Skip to main content

A role for artificial intelligence in molecular imaging of infection and inflammation

Abstract

The detection of occult infections and low-grade inflammation in clinical practice remains challenging and much depending on readers’ expertise. Although molecular imaging, like [18F]FDG PET or radiolabeled leukocyte scintigraphy, offers quantitative and reproducible whole body data on inflammatory responses its interpretation is limited to visual analysis. This often leads to delayed diagnosis and treatment, as well as untapped areas of potential application. Artificial intelligence (AI) offers innovative approaches to mine the wealth of imaging data and has led to disruptive breakthroughs in other medical domains already. Here, we discuss how AI-based tools can improve the detection sensitivity of molecular imaging in infection and inflammation but also how AI might push the data analysis beyond current application toward predicting outcome and long-term risk assessment.

Introduction

Artificial intelligence (AI) is considered to be the key to precision medicine and transforming health care (Denny and Collins 2021). In line with other imaging disciplines, such as microscopy (Meijering et al. 2016) and pathology (Colling et al. 2019; Laak et al. 2021), images obtained from routine clinical procedures represent rich and minable datasets on specific tissue characteristics (Gillies et al. 2016; Aerts et al. 2014). This realization urged the development of AI-based technologies to exploit these wealthy data sources (Parmar et al. 2018; Hosny et al. 2018). Although practical issues concerning data sharing, data safety and standardization are yet to be resolved (He et al. 2019; Currie and Hawk 2021), ongoing developments in AI will drive its implementation in the field of medical imaging (Currie and Rohren 2021). When it comes to application of AI-based technology for nuclear imaging modalities such as positron emission tomography (PET) and single-photon emission tomography (SPECT), excellent reviews which discuss modality-specific potential and limitations are available from the recent literature (Hatt et al. 2021; Uribe et al. 2019; Zukotynski et al. 2021; Decuyper et al. 2021). A key asset of nuclear imaging modalities is their whole body field-of-view and hence the capacity to quantify the distribution of tracers targeting specific biological processes where several organs and tissues are involved. Furthermore, dynamic imaging in nuclear medicine offers the possibility to temporally resolve systemic processes. Both aspects of nuclear imaging are extremely useful to develop a ‘systems biology’ approach based on nuclear imaging to characterize host’ immune responses in infection and inflammation.

AI methodology is evolving rapidly and it is beyond the scope of this review to provide a comprehensive overview on current concepts in image analysis. In general, AI-based approaches can be divided into supervised and unsupervised learning methods. Supervised learning requires data which is considered a ground truth or a gold standard, like histopathology. Supervised learning therefore is a mathematical way to approximate a model using a labeled training dataset which is then optimized in iterative steps. Typically, validation and test datasets are needed to assess the accuracy of the developed model. Unsupervised learning is trained to recognize patterns in unlabeled data without ground truth information. In unsupervised learning, algorithms are searching for regularities that can be used to define relationships like groups with similar features in an unlabeled dataset. Furthermore, unsupervised learning methods are used for capturing noise in data or to generate new data samples. Clustering methods like k-means are common unsupervised approaches to find patterns between data points in a dataset. More sophisticated approaches use, for example, trained neural networks which allow to model more complex relationships with only little assumptions (LeCun et al. 2015).

While in its early days, now is the time to also consider the potential roles of AI specifically in molecular imaging of infection and inflammation. ‘Precision medicine’ in the field of inflammation translates to early identification of patients at risk for inflammatory diseases and tailored treatment duration based on individual characteristics of a patients’ immune system.

In recent years it became evident that the activation of the immune system requires metabolic reprogramming, especially in regard to glucose metabolism (Gaber et al. 2017), thus in principle leads to effects measurable with 2-[18F]fluoro-2-deoxy-D-glucose ([18F]FDG) PET. Likewise, the spatial distribution of immune cells throughout the body determines the effectiveness of immune responses, which is assessable, for example, by radiolabeled leukocyte scintigraphy. While extensively studied in the aspect of cancer immunotherapy, these effects are similarly important in inflammatory diseases including infections, autoimmune disorders and atherosclerosis.

An important aspect of inflammation is the emerging concept of trained immunity: long-term functional reprogramming of the innate immune cells which co-determines responses to subsequent triggers (Netea et al. 2020a; Schultze et al. 2018). The development of trained immunity is determined by epigenetic reprogramming and profound rewiring of metabolic circuits in immune cells.

Current routine analysis of imaging techniques like [18F]FDG PET or radiolabeled leukocyte scintigraphy rely on the visual detection of foci in symptomatic patients, which heavily depends on the readers’ reference. Given the pivotal role of metabolic reprogramming in a range of inflammatory conditions, it emerges that at present only the tip of the iceberg of the available information is extracted out of the acquired data.

The recent advances in AI might push the analysis of current imaging techniques toward a more comprehensive understanding of inflammatory diseases including atherosclerosis and infections, to the detection of pathological immune responses even in asymptomatic patients on the long run.

In this short communication we propose three potential tasks for AI, ranging from practical to more hypothetical, where AI-based technologies can be applied to improve current practice.

Artificial intelligence to improve detection

Over the past years, [18F]FDG PET/CT has established its central role in the diagnosis and follow-up of infectious diseases and inflammatory conditions (Slart et al. 2018a; Signore et al. 2019; Chakfe et al. 2020; Jamar et al. 2013). Based on the high sensitivity and favorable whole-body view, the range of clinical indications continues to expand and the questions to be addressed are increasingly complex. At these far-end applications in inflammation imaging, current PET technologies have probably met their limits of detection and discriminative power (Fig. 1).

Fig. 1
figure 1

Artificial intelligence to improve detection. A graphical illustration of the typical dynamics of an immune response upon a trigger, with rapid increase, associated with increased glycolysis in effector cells which can be measured by [18F]FDG PET and expressed as maximum or mean standardized uptake values (SUVmax/mean). As soon as the causative trigger is cleared, inflammatory responses also include repair processes to gradually return to a state of tissue homeostasis

This holds true particularly for infections at occult sites, low-grade infections or low-grade inflammatory conditions that are diagnosed late and treated with delay (Hipfl et al. 2021; Laohapensang et al. 2017; Talha et al. 2020). Although this concerns a minority of patients, these cases consume a lot of health care related services, multiple diagnostic tests are being performed and prolonged treatment is required with increased likelihood to encounter complications. For example, the diagnosis of infectious (native or prosthetic valve) endocarditis currently requires a composite of clinical, microbiological and imaging (ultrasound and [18F]FDG PET) to accomplish reasonable sensitivity and specificity (Chakfe et al. 2020; Habib et al. 2015). However, this investment in diagnostic accuracy is mandatory as we have learned that insufficient treatment of even these small intravascular infectious foci is associated with increased mortality (Jaltotage et al. 2021; Chirillo 2021).

Moreover, in the premise of ‘precision medicine,’ (intravenous) antibiotic treatment durations tend to be shortened (Berrevoets et al. 2019; Kouijzer et al. 2021) to avoid overtreatment and reduce health care expenditure. Thus, discriminating persistent foci of active infection from tissue remodeling at the end of antibiotic or anti-inflammatory treatment will be an increasingly relevant, but challenging task for [18F]FDG PET imaging.

As there is a need for improved detection of low-grade or localized infections, which inherent features of nuclear imaging techniques are limiting? In comparison with computed tomography (CT) nuclear imaging techniques suffer from long acquisition times, inherently resulting in motion artifacts. Furthermore, nuclear imaging is constraint by radiation dose and the associated safety considerations, which together with the urge to detect low-grade of localized infection, call for further optimization of detection efficiency of PET and SPECT systems. The introduction of new PET scanners with digital detector technology and a long-axial field of view of 100 cm or more provide significant improvements in this regard. These developments have the potential to significantly reduce the data acquisition times in PET, which makes a high-resolution whole-body PET scan in less than 5 min possible (Alberts et al. 2021; Filippi and Schillaci 2022). Besides that, the substantial advances of this generation of scanners also allow a better temporal and spatial resolution as well as reduction of the administered radiation dose.

The relatively poor spatial resolution of PET (3–4 mm) and SPECT (8–9 mm) hampers the accurate assessment of anatomical regions with respiratory and cardiac motion. This is particularly relevant for imaging subtle changes in signal intensity in the myocardium, for example when endocarditis is suspected; or discrimination of [18F]FDG uptake in aortic root complications after recent vascular graft surgery.

Cardiac and respiratory motion, however, are highly standardized movements that can be modelled using AI-based technologies (Fig. 2). These are particularly suited to reconstruct images by incorporating previously learned information that compensate for motion. Indeed, data-driven approaches for PET image reconstruction that compensates for respiratory motion are increasingly available (Buther et al. 2016; Feng et al. 2019; Schleyer et al. 2011), paralleling developments in CT (Saeedan et al. 2021). Previously, electrocardiogram (ECG)-gated motion correction of [18F]-NaF uptake in coronary arteries in patients with myocardial infarction or stable angina had significant impact on lesion quantification (Rubeaux et al. 2016). Along the same line for endocarditis, more accurate detection of small infectious foci in the plane of cardiac valves would increase detection rates and allow better co-localization with findings on ultrasound or cardiac CT, which improves diagnostic accuracy (Hove et al. 2021). On a more general note, deep-learning methods to reconstruct whole body PET images without the signal-derived input for attenuation and scatter correction by its CT component are reported (Yang et al. 2021; Haggstrom et al. 2019), with non-inferior image quality, but much faster reconstruction times. Although large-scale comparative studies supporting AI-based motion correction are lacking, these approaches demonstrated that pre-learned information can be incorporated in AI-based reconstructions of PET acquisitions.

Fig. 2
figure 2

Potential AI workflow to improve image quality by cardiac and respiratory motion correction. The pre-learned, highly standardized movements of the heart and the lung can be integrated in the image reconstruction in order to optimize the image quality leading to advantages in the visual assessment by the nuclear medicine physician

In addition to factors that affect the measured signal intensity, the presence of noise in PET and SPECT data impairs accurate visual assessment and diagnostic accuracy of scan images in low-count statistics and the detection of small foci with little signal-to-noise ratio (Minarik et al. 2020). For example, radiolabeled autologous leukocytes for SPECT imaging have long been used to detect infectious foci and with sufficient specificity to discriminate infection from inflammation (de Vries et al. 2010; Roca et al. 2010). However, its inferior image quality due to low-count statistics and high levels of noise resulted in a rapid take-over by [18F]FDG PET/CT for these indications (Jamar et al. 2013), as image quality and system sensitivity were preferred despite the use of a less specific tracer. AI-based technologies can be exploited to reduce noise in such settings, which will positively impact image interpretation. For PET imaging the assignment of a line-of-response (LOR) for accurate image reconstruction can be corrupted by non-perpendicular coincidences, resulting in uncertainties in positioning the input signal. A deep learning estimator has been developed to predict the depth-of-interaction of incoming photons in pixelated detectors, which resulted in improved performance (Zatcepin et al. 2020). Improved positioning of input signals for monolithic detectors has been improved using convolutional neural networks that integrates the charge of silicon photomultipliers to predict locations of non-perpendicular coincidences (He et al. 2021). Compton scattering in the detection crystal results in incorrectly assigned LORs and contributes to system noise for PET and SPECT imaging. Deep learning algorithms trained on Monte Carlo simulation data showed improved LOR recovery rates and sensitivity by including accurate position of events in image reconstruction (Bergeron et al. 2014). Furthermore, in PET imaging, prediction of adverse cardiovascular events has recently been studied through the implementation of transfer learning, which allows for data economization while boosting image recognition capabilities and broadening the horizon of network architectures that can be constructed (Vos et al. 2019).

Image denoising based on deep learning methods is applied in general image restoration in cases of low or lack of spatial input. Several studies have now shown potential to convert low-count to high-count PET or SPECT images using U-Net (Kaplan and Zhu 2019; Dietze et al. 2019) or ResNet (Gong et al. 2019; Cui et al. 2019) algorithms. Lastly, on the hardware site of development, the new generation of long-axis field-of-view PET scanners have an even better sensitivity (Alberts et al. 2021; Badawi et al. 2019) which can also be exploited to increase signal-to-noise ratio.

Artificial intelligence to predict outcomes

Immune responses are a complex series of events that involve different immune cell populations and requires a concerted action in multiple body compartments (Spitzer et al. 2017; Chavakis et al. 2019). For example, upon infection, inflammatory monocytes are recruited from the bone marrow and spleen and increased myelo- and granulopoiesis should compensate for the loss of these effector cells in peripheral tissues (Hotchkiss et al. 2016). Indeed, most patients referred for nuclear imaging under suspicion of an infectious or inflammatory condition have symptoms and clinical signs indicative of systemic immune responses, such as fever, increased C-reactive protein and erythrocyte sedimentation rate and leukocytosis. The questions here are whether we can capture these systemic responses using molecular imaging and use this information to improve outcome prediction (Fig. 3).

Fig. 3
figure 3

Artificial intelligence to predict outcomes

The switch from a quiescent to an activated status inevitably comes with metabolic reprogramming of immune cells, resulting in increased glycolytic capacity (Netea et al. 2020a; Arts et al. 2017, 2018). As [18F]FDG-PET is a highly sensitive technique to quantify glycolysis on a whole-body scale, we and others have demonstrated that increased uptake of [18F]FDG-PET in organs involved in hematopoiesis and immune activation, e.g., bone marrow, spleen and vascular system, associates with the state of immune activation (van der Heijden et al. 2020; Bernelot Moens et al. 2016; Valk et al. 2016; Joseph et al. 2017; Stiekema et al. 2019; Ungar et al. 2020; Kalafati et al. 2020). Paralleling mechanisms might also play a role in the responsiveness of immune cells in anticancer immunity (Kalafati et al. 2020; Netea and Joosten 2018; Schwenck et al. 2020; Seith et al. 2020). Increased [18F]FDG uptake in bone marrow or spleen, as substrate of systemic immune activation, are associated with improved clinical outcome in melanoma patients under immune checkpoint inhibitors (Seban et al. 2021). This effect could potentially also be observed in overacting autoimmune events by [18F]FDG PET (Spitzer et al. 2017; Kalafati et al. 2020; Flint et al. 2017).

In patients with atherosclerotic cardiovascular diseases there is already evidence that increased [18F]FDG uptake in the arterial wall, spleen and bone marrow predicts future occurrence of cardiovascular events (Emami et al. 2015). To the contrary, clinical studies in sepsis showed that patients with decreased glycolytic capacity in leukocytes have a worse clinical outcome (Cheng et al. 2016; Hotchkiss et al. 2013; Kaufmann et al. 2018), a phenomenon called ‘immune paralysis.’

Thus, as the whole-body field of view of PET allows to assess body compartments involved in the systemic response to infection or inflammation, which element hampers the analysis of these potentially predictive data? At present, the integration of this additional data on immune metabolism in multiple body compartments depends on the limited human capacity to deal with multi-dimensional data. Systems biology studies are integrating large-scale (‘omics’) data, e.g., from different tissues (Kidd et al. 2014) and therefore are urged to implement AI-based technologies for data analysis (Camacho et al. 2018). These studies enabled a more comprehensive mechanistic insight in multidimensional complex diseases (Yang 2020), such as cardiovascular disease (Lempiainen et al. 2018; Makinen et al. 2014; Shu et al. 2017; Slart et al. 2021). These studies demonstrated that the net outcomes on patient level result from perturbations in multiple body compartment involving diverse cell types and molecular pathways. The integration of these different scales of data, in which the contribution of the individual components can vary from subject to subject, demonstrates that cardiovascular disease is promoted by increased inflammatory pathways in the liver, adipose tissue and vascular system (Libby et al. 2019), as well as by the immune response, which is not limited to the arterial wall as it is also detectable in the bone marrow and the spleen.

Similar to infections, effective anticancer immune responses require an integrated action from both innate and adaptive immune cells (Chiossone et al. 2018) including their activation in local and distant body compartments (Spitzer et al. 2017; Kalafati et al. 2020). These observations underscore the general concept that on a systems level, metabolism and immune responses are connected (Flint et al. 2017).

Thus, immune metabolism is a preeminent example of reciprocal interactions on a cellular, organ and system level (Lercher et al. 2020) that impact inflammatory and infectious diseases as well as the homeostasis of the immune system, as will be discussed later.

[18F]FDG PET is well-suited to measure metabolic activity across these multiple circuitries, provided that AI-based technologies are developed to extract and process these data in predictive models (Fig. 4).

Fig. 4
figure 4

AI could be used to develop a predictive score calculated from the extracted information on the vascular, lymphoid and hematopoietic system. This score characterizes the level of systemic inflammation, for example, in a patient with suspected vasculitis and therefore supports the assessment of the nuclear medicine physician

Artificial intelligence to provide prognostic information

Pathogen and damage-associated molecular patterns are sensed by cells of the innate immune system, inducing rapid activation and non-specific responses to eliminate the trigger. A growing body of evidence suggests that, in addition to these rapid ‘first line-of-defense’ responses, long-term functional reprogramming of innate immune cells occurs and co-determines responses to subsequent triggers. So, ‘immunological memory’ is no longer considered to be exclusive for cells of the adaptive immune system, but also occurs in innate immune cells, both in hematopoietic progenitor cells (central trained immunity) and in differentiated cells such as monocytes, macrophages and natural killer cells (peripheral trained immunity) (Netea et al. 2020a; Schultze et al. 2021). These ‘trained immunity’ phenotypes have implications for the response to future infections (Netea et al. 2020b). Central in the development of trained immunity is epigenetic reprogramming, which is closely intertwined with metabolic reprogramming, characterized by an increased glycolysis, glutaminolysis and mevalonate pathway, among others. This mechanism allows altered immune-metabolic circuits in immune cells to respond with faster and higher upregulation of aerobic glycolysis and subsequent cytokine production capacity upon subsequent infectious triggers (Dominguez-Andres and Netea 2019).

Environmental inflictions, and the associated inflammatory response on tissue level, culminate during life span and are considered an integral part of ageing (Lopez-Otin et al. 2013; Sugimoto et al. 2019). It is tempting to speculate that beyond the development of whole body [18F]FDG PET as a predictive imaging classifier based on immune metabolic phenotypes, there might be a prognostic role for [18F]FDG PET to determine long-term outcome associated with chronic inflammatory conditions (Fig. 5).

Fig. 5
figure 5

Artificial intelligence to provide prognostic information

In addition to trained immune cells, repetitively triggered stromal cells can also convert into a state of chronic low-grade inflammation (Bekkering et al. 2016a, b, 2018, 2019; Leentjens et al. 2018), with detrimental impact on long-term clinical outcomes. For example, endothelial cells of the vascular system, which are key in directing the trafficking of immune cells to inflamed tissues, also respond to systemic inflammatory mediators (Pober and Sessa 2007). Moreover, these endothelial cells are exposed to a multitude of noxes throughout a lifespan, e.g., hypercholesterolemia or hyperglycemia, inducing cell damage and low-grade inflammation aimed to maintain endothelial integrity. The role of [18F]FDG PET imaging in large vessel vasculitis is established (Jamar et al. 2013; Slart et al. 2018b) and is currently explored for chronic inflammatory conditions (van der Heijden et al. 2020; Noz et al. 2020; Valk et al. 2017). Defining a threshold on this sliding scale from overt vascular wall inflammation, e.g., in the context of vasculitis, associated with symptoms and representing a clinical entity, to low-grade inflammation, associated with chronic inflammatory conditions such as atherosclerosis, perhaps is a new prognostic task for [18F]FDG PET that comes within reach with the advent of AI-based technologies (Fig. 6). Here the gain in image quality of new digital PET scanners and especially of the recently introduced long-axis field of view PET scanner might favor the evaluation of vascular wall inflammation by PET, as it allows to acquire dynamic data that can provide more accurate quantification of the biological process, in this case [18F]FDG uptake rates in cell types involved in vascular wall inflammation. Secondly, it can assess the involvement of primary and secondary lymphoid organs throughout the whole body in systemic diseases versus tissue-confined local inflammatory process.

Fig. 6
figure 6

The extracted information about the vascular, lymphoid and hematopoietic system can be facilitated by AI to develop a patient-tailored prognostic score

Similarly, as defects in metabolism are commonly associated with impaired outcomes in various conditions, such as impaired regulation of glucose homeostasis in type 2 diabetes (Bernelot Moens et al. 2016; Lee et al. 2018; Hotamisligil 2017; Norata et al. 2015) or obesity in anticancer immunity (Thaiss et al. 2021; Ringel et al. 2020) and atherosclerosis (Bucerius et al. 2012, 2014), such prognostic role might be requested from [18F]FDG by clinical disciplines in the near future. As far as it concerns the immune system, this field is actively researched to develop therapeutic strategies (Mulder et al. 2019) to enhance or reduce inflammatory responses in anticancer immunity (Priem et al. 2020), autoimmune (Municio and Criado 2020) and infectious diseases (O'Neill and Netea 2020; Netea et al. 2020c). The higher sensitivity of the whole-body PET scanner enables the acquisition of low-dose PET images below 1–1.5 mSv, allowing its more frequent use in non-oncologic diseases for risk stratification assessment.

How should AI-based technologies be implemented to facilitate the development of [18F]FDG PET as a tool to determine prognostic immune metabolic profiles? The challenge in such task lies in the discrimination of bona fide inflammation from mala fide inflammation. This difference is expected to be subtle, as the inflammatory response in a distinct context of disease is beneficial rather than pathological, and temporally apart from an identified trigger.

Accurate appraisal of subtle differences requires large datasets for training and supporting ‘circumstantial evidence’ where possible. For example, it can be postulated that assessment of [18F]FDG uptake in the vascular wall as mala fide will be more accurate if not only metabolic activity in the hematopoietic system is taken into account, but also atherosclerotic calcifications and body composition in terms of subcutaneous and visceral adipose tissue versus muscle mass can be deduced from the low-dose CT (Laur et al. 2021) and incorporated in the risk assessment. Similarly, AI-algorithms are available to assess bone mineralization and emphysema score on low-dose CT (Ebrahimian et al. 2021), which could determine the host’ long-term responses to environmental inflictions like smoking. As for now, such information is not incorporated in current practice yet. Along with training AI-algorithms on large datasets come the need for harmonization, smart processing and modelling, each individual task is suited for AI-based technologies. In line with developments in CT imaging (Choe et al. 2019), AI-based algorithms can be trained to overcome center- or vendor-related differences in reconstruction settings (Arabi and Zaidi 2021) and allowing to extract radiomic features (Orlhac et al. 2018; Zwanenburg 2019).

Reiterating from the conceived potential of AI to transform healthcare (Denny and Collins 2021), contemplating whole body [18F]FDG PET images as huge interoperable datasets that meet the criteria of diversity and inclusion, implies that we need AI-technology to open up these big datasets and exploit its potential to approach immune metabolism on a systems level in clinical settings.

Challenges and potential solutions

Despite the sheer limitless methodological and technological advancements in AI-based technology, the widespread application of AI-tools in molecular imaging of infection and inflammation is facing some major challenges on its way into routine clinical use. One hurdle that will need to be overcome is to deal with the ‘black box’ stigma on AI-based algorithms; the lack of explainable correlations between in- and output leaves physicians often hesitant to rely on AI-based output. In addition, the subtle differences between physiological and pathological and potentially high variations between individual patients requires large datasets and/or labeled datasets based on ground truth, of which the latter is more difficult as it will require invasive procedures to obtain tissue samples to analyze immune cells’ metabolic profiles.

To address these issues, smaller studies with high translational design including flow cytometry, metabolomics and/or transcriptomic data from circulating immune cells or the hematopoietic system could provide proof-of-concept to correlate specific imaging findings to immune metabolic features in relevant cell populations (e.g., Hotchkiss et al. 2016). Subsequent studies can then build further upon these data and provide larger datasets for validation and to determine its value in real-life clinical setting. For such studies with large datasets, questions on harmonization of input data arise, which have partly been tackled in the EARL program for multicenter studies by the European Association for Nuclear Medicine. Moreover, computing higher order features from PET requires image normalization during data processing and training AI-based models on a wide range of scanner hardware can provide a solution that would be compatible with current deep-learning networks, provided that the computing power is sufficient. Nevertheless, the input for predictive or prognostic AI-based models as discussed in this communication should be ‘supervised’ as only PET parameters computed from predefined immune relevant organ systems, in line with current concepts on immune-metabolism, should serve as input data.

Another challenge will be the integration of the clinical experience from the nuclear medicine and radiology readers into a future AI-supported workflow of clinical decision making. The experience of the reader, who is also taking the case-specific clinical context into account, will be difficult to replace. Therefore at least in the coming few years, AI might support the clinical decisions if it is confirming the evaluation by the reading physician, but it is unclear how to proceed if human and AI-based assessment are coming to the contradictory results. In line with broader developments of AI-based technology in medical imaging, liability issues need to be addressed in the near future.

Conclusion

AI tools are increasingly used for a growing number of tasks in the imaging field ranging from technical applications which improve the sensitivity of scanners to biomedical applications for holistic data analysis. As proposed above, AI has the potential to improve the detection of inflammatory diseases and predict prognosis and outcome of patients under various immune-mediated conditions (Table 1). Furthermore, these tools are capable to provide a deeper understanding of the basic molecular mechanisms of inflammatory diseases.

Table 1 Key elements for the future: opportunities, challenges and solutions of AI in infection and inflammation molecular imaging

For a successful application in future health care in the context of personalized medicine the tight integration of the AI imaging tools with other diagnostic methods like genetic analysis, proteomics and metabolomics is the key to achieve reliable and impactful data which improves treatment decisions and ultimately patients’ well-being and survival.

Availability of data and materials

None.

Abbreviations

AI:

Artificial intelligence

CT:

Computed tomography

[18F]FDG:

2-[18F]fluoro-2-deoxy-D-glucose

LOR:

Line-of-response

MRI:

Magnetic resonance imaging

PET:

Positron emission tomography

SPECT:

Single-photon emission tomography

SUV:

Standardized uptake value

References

  • Aerts HJ, Velazquez ER, Leijenaar RT, Parmar C, Grossmann P, Carvalho S et al (2014) Decoding tumour phenotype by noninvasive imaging using a quantitative radiomics approach. Nat Commun 5:4006

    CAS  PubMed  Article  Google Scholar 

  • Alberts I, Hunermund JN, Prenosil G, Mingels C, Bohn KP, Viscione M et al (2021) Clinical performance of long axial field of view PET/CT: a head-to-head intra-individual comparison of the Biograph Vision Quadra with the Biograph Vision PET/CT. Eur J Nucl Med Mol Imaging 48(8):2395–2404

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  • Arabi H, Zaidi H (2021) Non-local mean denoising using multiple PET reconstructions. Ann Nucl Med 35(2):176–186

    PubMed  Article  Google Scholar 

  • Arts RJ, Gresnigt MS, Joosten LA, Netea MG (2017) Cellular metabolism of myeloid cells in sepsis. J Leukoc Biol 101(1):151–164

    CAS  PubMed  Article  Google Scholar 

  • Arts RJW, Moorlag S, Novakovic B, Li Y, Wang SY, Oosting M et al (2018) BCG vaccination protects against experimental viral infection in humans through the induction of cytokines associated with trained immunity. Cell Host Microbe 23(1):89-100.e5

    CAS  PubMed  Article  Google Scholar 

  • Badawi RD, Shi H, Hu P, Chen S, Xu T, Price PM et al (2019) First human imaging studies with the EXPLORER total-body PET scanner. J Nucl Med 60(3):299–303

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  • Bekkering S, Blok BA, Joosten LA, Riksen NP, van Crevel R, Netea MG (2016a) In vitro experimental model of trained innate immunity in human primary monocytes. Clin Vaccine Immunol 23(12):926–933

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  • Bekkering S, van den Munckhof I, Nielen T, Lamfers E, Dinarello C, Rutten J et al (2016b) Innate immune cell activation and epigenetic remodeling in symptomatic and asymptomatic atherosclerosis in humans in vivo. Atherosclerosis 254:228–236

    CAS  PubMed  Article  Google Scholar 

  • Bekkering S, Arts RJW, Novakovic B, Kourtzelis I, van der Heijden C, Li Y et al (2018) Metabolic induction of trained immunity through the mevalonate pathway. Cell 172(1–2):135–46.e9

    CAS  PubMed  Article  Google Scholar 

  • Bekkering S, Stiekema LCA, Bernelot Moens S, Verweij SL, Novakovic B, Prange K et al (2019) Treatment with statins does not revert trained immunity in patients with familial hypercholesterolemia. Cell Metab 30(1):1–2

    CAS  PubMed  Article  Google Scholar 

  • Bergeron M, Cadorette J, Tetrault MA, Beaudoin JF, Leroux JD, Fontaine R et al (2014) Imaging performance of LabPET APD-based digital PET scanners for pre-clinical research. Phys Med Biol 59(3):661–678

    PubMed  Article  Google Scholar 

  • Bernelot Moens SJ, Stoekenbroek RM, van der Valk FM, Verweij SL, Koelemay MJ, Verberne HJ et al (2016) Carotid arterial wall inflammation in peripheral artery disease is augmented by type 2 diabetes: a cross-sectional study. BMC Cardiovasc Disord 16(1):237

    PubMed  PubMed Central  Article  CAS  Google Scholar 

  • Berrevoets MAH, Kouijzer IJE, Slieker K, Aarntzen E, Kullberg BJ, Oever JT et al (2019) (18)F-FDG PET/CT-guided treatment duration in patients with high-risk Staphylococcus aureus bacteremia: a proof of principle. J Nucl Med 60(7):998–1002

    CAS  PubMed  Article  Google Scholar 

  • Bucerius J, Mani V, Moncrieff C, Rudd JH, Machac J, Fuster V et al (2012) Impact of noninsulin-dependent type 2 diabetes on carotid wall 18F-fluorodeoxyglucose positron emission tomography uptake. J Am Coll Cardiol 59(23):2080–2088

    PubMed  PubMed Central  Article  Google Scholar 

  • Bucerius J, Mani V, Wong S, Moncrieff C, Izquierdo-Garcia D, Machac J et al (2014) Arterial and fat tissue inflammation are highly correlated: a prospective 18F-FDG PET/CT study. Eur J Nucl Med Mol Imaging 41(5):934–945

    PubMed  PubMed Central  Article  Google Scholar 

  • Buther F, Vehren T, Schafers KP, Schafers M (2016) Impact of data-driven respiratory gating in clinical PET. Radiology 281(1):229–238

    PubMed  Article  Google Scholar 

  • Camacho DM, Collins KM, Powers RK, Costello JC, Collins JJ (2018) Next-generation machine learning for biological networks. Cell 173(7):1581–1592

    CAS  PubMed  Article  Google Scholar 

  • Chakfe N, Diener H, Lejay A, Assadian O, Berard X, Caillon J et al (2020) Editor’s Choice—European Society for Vascular Surgery (ESVS) 2020 clinical practice guidelines on the management of vascular graft and endograft infections. Eur J Vasc Endovasc Surg 59(3):339–384

    PubMed  Article  Google Scholar 

  • Chavakis T, Mitroulis I, Hajishengallis G (2019) Hematopoietic progenitor cells as integrative hubs for adaptation to and fine-tuning of inflammation. Nat Immunol 20(7):802–811

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  • Cheng SC, Scicluna BP, Arts RJ, Gresnigt MS, Lachmandas E, Giamarellos-Bourboulis EJ et al (2016) Broad defects in the energy metabolism of leukocytes underlie immunoparalysis in sepsis. Nat Immunol 17(4):406–413

    CAS  PubMed  Article  Google Scholar 

  • Chiossone L, Dumas PY, Vienne M, Vivier E (2018) Natural killer cells and other innate lymphoid cells in cancer. Nat Rev Immunol 18(11):671–688

    CAS  PubMed  Article  Google Scholar 

  • Chirillo F (2021) New approach to managing infective endocarditis. Trends Cardiovasc Med 31(5):277–286

    PubMed  Article  Google Scholar 

  • Choe J, Lee SM, Do KH, Lee G, Lee JG, Lee SM et al (2019) Deep learning-based image conversion of CT reconstruction kernels improves radiomics reproducibility for pulmonary nodules or masses. Radiology 292(2):365–373

    PubMed  Article  Google Scholar 

  • Colling R, Pitman H, Oien K, Rajpoot N, Macklin P, Group CM-PAiHW et al (2019) Artificial intelligence in digital pathology: a roadmap to routine use in clinical practice. J Pathol 249(2):143–150

    PubMed  Article  Google Scholar 

  • Cui J, Gong K, Guo N, Wu C, Meng X, Kim K et al (2019) PET image denoising using unsupervised deep learning. Eur J Nucl Med Mol Imaging 46(13):2780–2789

    PubMed  PubMed Central  Article  Google Scholar 

  • Currie G, Hawk KE (2021) Ethical and legal challenges of artificial intelligence in nuclear medicine. Semin Nucl Med 51(2):120–125

    PubMed  Article  Google Scholar 

  • Currie G, Rohren E (2021) Intelligent imaging in nuclear medicine: the principles of artificial intelligence, machine learning and deep learning. Semin Nucl Med 51(2):102–111

    PubMed  Article  Google Scholar 

  • de Vos BD, Berendsen FF, Viergever MA, Sokooti H, Staring M, Isgum I (2019) A deep learning framework for unsupervised affine and deformable image registration. Med Image Anal 52:128–143

    PubMed  Article  Google Scholar 

  • de Vries EF, Roca M, Jamar F, Israel O, Signore A (2010) Guidelines for the labelling of leucocytes with (99m)Tc-HMPAO Inflammation/Infection. Taskgroup of the European Association of Nuclear Medicine. Eur J Nucl Med Mol Imaging 37(4):842–848

    PubMed  PubMed Central  Article  Google Scholar 

  • Decuyper M, Maebe J, Van Holen R, Vandenberghe S (2021) Artificial intelligence with deep learning in nuclear medicine and radiology. EJNMMI Phys 8(1):81

    PubMed  PubMed Central  Article  Google Scholar 

  • Denny JC, Collins FS (2021) Precision medicine in 2030-seven ways to transform healthcare. Cell 184(6):1415–1419

    CAS  PubMed  Article  Google Scholar 

  • Dietze MMA, Branderhorst W, Kunnen B, Viergever MA, de Jong H (2019) Accelerated SPECT image reconstruction with FBP and an image enhancement convolutional neural network. EJNMMI Phys 6(1):14

    PubMed  PubMed Central  Article  Google Scholar 

  • Dominguez-Andres J, Netea MG (2019) Long-term reprogramming of the innate immune system. J Leukoc Biol 105(2):329–338

    CAS  PubMed  Article  Google Scholar 

  • Ebrahimian S, Digumarthy S, Bizzo B, Primak A, Zimmermann M, Tarbiah MM et al (2021) Artificial intelligence has similar performance to subjective assessment of emphysema severity on chest CT. Acad Radiol. https://doi.org/10.1016/j.acra.2021.09.007

    Article  PubMed  Google Scholar 

  • Emami H, Singh P, MacNabb M, Vucic E, Lavender Z, Rudd JH et al (2015) Splenic metabolic activity predicts risk of future cardiovascular events: demonstration of a cardiosplenic axis in humans. JACC Cardiovasc Imaging 8(2):121–130

    PubMed  PubMed Central  Article  Google Scholar 

  • Feng T, Wang J, Dong Y, Zhao J, Li H (2019) A novel data-driven cardiac gating signal extraction method for PET. IEEE Trans Med Imaging 38(2):629–637

    PubMed  Article  Google Scholar 

  • Filippi L, Schillaci O (2022) Total-body [(18)F]FDG PET/CT scan has stepped into the arena: the faster, the better. Is it always true? Eur J Nucl Med Mol Imaging. https://doi.org/10.1007/s00259-022-05791-z

    Article  PubMed  Google Scholar 

  • Flint TR, Fearon DT, Janowitz T (2017) Connecting the metabolic and immune responses to cancer. Trends Mol Med 23(5):451–464

    CAS  PubMed  Article  Google Scholar 

  • Gaber T, Strehl C, Buttgereit F (2017) Metabolic regulation of inflammation. Nat Rev Rheumatol 13(5):267–279

    PubMed  Article  Google Scholar 

  • Gillies RJ, Kinahan PE, Hricak H (2016) Radiomics: images are more than pictures, they are data. Radiology 278(2):563–577

    PubMed  Article  Google Scholar 

  • Gong K, Guan J, Liu CC, Qi J (2019) PET image denoising using a deep neural network through fine tuning. IEEE Trans Radiat Plasma Med Sci 3(2):153–161

    PubMed  Article  Google Scholar 

  • Habib G, Lancellotti P, Antunes MJ, Bongiorni MG, Casalta JP, Del Zotti F et al (2015) ESC Guidelines for the management of infective endocarditis: the Task Force for the Management of Infective Endocarditis of the European Society of Cardiology (ESC). Endorsed by: European Association for Cardio-Thoracic Surgery (EACTS), the European Association of Nuclear Medicine (EANM). Eur Heart J 36(44):3075–3128

    PubMed  Article  Google Scholar 

  • Haggstrom I, Schmidtlein CR, Campanella G, Fuchs TJ (2019) DeepPET: a deep encoder-decoder network for directly solving the PET image reconstruction inverse problem. Med Image Anal 54:253–262

    PubMed  PubMed Central  Article  Google Scholar 

  • Hatt M, Cheze Le Rest C, Antonorsi N, Tixier F, Tankyevych O, Jaouen V et al (2021) Radiomics in PET/CT: current status and future AI-based evolutions. Semin Nucl Med 51(2):126–133

    PubMed  Article  Google Scholar 

  • He J, Baxter SL, Xu J, Xu J, Zhou X, Zhang K (2019) The practical implementation of artificial intelligence technologies in medicine. Nat Med 25(1):30–36

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  • He W, Wang Y, Liang X, Zhou W, Zhu M, Han X et al (2021) High-performance coded aperture gamma camera based on monolithic GAGG: Ce crystal. Rev Sci Instrum 92(1):013106

    CAS  PubMed  Article  Google Scholar 

  • Hipfl C, Mooij W, Perka C, Hardt S, Wassilew GI (2021) Unexpected low-grade infections in revision hip arthroplasty for aseptic loosening : a single-institution experience of 274 hips. Bone Joint J 103-B(6):1070–1077

    PubMed  Article  Google Scholar 

  • Hosny A, Parmar C, Quackenbush J, Schwartz LH, Aerts H (2018) Artificial intelligence in radiology. Nat Rev Cancer 18(8):500–510

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  • Hotamisligil GS (2017) Inflammation, metaflammation and immunometabolic disorders. Nature 542(7640):177–185

    CAS  PubMed  Article  Google Scholar 

  • Hotchkiss RS, Monneret G, Payen D (2013) Sepsis-induced immunosuppression: from cellular dysfunctions to immunotherapy. Nat Rev Immunol 13(12):862–874

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  • Hotchkiss RS, Moldawer LL, Opal SM, Reinhart K, Turnbull IR, Vincent JL (2016) Sepsis and septic shock. Nat Rev Dis Primers 2:16045

    PubMed  PubMed Central  Article  Google Scholar 

  • Jaltotage B, Ali U, Dorai-Raj A, Rankin J, Sanfilippo F, Dwivedi G (2021) Q fever endocarditis: a review of local and all reported cases in the literature. Heart Lung Circ 30(10):1509–1515

    PubMed  Article  Google Scholar 

  • Jamar F, Buscombe J, Chiti A, Christian PE, Delbeke D, Donohoe KJ et al (2013) EANM/SNMMI guideline for 18F-FDG use in inflammation and infection. J Nucl Med 54(4):647–658

    PubMed  Article  Google Scholar 

  • Joseph P, Ishai A, Mani V, Kallend D, Rudd JH, Fayad ZA et al (2017) Short-term changes in arterial inflammation predict long-term changes in atherosclerosis progression. Eur J Nucl Med Mol Imaging 44(1):141–150

    CAS  PubMed  Article  Google Scholar 

  • Kalafati L, Kourtzelis I, Schulte-Schrepping J, Li X, Hatzioannou A, Grinenko T et al (2020) Innate immune training of granulopoiesis promotes anti-tumor activity. Cell 183(3):771-8512.e12

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  • Kaplan S, Zhu YM (2019) Full-dose PET image estimation from low-dose PET image using deep learning: a pilot study. J Digit Imaging 32(5):773–778

    PubMed  Article  Google Scholar 

  • Kaufmann SHE, Dorhoi A, Hotchkiss RS, Bartenschlager R (2018) Host-directed therapies for bacterial and viral infections. Nat Rev Drug Discov 17(1):35–56

    CAS  PubMed  Article  Google Scholar 

  • Kidd BA, Peters LA, Schadt EE, Dudley JT (2014) Unifying immunology with informatics and multiscale biology. Nat Immunol 15(2):118–127

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  • Kouijzer IJE, van Leerdam EJ, Gompelman M, Tuinte RAM, Aarntzen E, Berrevoets MAH et al (2021) Intravenous to oral switch in complicated Staphylococcus aureus bacteremia without endovascular infection: a retrospective single-center cohort study. Clin Infect Dis 73(5):895–898

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  • Laohapensang K, Arworn S, Orrapin S, Reanpang T, Orrapin S (2017) Management of the infected aortic endograft. Semin Vasc Surg 30(2–3):91–94

    PubMed  Article  Google Scholar 

  • Laur O, Weaver MJ, Bridge C, Chow E, Rosenthal M, Bay C et al (2021) Computed tomography-based body composition profile as a screening tool for geriatric frailty detection. Skeletal Radiol. https://doi.org/10.1007/s00256-021-03951-0

    Article  PubMed  PubMed Central  Google Scholar 

  • LeCun Y, Bengio Y, Hinton G (2015) Deep learning. Nature 521(7553):436–444

    CAS  PubMed  Article  Google Scholar 

  • Lee YS, Wollam J, Olefsky JM (2018) An integrated view of immunometabolism. Cell 172(1–2):22–40

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  • Leentjens J, Bekkering S, Joosten LAB, Netea MG, Burgner DP, Riksen NP (2018) Trained innate immunity as a novel mechanism linking infection and the development of atherosclerosis. Circ Res 122(5):664–669

    CAS  PubMed  Article  Google Scholar 

  • Lempiainen H, Braenne I, Michoel T, Tragante V, Vilne B, Webb TR et al (2018) Network analysis of coronary artery disease risk genes elucidates disease mechanisms and druggable targets. Sci Rep 8(1):3434

    PubMed  PubMed Central  Article  CAS  Google Scholar 

  • Lercher A, Baazim H, Bergthaler A (2020) Systemic immunometabolism: challenges and opportunities. Immunity 53(3):496–509

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  • Libby P, Buring JE, Badimon L, Hansson GK, Deanfield J, Bittencourt MS et al (2019) Atherosclerosis. Nat Rev Dis Primers 5(1):56

    PubMed  Article  Google Scholar 

  • Lopez-Otin C, Blasco MA, Partridge L, Serrano M, Kroemer G (2013) The hallmarks of aging. Cell 153(6):1194–1217

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  • Makinen VP, Civelek M, Meng Q, Zhang B, Zhu J, Levian C et al (2014) Integrative genomics reveals novel molecular pathways and gene networks for coronary artery disease. PLoS Genet 10(7):e1004502

    PubMed  PubMed Central  Article  CAS  Google Scholar 

  • Meijering E, Carpenter AE, Peng H, Hamprecht FA, Olivo-Marin JC (2016) Imagining the future of bioimage analysis. Nat Biotechnol 34(12):1250–1255

    CAS  PubMed  Article  Google Scholar 

  • Minarik D, Enqvist O, Tragardh E (2020) Denoising of scintillation camera images using a deep convolutional neural network: a Monte Carlo simulation approach. J Nucl Med 61(2):298–303

    PubMed  PubMed Central  Article  Google Scholar 

  • Mulder WJM, Ochando J, Joosten LAB, Fayad ZA, Netea MG (2019) Therapeutic targeting of trained immunity. Nat Rev Drug Discov 18(7):553–566

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  • Municio C, Criado G (2020) Therapies targeting trained immune cells in inflammatory and autoimmune diseases. Front Immunol 11:631743

    CAS  PubMed  Article  Google Scholar 

  • Netea MG, Joosten LAB (2018) Trained immunity and local innate immune memory in the lung. Cell 175(6):1463–1465

    CAS  PubMed  Article  Google Scholar 

  • Netea MG, Dominguez-Andres J, Barreiro LB, Chavakis T, Divangahi M, Fuchs E et al (2020a) Defining trained immunity and its role in health and disease. Nat Rev Immunol 20:375–388

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  • Netea MG, Dominguez-Andres J, Barreiro LB, Chavakis T, Divangahi M, Fuchs E et al (2020b) Defining trained immunity and its role in health and disease. Nat Rev Immunol 20(6):375–388

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  • Netea MG, Giamarellos-Bourboulis EJ, Dominguez-Andres J, Curtis N, van Crevel R, van de Veerdonk FL et al (2020c) Trained immunity: a tool for reducing susceptibility to and the severity of SARS-CoV-2 infection. Cell 181(5):969–977

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  • Norata GD, Caligiuri G, Chavakis T, Matarese G, Netea MG, Nicoletti A et al (2015) The cellular and molecular basis of translational immunometabolism. Immunity 43(3):421–434

    CAS  PubMed  Article  Google Scholar 

  • Noz MP, Bekkering S, Groh L, Nielen TM, Lamfers EJ, Schlitzer A et al (2020) Reprogramming of bone marrow myeloid progenitor cells in patients with severe coronary artery disease. Elife. https://doi.org/10.7554/eLife.60939

    Article  PubMed  PubMed Central  Google Scholar 

  • O’Neill LAJ, Netea MG (2020) BCG-induced trained immunity: can it offer protection against COVID-19? Nat Rev Immunol 20(6):335–337

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  • Orlhac F, Boughdad S, Philippe C, Stalla-Bourdillon H, Nioche C, Champion L et al (2018) A postreconstruction harmonization method for multicenter radiomic studies in PET. J Nucl Med 59(8):1321–1328

    CAS  PubMed  Article  Google Scholar 

  • Parmar C, Barry JD, Hosny A, Quackenbush J, Aerts H (2018) Data analysis strategies in medical imaging. Clin Cancer Res 24(15):3492–3499

    PubMed  PubMed Central  Article  Google Scholar 

  • Pober JS, Sessa WC (2007) Evolving functions of endothelial cells in inflammation. Nat Rev Immunol 7(10):803–815

    CAS  PubMed  Article  Google Scholar 

  • Priem B, van Leent MMT, Teunissen AJP, Sofias AM, Mourits VP, Willemsen L et al (2020) Trained immunity-promoting nanobiologic therapy suppresses tumor growth and potentiates checkpoint inhibition. Cell 183(3):786-801.e19

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  • Ringel AE, Drijvers JM, Baker GJ, Catozzi A, Garcia-Canaveras JC, Gassaway BM et al (2020) Obesity shapes metabolism in the tumor microenvironment to suppress anti-tumor immunity. Cell 183(7):1848–66.e26

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  • Roca M, de Vries EF, Jamar F, Israel O, Signore A (2010) Guidelines for the labelling of leucocytes with (111)In-oxine Inflammation/Infection. Taskgroup of the European Association of Nuclear Medicine. Eur J Nucl Med Mol Imaging 37(4):835–841

    PubMed  PubMed Central  Article  Google Scholar 

  • Rubeaux M, Joshi NV, Dweck MR, Fletcher A, Motwani M, Thomson LE et al (2016) Motion correction of 18F-NaF PET for imaging coronary atherosclerotic plaques. J Nucl Med 57(1):54–59

    CAS  PubMed  Article  Google Scholar 

  • Saeedan MB, Wang TKM, Cremer P, Wahadat AR, Budde RPJ, Unai S et al (2021) Role of cardiac CT in infective endocarditis: current evidence, opportunities, and challenges. Radiol Cardiothorac Imaging 3(1):e200378

    PubMed  PubMed Central  Article  Google Scholar 

  • Schleyer PJ, O’Doherty MJ, Marsden PK (2011) Extension of a data-driven gating technique to 3D, whole body PET studies. Phys Med Biol 56(13):3953–3965

    PubMed  Article  Google Scholar 

  • Schultze JL, consortium S, Rosenstiel P (2018) Systems medicine in chronic inflammatory diseases. Immunity 48(4):608–613

    CAS  PubMed  Article  Google Scholar 

  • Schwenck J, Schorg B, Fiz F, Sonanini D, Forschner A, Eigentler T et al (2020) Cancer immunotherapy is accompanied by distinct metabolic patterns in primary and secondary lymphoid organs observed by non-invasive in vivo (18)F-FDG-PET. Theranostics 10(2):925–937

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  • Seban RD, Champion L, Muneer I, Synn S, Schwartz LH, Dercle L (2021) Potential theranostic role of bone marrow glucose metabolism on baseline [18F]-FDG PET/CT in metastatic melanoma. J Nucl Med. https://doi.org/10.2967/jnumed.121.262361

    Article  PubMed  Google Scholar 

  • Seith F, Forschner A, Weide B, Guckel B, Schwartz M, Schwenck J et al (2020) Is there a link between very early changes of primary and secondary lymphoid organs in (18)F-FDG-PET/MRI and treatment response to checkpoint inhibitor therapy? J Immunother Cancer. https://doi.org/10.1136/jitc-2020-000656

    Article  PubMed  PubMed Central  Google Scholar 

  • Shu L, Chan KHK, Zhang G, Huan T, Kurt Z, Zhao Y et al (2017) Shared genetic regulatory networks for cardiovascular disease and type 2 diabetes in multiple populations of diverse ethnicities in the United States. PLoS Genet 13(9):e1007040

    PubMed  PubMed Central  Article  CAS  Google Scholar 

  • Signore A, Sconfienza LM, Borens O, Glaudemans A, Cassar-Pullicino V, Trampuz A et al (2019) Consensus document for the diagnosis of prosthetic joint infections: a joint paper by the EANM, EBJIS, and ESR (with ESCMID endorsement). Eur J Nucl Med Mol Imaging 46(4):971–988

    PubMed  PubMed Central  Article  Google Scholar 

  • Slart R, Glaudemans A, Lancellotti P, Hyafil F, Blankstein R, Schwartz RG et al (2018a) A joint procedural position statement on imaging in cardiac sarcoidosis: from the Cardiovascular and Inflammation & Infection Committees of the European Association of Nuclear Medicine, the European Association of Cardiovascular Imaging, and the American Society of Nuclear Cardiology. J Nucl Cardiol 25(1):298–319

    PubMed  Article  Google Scholar 

  • Slart R, Writing g, Reviewer g, Members of EC, Members of EI, Inflammation et al (2018b) FDG-PET/CT(A) imaging in large vessel vasculitis and polymyalgia rheumatica: joint procedural recommendation of the EANM, SNMMI, and the PET Interest Group (PIG), and endorsed by the ASNC. Eur J Nucl Med Mol Imaging 45(7):1250–1269

    PubMed  PubMed Central  Article  Google Scholar 

  • Slart R, Williams MC, Juarez-Orozco LE, Rischpler C, Dweck MR, Glaudemans A et al (2021) Position paper of the EACVI and EANM on artificial intelligence applications in multimodality cardiovascular imaging using SPECT/CT, PET/CT, and cardiac CT. Eur J Nucl Med Mol Imaging 48(5):1399–1413

    PubMed  PubMed Central  Article  Google Scholar 

  • Spitzer MH, Carmi Y, Reticker-Flynn NE, Kwek SS, Madhireddy D, Martins MM et al (2017) Systemic immunity is required for effective cancer immunotherapy. Cell 168(3):487-502.e15

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  • Stiekema LCA, Stroes ESG, Verweij SL, Kassahun H, Chen L, Wasserman SM et al (2019) Persistent arterial wall inflammation in patients with elevated lipoprotein(a) despite strong low-density lipoprotein cholesterol reduction by proprotein convertase subtilisin/kexin type 9 antibody treatment. Eur Heart J 40(33):2775–2781

    CAS  PubMed  Article  Google Scholar 

  • Sugimoto MA, Vago JP, Perretti M, Teixeira MM (2019) Mediators of the resolution of the inflammatory response. Trends Immunol 40(3):212–227

    CAS  PubMed  Article  Google Scholar 

  • Talha KM, DeSimone DC, Sohail MR, Baddour LM (2020) Pathogen influence on epidemiology, diagnostic evaluation and management of infective endocarditis. Heart 106(24):1878–1882

    PubMed  Article  Google Scholar 

  • Ten Hove D, Slart R, Sinha B, Glaudemans A, Budde RPJ (2021) (18)F-FDG PET/CT in infective endocarditis: indications and approaches for standardization. Curr Cardiol Rep 23(9):130

    PubMed  PubMed Central  Article  Google Scholar 

  • Thaiss WM, Gatidis S, Sartorius T, Machann J, Peter A, Eigentler TK et al (2021) Noninvasive, longitudinal imaging-based analysis of body adipose tissue and water composition in a melanoma mouse model and in immune checkpoint inhibitor-treated metastatic melanoma patients. Cancer Immunol Immunother 70(5):1263–1275

    CAS  PubMed  Article  Google Scholar 

  • Ungar B, Pavel AB, Robson PM, Kaufman A, Pruzan A, Brunner P et al (2020) A preliminary (18)F-FDG-PET/MRI study shows increased vascular inflammation in moderate-to-severe atopic dermatitis. J Allergy Clin Immunol Pract 8:3500–3506

    PubMed  PubMed Central  Article  Google Scholar 

  • Uribe CF, Mathotaarachchi S, Gaudet V, Smith KC, Rosa-Neto P, Benard F et al (2019) Machine learning in nuclear medicine: part 1—introduction. J Nucl Med 60(4):451–458

    PubMed  Article  Google Scholar 

  • van der Heijden C, Smeets EMM, Aarntzen E, Noz MP, Monajemi H, Kersten S et al (2020) Arterial wall inflammation and increased hematopoietic activity in patients with primary aldosteronism. J Clin Endocrinol Metab. https://doi.org/10.1210/clinem/dgz306

    Article  PubMed  PubMed Central  Google Scholar 

  • van der Laak J, Litjens G, Ciompi F (2021) Deep learning in histopathology: the path to the clinic. Nat Med 27(5):775–784

    PubMed  Article  CAS  Google Scholar 

  • van der Valk FM, Verweij SL, Zwinderman KA, Strang AC, Kaiser Y, Marquering HA et al (2016) Thresholds for arterial wall inflammation quantified by (18)F-FDG PET imaging: implications for vascular interventional studies. JACC Cardiovasc Imaging 9(10):1198–1207

    PubMed  PubMed Central  Article  Google Scholar 

  • van der Valk FM, Kuijk C, Verweij SL, Stiekema LCA, Kaiser Y, Zeerleder S et al (2017) Increased haematopoietic activity in patients with atherosclerosis. Eur Heart J 38(6):425–432

    PubMed  Google Scholar 

  • Yang X (2020) Multitissue multiomics systems biology to dissect complex diseases. Trends Mol Med 26(8):718–728

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  • Yang J, Sohn JH, Behr SC, Gullberg GT, Seo Y (2021) CT-less direct correction of attenuation and scatter in the image space using deep learning for whole-body FDG PET: potential benefits and pitfalls. Radiol Artif Intell 3(2):e200137

    PubMed  Article  Google Scholar 

  • Zatcepin A, Pizzichemi M, Polesel A, Paganoni M, Auffray E, Ziegler SI et al (2020) Improving depth-of-interaction resolution in pixellated PET detectors using neural networks. Phys Med Biol 65(17):175017

    CAS  PubMed  Article  Google Scholar 

  • Zukotynski K, Gaudet V, Uribe CF, Mathotaarachchi S, Smith KC, Rosa-Neto P et al (2021) Machine learning in nuclear medicine: part 2—neural networks and clinical aspects. J Nucl Med 62(1):22–29

    PubMed  Article  Google Scholar 

  • Zwanenburg A (2019) Radiomics in nuclear medicine: robustness, reproducibility, standardization, and how to avoid data analysis traps and replication crisis. Eur J Nucl Med Mol Imaging 46(13):2638–2655

    PubMed  Article  Google Scholar 

Download references

Funding

N.P.R. is funded by an IN-CONTROL CVON grant (CVON2018-27) of the Netherlands Heart Foundation, and a JTC2018 grant (‘MEMORY’) from the European Research Area Network on Cardiovascular Disease (ERA-CVD).

Author information

Authors and Affiliations

Authors

Contributions

EA developed the overall concept of the review. EA and JS wrote the manuscript. MK, NR, CF, UM and RS critically read and amended the manuscript. All authors read and approved the final manuscript.

Corresponding author

Correspondence to Johannes Schwenck.

Ethics declarations

Ethics approval and consent to participate

None.

Consent for publication

None.

Competing interests

The authors declare that they have no competing interests.

Additional information

Publisher's Note

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

Rights and permissions

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/.

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Schwenck, J., Kneilling, M., Riksen, N.P. et al. A role for artificial intelligence in molecular imaging of infection and inflammation. European J Hybrid Imaging 6, 17 (2022). https://doi.org/10.1186/s41824-022-00138-1

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/s41824-022-00138-1