
β
Sanders et al [19]
Markov model time
10 years
β
Assumption
Diagnostic test performance
  MR sensitivity
0.8
β
Xi et al [8]
  MR specificity
0.35
β
Xi et al [8]
  MR + AI sensitivity
0.92
β
Xi et al [8]
  MR + AI specificity
0.41
β
Xi et al [8]
Costs (acute)
  Biopsy
$1535
γ
Medicare (Code: 50,200) [20]
  MR
$453
γ
Medicare (Code: 74,182) [20]
  ceCT
$246
γ
Medicare (Code: 74,170) [20]
  Timely surgery + treatment (true positive)
$7652
γ
Medicare (Code: 52,355 and 50,543) [20]
  No further action (true negative and false negative)
$0.00
γ
Assumption
  Unnecessary surgery (false positive)
$4884
γ
Medicare (Code: 52,355) [20]
Costs (long-term)
  Yearly costs without tumor
$0
γ
Assumption
  Yearly costs with localized (1st year)
$492
γ
Donat et al [21]
  Yearly costs with localized (after 1st year)
$246
γ
Donat et al [21]
  Yearly costs with metastatic (1st year)
$70,703
γ
Shih et al [22]
  Yearly costs with metastatic (after 1st year)
$34,716
γ
Shih et al [22]
Utilities
  Alive, benign tumor
1
β
Assumption
  After biopsy (= FP 1st year)
0.995
β
Adapted from Feldmann et al and Gassert et al [23, 24]
  Alive, resected, 1st year (including resection)
0.97
β
Adapted from Jiang et al [25]
  Alive, resected from 2nd year on
1
β
Adapted from Rossi et al [26]
  Alive, non-resected
1
β
Assumption
  Alive, resected, occurrence of metastasis
0.66
β
de Groot et al [27]
  QoL after biopsy for 1 month
0.995
β
Adapted from Feldmann et al [23]
  QoL after surgery for 1 month
0.7
β
Kim et al als Referenzpunkt (SF 12 Physical Component Score) [28]
  Annual risk of death without malignancy
0
β
Assumption
Transition probabilities
  Probability of initial non-R0 resection
5.7%
β
Orosco et al [29]
  Risk of metastases in false negative patients
1.0%
β
Bensalah et al [30]
  Probability of local recurrence after resection
1.8%
β
Margulis et al [14]
  Yearly probability of occurrence of metastases after resection
0.18%
β
Manikandan et al [15]
  Probability of successful surgery of local recurrence
41.2%
β
Thomas et al [31]
  Annual risk of death with localized tumor
Age adjusted
β
US Life Tables [16]
  Death due to surgery
0%
β
Chaudery et al [32]
  Detection of initially undetected localized lesion
100%
β
Assumption
  Risk of death with metastasis
33.5%
β
Howlader et al [33]
AI artificial intelligence, ceCT contrast-enhanced computer tomography, ICER incremental cost-effectiveness ratio, QALY quality-adjusted life year, QoL quality of life, WTP willingness-to-pay
The Markov model runtime was set to 10 years with a duration per iteration of 1 year, resulting in a total of 10 iterations.

Population

The population in this study is based on Xi et al [8]. In their study, patients with renal lesions confirmed by histology or imaging were retrospectively identified from two prominent academic centers in the United States (US) (HUP and MAY), two hospitals in the People’s Republic of China (SXH and PHH), and the Cancer Imaging Archive (TCIA) [8]. Most patients with renal lesions came from the academic centers in the US. Therefore, further data and literature included in this evaluation was based on the US healthcare system. Although this algorithm could be used in any kind of medical institution, our assumption will most likely hold true for academic centers, where further evaluations and handling of renal tumors is performed.

Input parameters

Model input parameter included pre-test probability of a malignant lesion present, and mean age at the diagnostic procedure as well as diagnostic accuracy of the MRI strategy and the MRI + AI strategy. For the Markov model simulation, costs and utilities (measured as QoL) per iteration as well as transition probabilities between states were retrieved through review of current literature. These parameters are displayed in Table 1.

Diagnostic test performances

Sensitivity and specificity for the correct classification of incidental renal lesions through a MRI strategy and a MRI + AI strategy are based on the study by Xi et al [8]. In this investigation, a multicenter cohort of 1162 renal lesions with a confirmed pathology or imaging diagnosis (655 of which were malignant and 507 of which were benign) were separated into training, validation, and test sets (70:20:10). A bagging classifier was used to build an ensemble model based on ResNet that included clinical factors with T1C and T2WI MR images to predict the pathology of renal tumors. Performance of the final model was evaluated against professional analysis. The ensemble deep learning model outperformed the average performance of the four experts in terms of sensitivity (0.92 vs. 0.80), and specificity (0.41 vs. 0.35) [8].

Costs

Starting from a US healthcare perspective, costs were estimated based on Medicare data and available literature (Table 1). It was assumed that if a localized malignant tumor was diagnosed, surgery would follow in all cases. Costs of timely surgery and resection of a malignant renal tumor were set to USD 7652 (Medicare codes 52,355 and 50,543) [20]. After treatment, the long-term yearly costs of follow-up were estimated at USD 492 (corresponding to two contrast-enhanced CT scans) in the first year after resection and at USD 246 (corresponding to one contrast-enhanced CT scan) for every further year [21]. According to Shih et al, costs for patients with a metastasized renal tumor were at USD 70,703 for the first year after diagnosis and at USD 34,716 for every following year [22]. In patients with false positive results, only the costs of surgery were attributed with USD 4884.

Utilities

For determination of the overall effectiveness, quality-adjusted life years (QALY) gained from each diagnostic procedure were calculated based on the QoL in each disease state. QoL was set to 1 for healthy patients as well as patients with an undetected malignant renal mass, as we assumed the patient to change disease state in the Markov model if the lesion became symptomatic. Due to the possible complications associated with surgery, QoL in the first year after resection of localized tumor was set to 0.97 [25]. However, according to previous literature, from the second year after resection QoL was set to 1 again [26]. In a metastasized disease state, QoL was set to 0.66 [27].

Transition probabilities

Transition probabilities were derived from recent literature and are displayed in Table 1. The probability of detection of an initially undetected malignant renal mass was assumed to be at 100% after 1 year. The probability of occurrence of metastases in patients with a localized tumor was assumed to be 1%, which corresponds to the risk of metastases in false negatively diagnosed lesions [30]. The annual probability of occurrence of metastases after resection of a localized renal malignancy was set to 0.18%, based on recent literature [15]. As metastasis was thought to be the cause of the disease’s deadly outcome, the risk of death with a localized renal malignancy was assumed to be comparable to the risk of death without a malignant renal mass [33]. The age-dependent annual risk of death with localized renal malignancy was based on the current US Life Tables [16].

Cost-effectiveness analysis

Cost-effectiveness analysis was performed using Markov simulations with a run time of 10 years after the detection of solid renal masses. The discount rate was set to 3.0% and willingness-to-pay (WTP) was set to USD 100,000 per QALY, as recommended in current guidelines. Willingness-to-pay is the valuation of health benefit in monetary terms and defined as the costs a society or healthcare system is willing to pay for an additional QALY gained [19]. In the base-case scenario, additional costs for the MRI + AI strategy as compared to the MRI strategy were set to USD 0 as this parameter was to be determined by the analysis. A deterministic sensitivity analysis of costs and diagnostic parameters was performed to evaluate model uncertainty and impact of variation of those parameters on the potential costs of the AI. These results were presented in a tornado diagram and given as a difference of incremental cost-effectiveness ratio (ICER) compared to the MRI strategy. The ICER is defined as the difference in cost between two possible interventions, divided by the difference in their effect and therefore represents the incremental cost associated with one additional QALY gained [19]. A probabilistic sensitivity analysis based on Monte Carlo simulations with 30,000 iterations was performed to show the individual patients outcomes based on the individual variation of the input parameters for evaluation of the robustness of the model. A threshold analysis was performed to define the maximum costs of the AI algorithm at different values for the WTP.

Results

Cost-effectiveness analysis

In the base-case scenario over a period of 10 years, total costs were at $8054 for the MRI strategy and at $7939 for the MRI + AI strategy if additional diagnostic costs for the use of AI were at USD 0. In the same scenario, the model yielded a cumulative effectiveness of 8.76 QALYs for the MRI strategy and a cumulative effectiveness of 8.77 for the MRI + AI strategy. Therefore, in the base-case scenario, the use of AI was the dominant strategy from a cost-effectiveness point of view.

Probabilistic sensitivity analysis

The outcomes of the 30,000 iterations are shown in Fig. 3A, a scatterplot depicting the effectiveness and cost of the scenario “MRI” (red) versus the scenario “MRI + AI” (blue). Visualization reveals that, overall, iterations in the diagnostic pathway of the MRI strategy are more expensive and less effective compared to iterations in the diagnostic pathway of the MRI + AI strategy. Figure 3B shows the cost-effectiveness acceptability curve, which indicates the proportion of iterations, which are cost-effective for the MRI + AI strategy based on a variation of the WTP threshold. In the range of USD 0/QALY to USD 200,000/QALY, more than 50% of the iterations are cost-effective for the MRI + AI strategy as compared to MRI strategy.

Deterministic sensitivity analysis

A deterministic sensitivity analysis was performed to account for variation of input parameters in literature. The results are shown as a tornado diagram in Fig. 4. ICER is positive if specificity of the MRI strategy increases above 0.38 and specificity of the MRI + AI strategy decreases below 0.39. However, within the indicated ranges of the input parameter, ICER stayed below the WTP threshold of USD 100,000/QALY in all cases.

Threshold analysis

A threshold analysis was performed to define the maximum costs of the AI algorithm at different WTP thresholds. If additional costs for the use of AI as compared to the MRI strategy remained below USD 115 per application, the MRI + AI strategy remains the dominant strategy. When increasing the costs for the use of AI further, the assumed WTP of USD 100,000/QALY was only reached at USD 667 per application. Values for the maximum costs of AI at other WTP thresholds are depicted in Fig. 5.

Discussion

This study emphasizes the potential cost-effectiveness rather than providing a definitive technology assessment. It indicates that integrating a deep learning–based diagnostic support (with a sensitivity of 0.92 and specificity of 0.41) into routine MR imaging is a cost-effective alternative compared to using MRI alone for distinguishing benign and malignant incidental renal lesions. Acknowledging the underlying assumptions, higher accuracies could potentially decrease the need for biopsies or surgical resections of benign renal lesions, ultimately lowering costs and improving patient outcomes by reducing the risk of delayed detection of malignant renal lesions. This is reciprocated by our study as the MRI + AI strategy remained the dominant strategy up to the threshold of $115. At a WTP of $100,000/QALY, additional cost of AI may be as high as $667 to remain the cost-effective alternative compared to the use of MRI alone.
Preoperative differentiation between benign and malignant renal lesions using non-invasive imaging techniques is a crucial factor for treatment planning but remains challenging from imaging. According to current clinical guidelines, suspicious renal lesions should be analyzed histopathologically via biopsy or resection [5]. Several studies can be found in literature using traditional machine-learning techniques, such as support vector machine and random forest, to distinguish incidental renal lesions based on CT radiomics [34, 35]. The CNN developed by Xi et al is one of the first studies using deep learning for differentiation of these lesions. If adapted into routine clinical practice, it has the potential to reduce the number of unnecessary diagnostics through more accurate classification [8]. However, to this date, there is not yet an AI product commercially available in Europe [36] or the United States [37] for this purpose.
To our knowledge, currently there are very few studies that tested the cost-effectiveness of AI-based algorithms in radiology. A study conducted by Ziegelmayer et al showed that the use of a 3D-convolutional neural network can be cost-effective when applied to CT-based lung cancer screening [11]. Similar to the current study, where the MRI + AI strategy remained the dominant strategy up to a threshold of $115 per patient, additional AI support in lung cancer screening remained the dominant strategy to a threshold of $68 [11]. Furthermore, van Leeuwen et al conducted a cost effectiveness-analysis on AI support for the early detection of vessel occlusion in the brain [10]. This study also showed that AI support is cost-effective and can reduce the number of missed large vessel occlusion with lower costs (− $156) and higher effectiveness (0.0095 QALYs). Van Leeuwen et al discuss the importance of using such CEAs to deduce the potential patient outcome and the benefit of an AI system to help not only its development, but also the translation into routine clinical practice [10]. In a related context, Vithlani et al and Jiao et al conducted comprehensive literature reviews, shedding light on numerous HEEs associated with AI-based health interventions [12, 13]. Their findings consistently revealed potential positive impacts on health outcomes. Notably, both manuscripts underscored the imperative for increased CEAs to effectively bridge the gap between the rapid progress of AI in healthcare and its practical application in real-world settings [12, 13]. Overall, these studies show that use of additional AI-based algorithms in routine imaging techniques can be beneficial in terms of both patient outcome and economic aspects when implemented into clinical diagnostics.
This study has limitations. Firstly, our model assumes the quality of life in a patient with a non-resected malignant tumor, as a study determining the quality of life in an unresected known malignant tumor would be unethical and therefore does not exist. Additionally, our focus is solely on MRI, excluding other techniques like contrast-enhanced ultrasound (CEUS) and contrast-enhanced CT, which have been proven to be a feasible diagnostic method in incidental renal lesions [38, 39]. However, the goal of this study was to evaluate the use of AI in diagnostic imaging, particular in MR imaging. To our knowledge, no algorithm has been developed so far to differentiate between benign and malignant renal lesions in imaging modalities other than MRI. Furthermore, the imaging quality and therefore diagnostic outcome of CEUS is dependent of the clinician’s skill set and the availability.
Moreover, while MRI is not commonly the initial choice for detecting incidental renal lesions, it is recommended, especially in cases of indeterminate renal masses, according to the ACR incidental findings committee [40]. It is important to note that no single imaging method can definitively distinguish between benign and malignant renal lesions. Nevertheless, the incorporation of AI alongside MRI has demonstrated encouraging outcomes, justifying our choice of this imaging approach. Furthermore, we have not incorporated alternative techniques aimed at enhancing the effectiveness of malignant detection in MRI, such as the multi-parametric MRI approach explored by Suresh de Silva et al [41]. This is because our primary emphasis in this paper is on AI, and currently, there is insufficient available data concerning novel methodologies.
Additionally, while deterministic sensitivity analysis may consider some parameter change, recommendations for each individual situation cannot be obtained from the model since cost-effectiveness analysis using decision-based models depends heavily on the input parameters. Furthermore, AI performance metrics in this study are based on a single study from Xi et al of a research algorithm tested on a split from the dataset it was trained on [8]. Published data on AI algorithms can vary widely depending on the study design, training cohort, and scanners. Also, the majority of the cohort in Xi et al was from the US with a few patientsfrom China. Despite Xi et al being the biggest study that compared AI performance with four expert radiologists in the differentiation of incidental renal lesions, future studies on this topic may show different results [8]. Several studies have assessed the effectiveness of AI in distinguishing between benign and malignant renal masses [4244]. For instance, Maasa’a et al employed MRI-based radiomics and machine learning to assess the differentiation of 182 renal lesions in 160 patients [43]. Similarly, Wentland et al compared CT-based radiomics and machine learning with radiologist interpretation focusing on 148 renal masses [44]. Prioritizing realism, we chose Xi et al’s research for its large multicentric cohort in our investigation [8], opting for a specific example due to the challenges of integrating outcomes from multiple studies in this research context.
In conclusion, our study suggests that the potential integration of an AI-based algorithm could be a valuable tool for differentiating benign and malignant incidental renal lesions, considering patient benefits and economic factors. However, it is crucial to recognize that our analysis, rooted in various assumptions and diagnostic accuracy uncertainties, serves as an example for approaching these considerations in a hypothetical scenario based on a single publication. Establishing a benchmark for potential costs associated with AI application may contribute to the accelerated translation of such systems into clinical routine, bearing in mind the speculative nature of our findings.

Declarations

Guarantor

The scientific guarantor of this publication is Felix G. Gassert (Department of diagnostic and interventional Radiology, Klinikum rechts der Isar, School of Medicine, Technical University of Munich, Ismaninger Str. 22, 81675 Munich, Germany). Email: felix.gassert@tum.de.

Conflict of interest

JR is a member of the European Radiology Advisory Editorial Board. He has not taken part in the review or selection process of this article. The remaining authors of this manuscript declare no relationships with any companies, whose products or services may be related to the subject matter of the article.

Statistics and biometry

One of the authors (FG) has significant statistical expertise.
Not applicable.

Ethical approval

Not applicable.

Study subjects or cohorts overlap

Not applicable.

Methodology

• cost-effectiveness
• performed at one institution
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://​creativecommons.​org/​licenses/​by/​4.​0/​.

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Unsere Produktempfehlungen

e.Med Interdisziplinär

Kombi-Abonnement

Für Ihren Erfolg in Klinik und Praxis - Die beste Hilfe in Ihrem Arbeitsalltag

Mit e.Med Interdisziplinär erhalten Sie Zugang zu allen CME-Fortbildungen und Fachzeitschriften auf SpringerMedizin.de.

e.Med Radiologie

Kombi-Abonnement

Mit e.Med Radiologie erhalten Sie Zugang zu CME-Fortbildungen des Fachgebietes Radiologie, den Premium-Inhalten der radiologischen Fachzeitschriften, inklusive einer gedruckten Radiologie-Zeitschrift Ihrer Wahl.

Literatur
1.
Zurück zum Zitat Winder M, Owczarek AJ, Chudek J, Pilch-Kowalczyk J, Baron J (2021) Are we overdoing it? Changes in diagnostic imaging workload during the years 2010–2020 including the impact of the SARS-CoV-2 pandemic. Healthcare (Basel) 9:1557 Winder M, Owczarek AJ, Chudek J, Pilch-Kowalczyk J, Baron J (2021) Are we overdoing it? Changes in diagnostic imaging workload during the years 2010–2020 including the impact of the SARS-CoV-2 pandemic. Healthcare (Basel) 9:1557
2.
Zurück zum Zitat Gill IS, Aron M, Gervais DA, Jewett MA (2010) Clinical practice. Small renal mass. N Engl J Med 362:624–634CrossRefPubMed Gill IS, Aron M, Gervais DA, Jewett MA (2010) Clinical practice. Small renal mass. N Engl J Med 362:624–634CrossRefPubMed
3.
Zurück zum Zitat Silverman SG, Pedrosa I, Ellis JH et al (2019) Bosniak Classification of Cystic Renal Masses, Version 2019: an update proposal and needs assessment. Radiology 292:475–488CrossRefPubMed Silverman SG, Pedrosa I, Ellis JH et al (2019) Bosniak Classification of Cystic Renal Masses, Version 2019: an update proposal and needs assessment. Radiology 292:475–488CrossRefPubMed
4.
Zurück zum Zitat Nicolau C, Antunes N, Paño B, Sebastia C (2021) Imaging characterization of renal masses. Medicina (Kaunas) 57:51 Nicolau C, Antunes N, Paño B, Sebastia C (2021) Imaging characterization of renal masses. Medicina (Kaunas) 57:51
5.
Zurück zum Zitat Escudier B, Porta C, Schmidinger M et al (2019) Renal cell carcinoma: ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up. Ann Oncol 30:706–720CrossRefPubMed Escudier B, Porta C, Schmidinger M et al (2019) Renal cell carcinoma: ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up. Ann Oncol 30:706–720CrossRefPubMed
6.
Zurück zum Zitat Patel U, Sokhi H (2012) Imaging in the follow-up of renal cell carcinoma. AJR Am J Roentgenol 198:1266–1276CrossRefPubMed Patel U, Sokhi H (2012) Imaging in the follow-up of renal cell carcinoma. AJR Am J Roentgenol 198:1266–1276CrossRefPubMed
7.
Zurück zum Zitat Sun XY, Feng QX, Xu X et al (2020) Radiologic-radiomic machine learning models for differentiation of benign and malignant solid renal masses: comparison with expert-level radiologists. AJR Am J Roentgenol 214:W44-w54CrossRefPubMed Sun XY, Feng QX, Xu X et al (2020) Radiologic-radiomic machine learning models for differentiation of benign and malignant solid renal masses: comparison with expert-level radiologists. AJR Am J Roentgenol 214:W44-w54CrossRefPubMed
8.
Zurück zum Zitat Xi IL, Zhao Y, Wang R et al (2020) Deep learning to distinguish benign from malignant renal lesions based on routine MR imaging. Clin Cancer Res 26:1944–1952CrossRefPubMed Xi IL, Zhao Y, Wang R et al (2020) Deep learning to distinguish benign from malignant renal lesions based on routine MR imaging. Clin Cancer Res 26:1944–1952CrossRefPubMed
9.
Zurück zum Zitat Tadavarthi Y, Vey B, Krupinski E et al (2020) The state of radiology AI: considerations for purchase decisions and current market offerings. Radiol Artif Intell 2:e200004CrossRefPubMedPubMedCentral Tadavarthi Y, Vey B, Krupinski E et al (2020) The state of radiology AI: considerations for purchase decisions and current market offerings. Radiol Artif Intell 2:e200004CrossRefPubMedPubMedCentral
10.
Zurück zum Zitat van Leeuwen KG, Meijer FJA, Schalekamp S et al (2021) Cost-effectiveness of artificial intelligence aided vessel occlusion detection in acute stroke: an early health technology assessment. Insights Imaging 12:133CrossRefPubMedPubMedCentral van Leeuwen KG, Meijer FJA, Schalekamp S et al (2021) Cost-effectiveness of artificial intelligence aided vessel occlusion detection in acute stroke: an early health technology assessment. Insights Imaging 12:133CrossRefPubMedPubMedCentral
11.
Zurück zum Zitat Ziegelmayer S, Graf M, Makowski M, Gawlitza J, Gassert F (2022) Cost-effectiveness of artificial intelligence support in computed tomography-based lung cancer screening. Cancers (Basel) 14:1729 Ziegelmayer S, Graf M, Makowski M, Gawlitza J, Gassert F (2022) Cost-effectiveness of artificial intelligence support in computed tomography-based lung cancer screening. Cancers (Basel) 14:1729
12.
Zurück zum Zitat Jiao W, Zhang X, D’Souza F (2023) The economic value and clinical impact of artificial intelligence in healthcare: a scoping literature review. IEEE Access 11:123445–123457CrossRef Jiao W, Zhang X, D’Souza F (2023) The economic value and clinical impact of artificial intelligence in healthcare: a scoping literature review. IEEE Access 11:123445–123457CrossRef
13.
Zurück zum Zitat Vithlani J, Hawksworth C, Elvidge J, Ayiku L, Dawoud D (2023) Economic evaluations of artificial intelligence-based healthcare interventions: a systematic literature review of best practices in their conduct and reporting. Front Pharmacol 14:1220950 Vithlani J, Hawksworth C, Elvidge J, Ayiku L, Dawoud D (2023) Economic evaluations of artificial intelligence-based healthcare interventions: a systematic literature review of best practices in their conduct and reporting. Front Pharmacol 14:1220950
14.
Zurück zum Zitat Margulis V, McDonald M, Tamboli P, Swanson DA, Wood CG (2009) Predictors of oncological outcome after resection of locally recurrent renal cell carcinoma. J Urol 181:2044–2051CrossRefPubMed Margulis V, McDonald M, Tamboli P, Swanson DA, Wood CG (2009) Predictors of oncological outcome after resection of locally recurrent renal cell carcinoma. J Urol 181:2044–2051CrossRefPubMed
15.
Zurück zum Zitat Manikandan R, Srinivasan V, Rané A (2004) Which is the real gold standard for small-volume renal tumors? Radical nephrectomy versus nephron-sparing surgery. J Endourol 18:39–44CrossRefPubMed Manikandan R, Srinivasan V, Rané A (2004) Which is the real gold standard for small-volume renal tumors? Radical nephrectomy versus nephron-sparing surgery. J Endourol 18:39–44CrossRefPubMed
16.
Zurück zum Zitat Arias E, Xu J, Kochanek KD (2019) United States Life Tables, 2016. National vital statistics reports: from the Centers for Disease Control and Prevention 68:1–66 Arias E, Xu J, Kochanek KD (2019) United States Life Tables, 2016. National vital statistics reports: from the Centers for Disease Control and Prevention 68:1–66
17.
Zurück zum Zitat O’Connor SD, Pickhardt PJ, Kim DH, Oliva MR, Silverman SG (2011) Incidental finding of renal masses at unenhanced CT: prevalence and analysis of features for guiding management. AJR Am J Roentgenol 197:139–145CrossRefPubMed O’Connor SD, Pickhardt PJ, Kim DH, Oliva MR, Silverman SG (2011) Incidental finding of renal masses at unenhanced CT: prevalence and analysis of features for guiding management. AJR Am J Roentgenol 197:139–145CrossRefPubMed
18.
Zurück zum Zitat Karakiewicz PI, Jeldres C, Suardi N et al (2008) Age at diagnosis is a determinant factor of renal cell carcinoma-specific survival in patients treated with nephrectomy. Can Urol Assoc J 2:610–617CrossRefPubMedPubMedCentral Karakiewicz PI, Jeldres C, Suardi N et al (2008) Age at diagnosis is a determinant factor of renal cell carcinoma-specific survival in patients treated with nephrectomy. Can Urol Assoc J 2:610–617CrossRefPubMedPubMedCentral
19.
Zurück zum Zitat Sanders GD, Neumann PJ, Basu A et al (2016) Recommendations for conduct, methodological practices, and reporting of cost-effectiveness analyses: second panel on cost-effectiveness in health and medicine. JAMA 316:1093–1103CrossRefPubMed Sanders GD, Neumann PJ, Basu A et al (2016) Recommendations for conduct, methodological practices, and reporting of cost-effectiveness analyses: second panel on cost-effectiveness in health and medicine. JAMA 316:1093–1103CrossRefPubMed
21.
Zurück zum Zitat Donat SM, Diaz M, Bishoff JT et al (2013) Follow-up for Clinically Localized Renal Neoplasms: AUA Guideline. J Urol 190:407–416CrossRefPubMed Donat SM, Diaz M, Bishoff JT et al (2013) Follow-up for Clinically Localized Renal Neoplasms: AUA Guideline. J Urol 190:407–416CrossRefPubMed
22.
Zurück zum Zitat Shih Y-CT, Xu Y, Chien C-R et al (2019) Rising economic burden of renal cell carcinoma among elderly patients in the USA: part II-an updated analysis of SEER-Medicare data. Pharmacoeconomics 37:1495–1507CrossRefPubMedPubMedCentral Shih Y-CT, Xu Y, Chien C-R et al (2019) Rising economic burden of renal cell carcinoma among elderly patients in the USA: part II-an updated analysis of SEER-Medicare data. Pharmacoeconomics 37:1495–1507CrossRefPubMedPubMedCentral
23.
Zurück zum Zitat Feldmann Y, Böer K, Wolf G, Busch M (2018) Complications and monitoring of percutaneous renal biopsy - a retrospective study. Clin Nephrol 89:260–268CrossRefPubMed Feldmann Y, Böer K, Wolf G, Busch M (2018) Complications and monitoring of percutaneous renal biopsy - a retrospective study. Clin Nephrol 89:260–268CrossRefPubMed
24.
Zurück zum Zitat Gassert FG, Rübenthaler J, Cyran CC et al (2021) (18)F FDG PET/MRI with hepatocyte-specific contrast agent for M staging of rectal cancer: a primary economic evaluation. Eur J Nucl Med Mol Imaging 48:3268–3276CrossRefPubMedPubMedCentral Gassert FG, Rübenthaler J, Cyran CC et al (2021) (18)F FDG PET/MRI with hepatocyte-specific contrast agent for M staging of rectal cancer: a primary economic evaluation. Eur J Nucl Med Mol Imaging 48:3268–3276CrossRefPubMedPubMedCentral
25.
Zurück zum Zitat Jiang J, Zheng X, Qin J et al (2009) Health-related quality of life after hand-assisted laparoscopic and open radical nephrectomies of renal cell carcinoma. Int Urol Nephrol 41:23–27CrossRefPubMed Jiang J, Zheng X, Qin J et al (2009) Health-related quality of life after hand-assisted laparoscopic and open radical nephrectomies of renal cell carcinoma. Int Urol Nephrol 41:23–27CrossRefPubMed
26.
Zurück zum Zitat Rossi SH, Klatte T, Stewart GD (2018) Quality of life outcomes in patients with localised renal cancer: a literature review. World J Urol 36:1961–1972CrossRefPubMedPubMedCentral Rossi SH, Klatte T, Stewart GD (2018) Quality of life outcomes in patients with localised renal cancer: a literature review. World J Urol 36:1961–1972CrossRefPubMedPubMedCentral
27.
Zurück zum Zitat de Groot S, Redekop WK, Versteegh MM et al (2018) Health-related quality of life and its determinants in patients with metastatic renal cell carcinoma. Qual Life Res 27:115–124CrossRefPubMed de Groot S, Redekop WK, Versteegh MM et al (2018) Health-related quality of life and its determinants in patients with metastatic renal cell carcinoma. Qual Life Res 27:115–124CrossRefPubMed
28.
Zurück zum Zitat Kim SB, Williams SB, Cheng SC, Sanda MG, Wagner AA (2012) Evaluation of patient-reported quality-of-life outcomes after renal surgery. Urology 79:1268–1273CrossRefPubMed Kim SB, Williams SB, Cheng SC, Sanda MG, Wagner AA (2012) Evaluation of patient-reported quality-of-life outcomes after renal surgery. Urology 79:1268–1273CrossRefPubMed
30.
Zurück zum Zitat Bensalah K, Pantuck AJ, Crepel M et al (2008) Prognostic variables to predict cancer-related death in incidental renal tumours. BJU Int 102:1376–1380CrossRefPubMed Bensalah K, Pantuck AJ, Crepel M et al (2008) Prognostic variables to predict cancer-related death in incidental renal tumours. BJU Int 102:1376–1380CrossRefPubMed
31.
Zurück zum Zitat Thomas AZ, Adibi M, Borregales LD et al (2015) Surgical management of local retroperitoneal recurrence of renal cell carcinoma after radical nephrectomy. J Urol 194:316–322CrossRefPubMedPubMedCentral Thomas AZ, Adibi M, Borregales LD et al (2015) Surgical management of local retroperitoneal recurrence of renal cell carcinoma after radical nephrectomy. J Urol 194:316–322CrossRefPubMedPubMedCentral
32.
Zurück zum Zitat Chaudery H, MacDonald N, Ahmad T et al (2019) Acute kidney injury and risk of death after elective surgery: prospective analysis of data from an international cohort study. Anesth Analg 128:1022–1029CrossRefPubMed Chaudery H, MacDonald N, Ahmad T et al (2019) Acute kidney injury and risk of death after elective surgery: prospective analysis of data from an international cohort study. Anesth Analg 128:1022–1029CrossRefPubMed
34.
Zurück zum Zitat Raman SP, Chen Y, Schroeder JL, Huang P, Fishman EK (2014) CT texture analysis of renal masses: pilot study using random forest classification for prediction of pathology. Acad Radiol 21:1587–1596CrossRefPubMedPubMedCentral Raman SP, Chen Y, Schroeder JL, Huang P, Fishman EK (2014) CT texture analysis of renal masses: pilot study using random forest classification for prediction of pathology. Acad Radiol 21:1587–1596CrossRefPubMedPubMedCentral
35.
Zurück zum Zitat Hodgdon T, McInnes MD, Schieda N, Flood TA, Lamb L, Thornhill RE (2015) Can quantitative CT texture analysis be used to differentiate fat-poor renal angiomyolipoma from renal cell carcinoma on unenhanced CT images? Radiology 276:787–796CrossRefPubMed Hodgdon T, McInnes MD, Schieda N, Flood TA, Lamb L, Thornhill RE (2015) Can quantitative CT texture analysis be used to differentiate fat-poor renal angiomyolipoma from renal cell carcinoma on unenhanced CT images? Radiology 276:787–796CrossRefPubMed
38.
Zurück zum Zitat Gassert F, Schnitzer M, Kim SH et al (2021) Comparison of magnetic resonance imaging and contrast-enhanced ultrasound as diagnostic options for unclear cystic renal lesions: a cost-effectiveness analysis. Ultraschall Med 42:411–417CrossRefPubMed Gassert F, Schnitzer M, Kim SH et al (2021) Comparison of magnetic resonance imaging and contrast-enhanced ultrasound as diagnostic options for unclear cystic renal lesions: a cost-effectiveness analysis. Ultraschall Med 42:411–417CrossRefPubMed
39.
Zurück zum Zitat Runtemund J, Rübenthaler J, von Münchhausen N et al (2022) Diagnostic workup for patients with solid renal masses: a cost-effectiveness analysis. Cancers 14:2235CrossRefPubMedPubMedCentral Runtemund J, Rübenthaler J, von Münchhausen N et al (2022) Diagnostic workup for patients with solid renal masses: a cost-effectiveness analysis. Cancers 14:2235CrossRefPubMedPubMedCentral
40.
Zurück zum Zitat Herts BR, Silverman SG, Hindman NM et al (2018) Management of the incidental renal mass on CT: a white paper of the ACR Incidental Findings Committee. J Am Coll Radiol 15:264–273CrossRefPubMed Herts BR, Silverman SG, Hindman NM et al (2018) Management of the incidental renal mass on CT: a white paper of the ACR Incidental Findings Committee. J Am Coll Radiol 15:264–273CrossRefPubMed
41.
Zurück zum Zitat de Silva S, Lockhart KR, Aslan P et al (2022) Differentiation of renal masses with multi-parametric MRI: the de Silva St George classification scheme. BMC Urol 22:141CrossRefPubMedPubMedCentral de Silva S, Lockhart KR, Aslan P et al (2022) Differentiation of renal masses with multi-parametric MRI: the de Silva St George classification scheme. BMC Urol 22:141CrossRefPubMedPubMedCentral
42.
Zurück zum Zitat Erdim C, Yardimci AH, Bektas CT et al (2020) Prediction of benign and malignant solid renal masses: machine learning-based CT texture analysis. Acad Radiol 27:1422–1429CrossRefPubMed Erdim C, Yardimci AH, Bektas CT et al (2020) Prediction of benign and malignant solid renal masses: machine learning-based CT texture analysis. Acad Radiol 27:1422–1429CrossRefPubMed
43.
Zurück zum Zitat Massa’a RN, Stoeckl EM, Lubner MG et al (2022) Differentiation of benign from malignant solid renal lesions with MRI-based radiomics and machine learning. Abdom Radiol (NY) 47:2896–2904CrossRefPubMed Massa’a RN, Stoeckl EM, Lubner MG et al (2022) Differentiation of benign from malignant solid renal lesions with MRI-based radiomics and machine learning. Abdom Radiol (NY) 47:2896–2904CrossRefPubMed
44.
Zurück zum Zitat Wentland AL, Yamashita R, Kino A et al (2023) Differentiation of benign from malignant solid renal lesions using CT-based radiomics and machine learning: comparison with radiologist interpretation. Abdom Radiol (NY) 48:642–648CrossRefPubMed Wentland AL, Yamashita R, Kino A et al (2023) Differentiation of benign from malignant solid renal lesions using CT-based radiomics and machine learning: comparison with radiologist interpretation. Abdom Radiol (NY) 48:642–648CrossRefPubMed
Metadaten
Titel
Artificial intelligence support in MR imaging of incidental renal masses: an early health technology assessment
verfasst von
Alexander W. Marka
Johanna Luitjens
Florian T. Gassert
Lisa Steinhelfer
Egon Burian
Johannes Rübenthaler
Vincent Schwarze
Matthias F. Froelich
Marcus R. Makowski
Felix G. Gassert
Publikationsdatum
23.02.2024
Verlag
Springer Berlin Heidelberg
Erschienen in
European Radiology
Print ISSN: 0938-7994
Elektronische ISSN: 1432-1084
DOI
https://doi.org/10.1007/s00330-024-10643-5

Update Radiologie

Bestellen Sie unseren Fach-Newsletter und bleiben Sie gut informiert.