The Human Olfactory Dysfunction

Having received relatively little attention over the last 100 years, recent evidence suggests that the human sense of smell is not a negligible entity (McGann 2017), and plays an important role in food ingestion, harm avoidance, and social communication (Hummel et al. 2007; Stevenson 2010). The loss of olfactory functions can lead to a drastic change in the enjoyment of eating, changes in body weight, changes in daily activities, sexuality and disturbances in affective behavior, and, hence, a reduced quality of life (Miwa et al. 2001; Santos et al. 2004; Frasnelli and Hummel 2005; Croy et al. 2014). Population-based studies of olfactory loss indicate a prevalence of 22% (25 to 75 years) (Vennemann et al. 2008), 19% (≥ 20 years) (Bramerson et al. 2004), or 24% (≥ 53 years) (Murphy et al. 2002), with highest prevalence in older men.

Olfactory loss is characterized as a reduced ability to smell and to detect odors. Clinically, people with olfactory loss are classified as hyposmia (the sense of smell is partially reduced) or functional anosmia (the sense of smell is reduced to the extent that is not useful in daily life) (Hummel et al. 2016). Besides, approximately 1 in 8000 people are born without a sense of smell which is the congenital anosmia. Apart from age-related olfactory function decline, the major causes for olfactory loss include sinonasal disease, traumatic brain injury, upper respiratory tract infection, congenital anosmia, or idiopathic causes (Temmel et al. 2002). Moreover, olfactory loss is among the early symptoms of Parkinson’s disease (Hawkes et al. 1997; Haehner et al. 2007; Ross et al. 2008), Alzheimer’s disease (Marine and Boriana 2014), incipient dementia schizophrenia (Moberg et al. 1999), or depression (Croy and Hummel 2017). Hence, olfactory dysfunction is associated with multiple neurological disorders that may affect the overall health of the population.

The Central Olfactory System

Odor perception requires a functional peripheral sensory organ and central pathways. Odor molecules first arrive at the nasal epithelium and make contact with the sensory endings of olfactory receptor neurons. Axonal projections are then conveyed with a relay in the olfactory bulbs (OB) to the cortex and terminate in several areas in the basal frontal and medial temporal lobes, including the piriform cortex, peri and anterior amygdaloid cortex, and rostral entorhinal cortex (Gottfried 2010; Van Hartevelt and Kringelbach 2011). These cortical regions that receive direct input from the OB are collectively referred to as the ‘primary olfactory cortex’. The cortical targets of the primary olfactory cortex include various brain regions in the limbic lobe and frontal cortex, including the additional amygdala subnuclei, orbitofrontal cortex (OFC), hippocampus, parahippocampal gyrus, cingulate cortex, insula, striatum hypothalamus, and the mediodorsal thalamus (Gottfried and Zald 2005; Fjaeldstad et al. 2017; Zhou et al. 2019). Collectively, these areas are commonly referred to as the secondary olfactory cortex (Fig. 1).

Fig. 1
figure 1

The human central olfactory system

Full size image

Olfactory dysfunctions led to a significant decreased quality of life (Croy et al. 2014) as well as structural and functional alterations of the brain (Reichert and Schopf 2017). The magnetic resonance imaging (MRI), providing a non-invasive approach to investigate the structural and functional features of the working brain, has been applied in research on human olfaction and olfactory dysfunctions. The present review will provide a summary of the current state of MRI research on functional and structural changes in the central olfactory system due to olfactory dysfunction and will discuss the potential of MRI-based investigations in clinical routine.

Functional Changes After Olfactory Loss

Odor-Induced Brain Responses in Olfactory Loss

Chemosensory odor perception is partly the result of the interaction between the olfactory and the trigeminal systems. Whereas the olfactory system mediates the quality percept of an odor, the trigeminal system conveys sensations such as burning, pungency, or stinging, as well as touch, pressure, and temperature (Albrecht et al. 2010). Using MRI, a number of studies has investigated disrupted brain responses to odors in patients with olfactory loss. Compared to healthy controls, patients demonstrate widespread decrease of odor-induced brain activation in the olfactory related regions, including the piriform cortex, amygdala, OFC, insula, and anterior cingulate cortex (Levy et al. 1998; Levy et al. 1999a, b; Pellegrino et al. 2016; Han et al. 2018a, b; Moon et al. 2018). Due to intimate connections between the olfactory and trigeminal systems, the olfactory loss is also linked to alterations for cerebral processing of trigeminal or bimodal olfactory stimulation. In two fMRI studies, patients with olfactory loss exhibited reduced activation in the primary motor area and cerebellum in response to either trigeminal or bimodal odor stimulations (Iannilli et al. 2007, 2011).

Attenuated odor-induced brain responses were also seen in patients with neurodegenerative disorders, including Parkinson’s disease (Hummel et al. 2010, Takeda et al. 2010), Alzheimer’s disease (Vasavada et al. 2015, 2017) or Autism Spectrum Disorder (Koehler et al. 2018). Hummel et al. (2010) found different patterns for processing pleasant or unpleasant odors among patients with Parkinson’s disease. Li et al. (2010) demonstrated that patients with early-stage Alzheimer’s disease exhibited a blunted cross-adaptation response in the posterior piriform cortex after the sequential presentation of qualitatively similar (versus qualitatively different) odor pairs. This finding reflected a functional disruption of odor quality coding in this olfactory brain area, which may provide evidence for perceptual deficits in odor identification among those patients. Taken together, olfactory loss affects the central olfactory system and led to a blunted neural response to odor stimulation. However, the patients included in most previous studies were heterogeneous in terms of the cause for olfactory loss, which limited the detection of specific differences between groups. Moreover, few fMRI studies had been done among patients with olfactory loss focusing on olfactory sub-functions during odor perception, such as odor discrimination or identification.

Olfactory Loss and Top-Down Cognitive Modulation

Olfactory perception is modulated by higher levels of cognitive processing, such as memory or learning, which is known as the olfactory “top-down” modulation (Rolls 2011). Previous studies have shown the involvement of multiple olfactory brain regions in the olfactory “top-down” processing. For example, regions activated during  odor perception such as the piriform cortex, OFC, and insula were also seen activation during olfactory imagery (Bensafi et al. 2003; Djordjevic et al. 2005; Bensafi et al. 2007) or odor expectation (Zelano et al. 2011). Besides, non-odorous but odor-related stimuli such as words (Gonzalez et al. 2006; Arshamian et al. 2013), metaphors (Pomp et al. 2018) or labels (de Araujo et al. 2005; Gonzalez et al. 2006; Pomp et al. 2018), as well as sniffing (Sobel et al. 1998; Kareken et al. 2004) recruited activation in brain regions including the piriform cortex, OFC, and insula, suggesting the role of these brain regions in semantic and odor coding integration (Olofsson and Gottfried 2015).

For patients with olfactory loss, two early fMRI studies investigated olfactory imagery among patients with hyposmia (Levy et al. 1999a, b; Henkin and Levy 2002). Both studies reported increased activation in olfactory regions following olfactory imagery. The activations were interpreted as a sign of preserved olfactory imagery capacity in patients with acquired olfactory loss. However, both studies suffered from methodological problems with few participants (n = 3) in the experimental condition (Levy et al. 1999a, b) or without having a control group (Henkin and Levy 2002).

Flohr et al. (2014) reported that olfactory loss leads to deficits in odor imagery. They found that patients with olfactory loss, compared to controls, experienced lower vividness when imaging odors, and had lower levels of brain activation in the hippocampus and insula. However, the patient group had a higher level of brain activation in the dorsal lateral prefrontal cortex (dlPFC), cerebellum and precuneus, regions that are related to working memory and autobiographical odor information retrieval (Dade et al. 2001; Curtis and D’Esposito 2003; Cavanna and Trimble 2006; Arshamian et al. 2013). Moreover, the level of the dlPFC activation was positively correlated with the duration of olfactory loss among patients (Flohr et al. 2014). These findings indicated that during odor imagery, patients were not able to image odors in a conventional way and needed more attention resources as the duration of odor deprivation increased.

During expectation of words with strong olfactory associations, Han et al. (2019) found that patients with acquired smell loss showed increased activation in the left inferior frontal gyrus extending to insular cortex and the bilateral angular gyrus, areas related to semantic knowledge retrieval and olfactory processing. This may indicate a higher effort and attention loads towards olfactory cues among patients, consequently from smell loss. Besides, during reading words with olfactory associations, patients and control participants with superior odor identification abilities showed reduced activation in the OFC and putamen, respectively. The lower brain activation may be linked to a higher degree of efficiency in information processing, a similar mechanism observed in olfactory experts during odor imagery (Plailly et al. 2012). However, for patients with congenital anosmia, less brain activation in the anterior cingulate cortex and the middle frontal gyrus was observed as compared to normal controls while reading of words with olfactory association (Yoshi et al. Unpublished). Based on these findings it can be hypothesized that prior olfactory experience, which differed between patients with acquired and congenital anosmia, had different impacts on the top-down processing of olfactory-related information.

Taken together, the olfactory-loss related deprivation of odor information input has a profound impact on both the bottom-up and top-down olfactory pathways. Moreover, it has to be taken into account that different causes of smell loss could have different impacts on the top-down olfactory modulation. A more central cause for the smell loss would therefore be likely to affect the top-down olfactory process more than a peripheral cause. Understanding of the top-down process in olfactory loss may help future direction for intervention or treatment such as olfactory training, which may drive associative learning-related brain plasticity in the olfactory system (Mandairon and Linster 2009; McGann 2015).

Changes of Functional Network Related to Olfactory Loss

Functional connectivity (FC) is a measurement of the temporal correlation of neuronal activity between anatomically separated brain regions. It reflects the functional communication between brain regions (Damoiseaux et al. 2006). The FC measured during resting state or during tasks are used to quantify the strength of inter-regional brain connectivity (van den Heuvel and Hulshoff Pol 2010).

During rest, FC has been shown between the primary olfactory regions (piriform cortex and OFC) and thalamus, medial prefrontal cortex, caudate, nucleus accumbens, parahippocampal gyrus and hippocampus (Tobia et al. 2016). Olfactory loss is associated with reduced FC in patients with Parkinson’s disease (Su et al. 2015; Yoneyama et al. 2018). Patients with hyposmia showed decreased FC within the olfactory related limbic or paralimbic regions, including the rectal gyrus and insula (Su et al. 2015). FC between the amygdala and inferior parietal lobule, lingual gyrus, and fusiform gyrus were associated with the severity of hyposmia (Yoneyama et al. 2018).

During passive odor smelling, dense connections between the primary (piriform cortex) and secondary olfactory regions (OFC, insula, and hippocampus) were identified (Nigri et al. 2013). With odor stimulation, disruption of the FC between the orbitofrontal cortex and medial temporal lobes, as part of the olfactory area, was observed in normal aging (Murphy et al. 2005), however, the olfactory functions were not reported for the group of elderly participants. Compared to people with normal olfaction, patients with anosmia caused by upper respiratory infection were found to have maintained but reduced number of FC in the somatosensory and integrative networks in response to odorous stimuli (Kollndorfer et al. 2015a, b). Applying the whole-brain network graph approach, Kollndorfer et al. (2015a, b) showed reduced FC of the olfactory brain areas including the anterior prefrontal cortex, anterior cingulate cortex, the entorhinal cortex and the cerebellum among patients with anosmia. Besides, a recent study showed the recruitment of a sensory processing network, as well as a cerebellar network and an occipital network during odor stimulation (Reichert et al. 2018), and the FC of these networks were correlated with olfactory functions measured by the Sniffin’ Sticks test. Besides, disrupted FC beyond the sensory or olfactory network had also been reported in patients with olfactory loss (Kollndorfer et al. 2015a, b; Su et al. 2015).

Taken together, olfactory dysfunction is related to a widespread reduction of FC in olfactory and non-olfactory networks. Future research needs to examine the links between the FC and specific olfactory functions (sensitivity, discrimination or identification), as well as the various of odor perceptual features (e.g. odor quality, intensity or valence). This could lead to further understanding of the interacting brain regions, rather than isolated regions that are involved in olfactory subset performances among patients with olfactory dysfunction.

Structural Brain Changes of Olfactory Loss

Olfactory Bulb Volume

As the first structural rely upon the olfactory pathway, the olfactory bulb (OB) integrates peripheral odor inputs and central activity of higher cortical regions. A large number of studies have demonstrated a positive correlation between the OB volume and olfactory function (Yousem et al. 1998; Buschhuter et al. 2008; Seubert et al. 2013; Mazal et al. 2016). Olfactory loss leading to a reduced sensory input to the OB is associated with atrophy of the OB. It has been demonstrated that the decreased olfactory function is associated with decreased OB volumes in patients suffering from a wide range of pathologies including post-traumatic olfactory disorder (Yousem et al. 1996; Rombaux et al. 2006a, b, c), post-infectious olfactory disorder (Mueller et al. 2005a, b; Rombaux et al. 2006a, b, c), sinonasal related olfactory disorders (Rombaux et al. 2008), idiopathic olfactory loss (Rombaux et al. 2010), and neurodegenerative diseases (Mueller et al. 2005a, b; Wang et al. 2011).

The degree of the OB volume reduction is related to the severity of olfactory loss. For example, in both post-infectious and post-traumatic olfactory loss, OB volumes in patients with anosmia were found to be smaller than those of hyposmia (Rombaux et al. 2006a, b, c; Han et al. 2018a, b). Besides, OB volume reduction was also affected by the causes of olfactory loss. Trauma patients had smaller OB compared to chronic rhinosinusitis or post-infectious patients (Hummel et al. 2015), possibly due to the direct damage of the OB or adjunct regions during the traumatic brain injury. In another study, the OB volume in olfactory-loss patients with chronic rhinosinusitis did not show a significant difference as compared to controls (Han et al. 2017). This is possibly due to intermittent odor input to the OB among those patients that prevents long-term plastic remodeling (Han et al. 2017). It has also been reported that the OB volume change in olfactory-loss patients with sinonasal disease may be more sensitive to the degree of sinonasal inflammation (Rombaux et al. 2008) than the impaired olfactory function observed with psychophysical testing.

Grey Matter and White Matter

Olfactory loss also results in widespread structural alterations in higher-order brain areas. Studies applied voxel-based morphometry, a method that allows uncovering of regional differences in gray matter volume (GMV) from MR images, had demonstrated a decrease of GMV across the primary and secondary olfactory cortex in patients with olfactory losses due to various causes (Table 1). These structural changes are thought to result from diminished sensory input (Bitter et al. 2010a, b). Although there is overlap between studies regarding the brain regions with GMV reduction, apparent differences exist. For example, patients with olfactory loss due to traumatic brain injury were found to have wide-spread GMV reduction in multiple areas belonging to the primary and secondary olfactory cortex (Han et al. 2018a, b), whereas GMV reduction was only observed in a few secondary olfactory areas (e.g. insula, OFC) among patients with smell loss due to sinonasal disease (Han et al. 2017) or upper respiratory tract infection (Gellrich et al. 2018; Yao et al. 2018). This suggests that the etiology of olfactory loss entails a significant contribution to the structural alteration of the central olfactory system. However, direct comparison between studies is difficult due to the variability regarding the sample size, severity of olfactory loss, and the time since the onset of olfactory loss. Thus, further studies including larger samples and with a stronger focus on participant selection are necessary to extend the results and to better understand the etiology of specific structural changes following olfactory loss. Besides, a few studies have investigated the GMV of the olfactory system in patients with congenital anosmia. Compared to control participants, those patients showed an increased GMV of the olfactory cortex, such as the piriform cortex (Frasnelli et al. 2013; Karstensen et al. 2018) and superior frontal sulcus (Karstensen et al. 2018). Therefore, the lifelong olfactory deprivation seems to trigger different changes regarding the cortical GMV as compared to changes found in acquired olfactory loss.

Table 1 Summary of literature showing gray matter volume reductions in patients with olfactory loss
Full size table

In addition to the regional structural alterations, olfactory loss also leads to changes in the anatomical connections between distant brain regions. Diffusion-weighted MRI, which characterizes the fiber architecture of tissues within the white matter, had been applied to explore anatomical brain connections (Jones et al. 2013). Using this technique, the measurement of the structural network includes the integrity, degree (number of connections), strength (number of fibers) and percentage of fibers per connections (Fernandes et al. 2015). The Fractional anisotropy (FA) as the most widely used parameter, describing the directionality of the molecular diffusion, is thought to reflect axon integrity (Jones et al. 2013) (FA value ranges from 0 to 1, with higher values indicating greater directionality of diffusion and preserved microstructure) (Jones and Leemans 2011). Until now, only a few studies had  focused on the structural connectivity in patients with olfactory loss. Patients with multiple sclerosis exhibited significantly reduced FA in olfactory brain regions related to impairment of the ability to identify odors (Erb et al. 2012; Erb-Eigner et al. 2014). A similar decline in olfactory performance due to normal aging was found to be correlated with low FA and mean diffusivity level in the corpus callosum and the superior longitudinal fasciculi (Segura et al. 2013). Ibarretxe-Bilbao et al. (2010) found that PD patients with olfactory loss showed FA decrease in areas adjacent to gyrus rectus and the primary olfactory area. The FA value in these areas was positively correlated with olfactory performances. Sobhani et al. (2019) observed positive associations between white matter connectivity and olfactory function in PD patients. Anatomical MRI could help to understand the substrates of olfaction dysfunctions and would, therefore, be a clinical marker in future studies.

MRI Studies of the Olfactory System with Olfactory Regain

Both functional and structural MRI had been used to assess the olfactory system in olfactory loss patients in relation to treatment of the disorder. MRI-based results indicated the high plasticity of the olfactory system. It was also demonstrated that olfactory training or treatment in individuals with olfactory deficits were accompanied by changes in the olfactory system. At a structural level, studies on olfactory loss patients with chronic rhinosinusitis have shown that improvement of olfactory function after surgical treatment was accompanied by increased OB volume (Gudziol et al. 2009), increased GMV in the primary and secondary olfactory brain regions (Whitcroft et al. 2018), as well as WM connectivity improvement in the anterior cingulate cortex and amygdala (Gullmar et al. 2017). Improvements in olfactory function (odor threshold) due to spontaneous recovery are related to the change of OB volume (Haehner et al. 2008). Besides, the “olfactory training”, i.e. systematic repeated exposure to odors, improved olfactory function and the volume of OB in patients with smell loss and was also shown to be associated with an increase in GMV (Gellrich et al. 2018). At the functional level, a 12-week “olfactory training” in a group of olfactory loss patients after upper respiratory infection helped to reestablish olfactory function, and restored the intensity of functional connectivity for the olfactory networks as well as the integrative and somatosensory networks (Kollndorfer et al. 2015a, b). In addition, olfactory training affected functional responses to odor stimulation, although brain activation may be driven by top-down processes (Pellegrino et al. 2019).

From Group Inference to Individual Prediction

Individual fMRI Biomarker for Olfactory Loss

Searching for reliable brain activation patterns as a potential clinical biomarker is important for the future application of olfactory fMRI research (Frohner et al. 2019). The odor-induced fMRI signals, especially in the primary olfactory cortex, was argued to have low signal-to-noise ratio and high inter-individual variability and thus has major limitations for diagnosis and other clinical application (Morrot et al. 2013). However, a recent study found that the detectability and reproducibility of olfactory activation in the primary olfactory cortex was at the same level of that in the visual cortex (given the visual cortex activation as an ideal benchmark which produces the most robust fMRI data), and demonstrates the usefulness of this application to clinical studies as a reliable imaging tool (Lu et al. 2018). Besides, several studies reported the improvement of detectability and reproducibility of the olfactory fMRI signals in the olfactory cortex through the inclusion of short odor stimulation time (Kleinhans et al. 2019), rapid repetition time (< 1 s) during brain image acquisition (Georgiopoulos et al. 2018), or specific task protocols (e.g. synchronization of the breathing cycle with odor stimulation) (Kleinhans et al. 2019).

Currently, application of fMRI response patterns as individual “biomarker” for diagnosis of olfactory dysfunction is far from clinical routine. The principle is to have one or more stable and reliable brain activation patterns within and across individuals. Classical experimental research aims to minimize inter-individual variability by identifying robust generalizable group-level effects. For olfactory fMRI, the optimization of between-subject variance needs to consider factors such as odor selection, stimulation paradigm, or choice of contrasts for group inference. For example, passive odor exposure or active sniffing has a major influence on the BOLD signal in the primary olfactory area (Wang et al. 2014).

The intra-individual variability and stability are dependent on the regions of interest and tasks applied (Vetter et al. 2017). Moreover, research is also required to explore intra-individual reliability over time before they can be routinely applied for prediction or classification of patients with olfactory loss (Frohner et al. 2019). Ultimately, this may help in identifying paradigms, contrasts, and conditions yielding sufficiently high reliability for prediction and classification of individuals undergoing olfactory fMRI data.

Machine Learning in Olfactory Loss Research

Broadly speaking, machine learning explicitly focus on learning statistical functions using the computational strategy from multidimensional data sets to make generalizable predictions about individuals (Dwyer et al. 2018). Using MRI or fMRI data, machine learning has been used in clinical populations to identify individuals with psychiatric disorders, such as Alzheimer’s disease (Kloppel et al. 2008), depression (Fu et al. 2008), schizophrenia (Davatzikos et al. 2008), anxiety disorders (Lueken et al. 2015) and anorexia (Lavagnino et al. 2018). Using resting-state data, multivariate signatures that are valid at single-subject level could be used as biomarkers to monitor the progress of the olfactory loss or the effectiveness of treatment or training (Woo et al. 2017).

Machine learning has also been applied in different aspects of olfactory research [for a review, see Lotsch et al. (2019)]. For example, using psychophysical datasets, it was able to predict olfactory deficits and the underlying etiologies from olfactory subtest results (Lotsch et al. 2016). However, for olfaction, there has been no published paper using MRI or fMRI data involving machine learning. One barrier for this could be that the functional brain activation, even in the primary olfactory cortex, involves cognitive modulations, as discussed above. Besides, another barrier could be the small sample size and the limited number of stimulations per participant.

Summary

In this article, we have attempted to review and discuss the progress of research on human olfactory loss using MRI. To date, many studies have shown a multitude of changes in the central olfactory system following smell loss—on both functional and structural levels. However, many open questions and challenges remain which need future investigation. First, longitudinal studies focusing on the rehabilitation of olfactory functions are needed to distinguish more clearly between cause and effect of olfactory loss. Besides, research on the neural basis for specific olfactory dysfunctions (e.g. impairment regarding odor quality discrimination or identification, odor valence) could provide insights to the mechanism of some olfactory dysfunctions present in patients with certain disorders such as Parkinson’s disease or schizophrenia (Atanasova et al. 2008). Moreover, taking advantage of the close connection between brain structure and function, the joint analysis of the anatomical and functional MRI data (Hermundstad et al. 2013) will contribute to the detection of subtle reorganizations in the central olfactory system. From a clinical perspective, the establishment and validation of MRI-based biomarkers could help as a non-invasive method to achieve a better diagnosis and treatment of olfactory loss.