Diagnostic accuracy of proton magnetic resonance spectroscopy and perfusion-weighted imaging in brain gliomas follow-up: a single institutional experience

Abstract

Objectives

The objective of this study was to evaluate whether proton magnetic resonance spectroscopy and perfusion magnetic resonance imaging (MRI) are able to increase diagnostic accuracy in the follow-up of brain gliomas, identifying the progression of disease before it becomes evident in the standard MRI; also to evaluate which of the two techniques has the best diagnostic accuracy.

Methods

Eighty-three patients with cerebral glioma (50 high-grade gliomas (HGGs), 33 low-grade gliomas (LGGs)) were retrospectively enrolled. All patients underwent standard MRI, H spectroscopic and perfusion echo-planar imaging MRI. For spectroscopy variations of choline/creatine, choline/N-acetyl-aspartate ratio, and lipids and lactates peak were considered. For perfusion 2.0 was considered the cerebral blood volume cut-off for progression. The combination of functional parameters gave a multiparametric score (0–2) to predict outcome. Diagnostic performance was determined by the receiver operating characteristic curve, with sensitivity, specificity, positive predictive and negative predictive values.

Results

In patients with LGGs a combined score of at least 1 was the best predictor for progression (odds ratio (OR) 3.91) with 8.4 months median anticipation of diagnosis compared to standard MRI. The individual advanced magnetic resonance technique did not show a diagnostic accuracy comparable to the combination of the two. Overall diagnostic accuracy area under the curve (AUC) was 0.881. In patients with HGGs the multiparametric score did not improve diagnostic accuracy significantly. Perfusion MRI was the best predictor of progression (OR 3.65), with 6.7 months median anticipation of diagnosis. Overall diagnostic accuracy AUC was 0.897. Then spectroscopy and perfusion MRI are able to identify tumour progression during follow-up earlier than standard MRI.

Conclusion

In patients with LGGs the combination of the functional parameters seems to be the best method for diagnosis of progression. In patients with HGGs perfusion is the best diagnostic method.

Introduction

Gliomas represent 27% of all brain tumours and 80% of all malignant brain tumours;^1,2 they are characterised by significant heterogeneous genetic, neuropathological, molecular and prognostic features. The 2016 World Health Organization (WHO) Classification of Tumors of the Central Nervous System, incorporating molecular findings into brain tumour diagnoses, presents major restructuring of diffuse gliomas over its 2007 predecessor; while in the past all astrocytic tumours had been grouped together, now all diffusely infiltrating gliomas (whether astrocytic or oligodendroglial) are grouped together.

This new classification distinguishes between ‘diffuse astrocytic and oligodendroglial tumours’ and ‘neuronal and mixed neuronal-glial tumours’: astrocytoma, oligodendroglioma, oligoastrocytoma, their anaplastic variants and glioblastoma belong to the former category, gangliocytoma and ganglioglioma are included in the latter.³

Gliomas can be divided into two major groups: low-grade gliomas (LGGs; grade II), representing 20–30% of all gliomas (5-year survival 65–78%), and high-grade gliomas (HGGs; grades III and IV), including about 50% of all gliomas (5-year survival 5% for glioblastoma).⁴

In the 2016 update, the diffuse gliomas group includes WHO grade II and grade III astrocytic tumours, grades II and III oligodendrogliomas and grade IV glioblastomas.³

Histological analysis of tissue specimens obtained through biopsy, either stereotactic or open, is the gold standard to establish brain tumour histological type and grade, but this procedure is invasive⁵ and may present sampling error if the most malignant portion of the tumour is not sampled. Magnetic resonance imaging (MRI) is an alternative non-invasive tool for the identification and characterisation of intracerebral gliomas. It is able to provide a result closely corresponding to the histological diagnosis⁶ also for the patient’s follow-up after treatment (radiotherapy, chemotherapy or radio-chemotherapy).

Response evaluation criteria in solid tumours (RECIST), based on both T2/fluid-attenuated inversion recovery (FLAIR) images and post-contrast images, are useful to identify tumour progression on standard MRI.⁴

Even if the main predictor for the assessment of treatment response on conventional MRI is the change in the volume of the tumour, which however will take some time to develop and to be depicted, the specificity of this technique has limitations to assess glioma degree and tumour margins, differentiate tumour from surrounding oedema, and differentiate tumour recurrence from therapy-related necrosis, pseudo-progression or pseudo-response.^7–9

The phenomenon of so-called ‘pseudo-response’ may occur in patients who receive anti-angiogenetic agent therapy that improves the brain–blood barrier (BBB) and corresponds to a reduction of contrast enhancement on T1-weighted images without any effect on tumour cell viability.¹⁰ This condition has been observed in multiple studies and it is estimated to occur in approximately 20% of patients following glioblastoma multiform treatment.¹¹

Instead, the so-called ‘pseudo-progression’ is defined as the increase of the enhancing area of the tumour, without clinical symptoms worsening, which disappears in the successive MRI, without any change in therapy.^11–13 It can develop after radiotherapy alone or, more frequently, after concomitant radiotherapy and chemotherapy with temozolamide.

It is very important to recognise this situation, because, as it happens also in cases of radiation necrosis, the appearance of an enhancing area might lead us to suspect a tumour recurrence that requires more aggressive treatment in order to have an impact on the course of disease.

So, during the follow-up of these patients, in order to differentiate stable disease, responder and non-responder patients, it is very important to assess functional parameters, which would be extremely helpful to differentiate tumour recurrence from therapy-related effects.

The advanced MRI techniques, in particular spectroscopic (¹H-MRS) and perfusion MRI, are able to overcome the limitations of conventional MRI, giving us pathobiological information about the proliferation rate, cellular metabolism (spectroscopy), microcirculation (perfusion),⁴ and always allowing a more precise diagnosis and satisfaction for neurosurgeons, neuro-oncologists and radiotherapists. They also represent a valuable tool in the assessment of therapy effects and monitoring of operated glioma patients, allowing us to identify early therapy-induced changes and to decide early if the treatment plan should be continued, adapted or completely changed.

The aim of this retrospective study was to evaluate whether spectroscopic and perfusion MRI are able to increase diagnostic accuracy during brain glioma follow-up, identifying progression of the disease before it becomes evident in the standard MRI; also to evaluate which of the two techniques has the best diagnostic accuracy.

Patients and methods

From January 2008 to March 2015, we retrospectively selected 83 patients with cerebral glioma submitted to MRI examination in our institute; 50 with HGGs and 33 with LGGs; all patients received MRI every 3 or 6 months.

Patients with LGGs underwent complete or partial surgical resection; this latter group includes the subjects who underwent diagnostic biopsy.

All patient with HGGs underwent macroscopically complete or partial surgical resection, radiotherapy or radiochemotherapy according to the plan of Stupp et al.¹⁴ It consists of a combination of radiotherapy and chemotherapy. The patients receive radiotherapy plus continuous daily temozolamide (7 days per week from the first to the last day of radiotherapy, at a dose of 75 mg/m² of body surface area per day) followed by six cycles of adjuvant temozolamide (150–200 mg/m² for 5 days during each 28-day cycle).¹⁴

Some patients also underwent treatment with antiangiogenetic agents.

Patients with LGGs were 18 men and 15 women, median age 41 years ± 13.32 (Table 1). Patients with HGGs were 31 men and 19 women, median age 60 years ± 9.09 (Table 2). All patients were informed about the procedure and gave their informed consent.

Table 1.

Patient characteristics low-grade gliomas (LGGs).

Characteristics	No. of patients = 33
Age, median (range), years	41 (17–69)
Sex
Male	18
Female	15
Histology
Oligodendroglioma	11
Astrocytoma	19
Oligoastrocytoma	1
Ganglioglioma	2
Therapy
Complete surgical resection	10
Partial surgical resection/biopsy	23

Table 2.

Patient characteristics high-grade gliomas (HGGs).

Characteristics	No. of patients = 50
Age, median (range), years	60 (38–84)
Sex
Male	31
Female	19
Histology
Glioblastoma	29
Anaplastic astrocytoma	10
Anaplastic oligodendroglioma	7
Anaplastic oligoastrocytoma	4
Therapy
Complete surgical resection	33
Partial surgical resection	17
Radio-chemotherapy	50

Inclusion criteria

• Patients with histologically confirmed LGGs and HGGs;

• No contraindication to MRI examination;

• No oncological treatment before MRI examination.

Exclusion criteria

• Patients with oncological history;

• Patients with allergy to contrast medium;

• Patients with contraindication to MRI examination;

• Patients who could not remain still;

• Pregnancy.

All patients underwent an MRI examination using a whole-body scanner (Sigma Horizon 1.5 T, General Electric Healthcare); standard (before and after contrast administration) and functional MRI (H spectroscopic and perfusion MRI) were performed during the same session.

Standard MRI

The standard MRI protocol comprised anatomical sequences including axial T1 spin echo (SE) (TR/TE = 360/8 ms), before and after contrast administration, axial DP-T2-SE (TR = 2660 and TE = 98.96 and 14.84), coronal and axial FLAIR (TR/TE/TI = 8000/85.25/1900), diffusion-weighted imaging (DWI) (TR/TE = 8000/104) and sagittal T1 three-dimensional fast spoiled gradient echo after-contrast administration (TR/TE = 9.52/4.20 slice thickness 1 mm).

The contrast material used was gadobutrol 1 mol, 0.1 ml/kg body weight.

In our study, we decided to define the tumour progression on standard MRI using RECIST¹⁵ because we evaluated only the MRI changes without correlation with clinical aspects and/or corticosteroid therapy.

RECIST criteria, used routinely for clinical practice, have been reported to have a high concordance with two-dimensional (2D), total tumour volume, and enhancing tumour volume methods in assessing tumour progression.

According to these criteria, the disease progression in LGGs was defined by the increase of the longest tumour diameter on FLAIR or T2-weighted images and in HGGs by the increase of the longest diameter of the tumour enhancing area in T1-weighted image after contrast administration at least 20%. In the case of multiple lesions per patient, the increase in the sum of the longest tumour diameters per patient had to be at least 20%. As a second criterion for tumour progression in the group of LGGs, we used a new contrast enhancement that persisted during subsequent follow-up examinations, thus excluding radiation necrosis. The third criterion is the appearance of one or more new lesions.⁴

Progression of the disease was assumed when at least one of the aforementioned three criteria was met.

Spectroscopy MRI

Single-section 2D multi-voxel and single-voxel ¹H-MRS were performed using a point-resolved spectroscopy sequence (PRESS) before and/or after contrast administration.

Multi-voxel sequence parameters included TR of 1000 ms, TE 144 ms, nex 1, field of vision (FOV) 18, matrix size 256 × 256, the voxel size was 1.2 cm³ and lasts about 3–4 minutes.

The volume of interest (VOI) was drawn on FLAIR images to include the area of altered signal (neoplasm, residual disease, relapsed disease, gliosis and oedema) and contralateral apparently normal brain tissue in the same plane, being careful to avoid areas of scalp, skull base and sinuses.

Single-voxel ¹H-MRS with PRESS was applied with a voxel size of 1.5 cm³ with an acquisition time of 3–4 minutes; standard parameters are the following: TR of 1500 ms, TE 144 and TE 35 ms, nex 8, FOV 24.00, matrix size 516 × 516. The VOI was drawn on altered signal areas on a FLAIR image or on an enhancing area.

In addition, saturation slabs were placed outside the VOI to suppress lipid signals from bone and scalp. For each patient, we chose to use multi-voxel and/or single-voxel spectroscopy independently of the tumour grading but taking into consideration the anatomical area, the size of the lesion to be examined and the heterogeneity of the area, in order to obtain spectra with the highest possible quality.

If the investigated region was widespread, it was necessary to use more than one slab for the multi-voxel technique; variation in the metabolic ratio was always evaluated on voxels corresponding to the area which shows the greater signal alteration on FLAIR and/or enhancement after contrast administration.

At each subsequent control, in order to optimise the reproducibility of diagnostic investigation, we placed the VOI for the single-voxel technique and the spectroscopic grid for the multi-voxel technique in the same area and layers of the previous control.

New areas of altered signal were also investigated.

The raw spectroscopic data were processed using Functool GE, to obtain the spectra and an accurate quantification of the ratio metabolites; the peak amplitude has been considered while measuring the metabolite ratio.

Creatine was considered as a benchmark.

To assess tumour progression by spectroscopy, we considered variations of both the choline (Cho)/creatine (Cr) and Cho/N-acetyl-aspartate (NAA) ratio and the presence of lipids and lactates (Lip/Lac) peak.

In particular, to assess tumour progression to HGG we considered these cut-off values: 1.9 for Cho/Cr, 1.5 for Cho/NAA, presence of the Lip/Lac peak, considering this tool indicative for a change to high grade when almost one of these three ratios was significantly altered.

Perfusion MRI

Magnetic resonance perfusion-weighted imaging is a rapid, non-invasive and useful technique that is used to characterise brain tumour.

The most commonly used MRI perfusion technique in clinical practice is dynamic susceptibility weighted bolus tracking MRI obtained after the intravenous injection of a bolus of paramagnetic contrast medium and first-pass acquisition of T2*-weighted gradient echo-planar imaging (EPI) perfusion sequence of the region of interest (ROI).

MRI perfusion with GR-EPI sequences was acquired during the intravenous infusion of gadobutrolo 1 mol, after the administration of a pre-bolus (2 ml) that minimises errors in perfusion quantification due to BBB leakage.

The parameters of this sequence are the following: TR of 1800 ms, TE 32.90, nex 1, FOV 24.00, matrix size 128 × 128, slice thickness 5 mm.

With this sequence, the images of 19 slices were repeated 50 times for a total acquisition time of 1.30 minutes. The sequence and the contrast agent injection began simultaneously, by a power injector at an injection rate of 6 ml/second with an injection delay of 10 seconds.

The processing of raw perfusion data and cerebral blood volume (CBV) calculation were made using Functool GE. The measurements of intralesional CBV were obtained from the processed CBV maps by manual ROI placement in highest CBV areas, being careful to avoid uninvolved adjacent grey matter structures; each ROI measured about 50 mm².

Normal-appearing white matter ROIs (judged as normal on corresponding T2-weighted, FLAIR and contrast-enhanced T1-weighted images) were placed in the contralateral hemisphere, and a maximum CBV ratio (rCBV) was then calculated for each tumour using the formula: highest average CBV in lesion/highest average CBV in the contralateral white matter.

In the perfusion study we measured rCBV differences between the altered area and apparently normal area using 2.0 as the cut-off for tumour progression.

Unprocessed perfusion images were evaluated and compared with co-registered axial-FLAIR and contrast-enhanced T1-weighted images to ensure that ROIs were not placed over blood vessels.

For each patient, the combination of functional parameters (spectroscopy and perfusion) gave a multiparametric score of (0–2) for the prediction of the outcome.

Statistical analyses

Continuous variables were summarised as medians and range values. Dichotomous variables were summarised as absolute and relative frequencies. Diagnostic performances were determined by receiver operating characteristic curve, with sensitivity (SE), specificity (SP), positive predictive value (PPV) and negative predictive value (NPV). Statistical calculations were performed using SPSS for Windows version 19.0 (SPSS, Chicago, IL, USA). Findings with a P value less than 0.05 were considered statistically significant. Functional parameter values are expressed as means.

Results

In LGGs, 22 out of 33 examined patients were stable to control MRI while 11 patients had disease progression with transformation to high grade, with mean values of each functional parameter in stable patients and in progressing patients reported in Figure 1.

Figure 1.

Comparison of mean value of Cho/Cre, Cho/NAA ratio and rCBV and presence of Lip/lac peak in stable and progression LGG patients.

The mean follow-up of all patients with LGGs was 32.2 months (lowest value 20 months, highest value 50). The mean follow-up of patients with stable disease at MRI was 33 months (lowest value 20 months, highest value 50). The Kaplan–Meier survival curve depicting the overall survival of patients radiologically progressed versus patients not progressed was reported in Figure 2.

Figure 2.

Curve of survival in LGG patients.

In patients with LGGs, a combined score of at least 1 was the best predictor for disease progression (odds ratio (OR) 3.91) with a 8.4 months median anticipation of diagnosis compared to standard MRI.

RECIST: No single advanced technique showed a diagnostic accuracy comparable to the combination of the two, as in the case shown in Figure 3.

Figure 3.

Follow-up of patient with low grade glioma (LGG): December '09 (A) shows an increase of CBV in the left semi-oval center (A3): this finding preceded by 6 months (B) the appearance of a greater extension of the alteration signal (B1) and by 8 months (C) the appearance of a nodule enhancement (C2). The alterations of metabolic parameters compatible with a change in high-grade gliomas are also evident at 6 to 8 months (B4 and C4).

Overall diagnostic accuracy (AUC) was 0.881 (SE, SP, PPV, NPV: 84.1%, 89.9%, 84.7%, 81.5%). None of the individual advanced techniques presented an AUC comparable to their combination.

In fact, for the spectroscopy alone the overall diagnostic accuracy (AUC) was 0.785 (SE, SP, VVP, VPN: 68.2%, 88.9%, 83.3%, 77.4%); instead for the perfusion alone the overall diagnostic accuracy (AUC) was 0.82 (SE, SP, VVP, VPN: 77.2%, 92.59%, 89.5%, 83.3%) (Figure 4).

Figure 4.

ROC curve in low-grade gliomas.

Regarding the HGGs, 10 out of the 50 examined patients were stable in the MRI examinations, and 40 patients had disease progression, with average values of each functional parameter in stable patients and in progressing patients reported in Figure 5.

Figure 5.

Comparison of mean value of Cho/Cre, Cho/NAA ratio and rCBV and presence of Lip/lac peak in stable and progression HGG patients.

The mean follow-up of all patients with HGGs was 13.3 months (lowest value 8 months, highest value 20). The mean follow-up of patients with stable disease at MRI was 16.3 months (lowest value 12 months, highest value 20). The Kaplan–Meier survival curve depicting the overall survival of patients radiologically progressed versus patients not progressed was reported in Figure 6.

Figure 6.

Curve of survival in HGG patients.

Spectroscopy had no diagnostic quality in nine HGG patients due to the artefact occurring at the surgical site.

One patient with pseudo-progression (Figure 7) and one patient with pseudo-response (Figure 8) were included, respectively, in the group of stable patients and in the group of patients in progression.

Figure 7.

Follow-up of patient with high-grade glioma (HGG); February 2013 some enhancement nodules appeared on the left frontal region (A3), where, however, it did not correspond an increase of CBV (A3), findings suggested a pseudo-progression; at the next check up in July, after the intravenous administration of contrast medium, the nodules were no longer evident (B2). Stable even at the last check up in April 2014 (C2).

Figure 8.

Follow up of patient with high grade glioma (HGG) with anti-angiogenic therapy. In February 2013, MRI showed extensive enhancement areas around chirurgic cable in left temporal polar region (A1-A3) which corresponded to high CBV value (A2, A4); in the next check up in May 2013, the MRI showed a reduction of enhancement at the same level (B1-B3); however, the high perfusion remained (B2-B4): this suggested the hypothesis of pseudo-response, confirmed at the next check up, after two months there was a new increase in the enhancement area (C1-C3).

In patients with HGGs the multiparametric score did not improve the diagnostic accuracy significantly; perfusion was the best predictor of disease progression (OR 3.65), with an average anticipation in diagnosis of 6.7 months compared to standard MRI as in the case shown in Figure 9. In fact, for spectoscopy alone the overall diagnostic accuracy (AUC) was 0.792 (SE, SP, VVP, VPN: 71.2%, 90.2%, 73.2%, 89.4%); instead for the combination of spectroscopy and perfusion the overall diagnostic accuracy (AUC) was 0.792 (SE, SP, VVP, VPN: 71.2%, 90.2%, 73.2%, 89.4%). Overall diagnostic accuracy (AUC) was 0.897 (SE, SP, VPP, VPN: 80.1%, 100%, 100%, 64.6%) (Figure 10).

Figure 9.

Follow-up of patient with high grade glioma (HGG). HGG operated in May 2010 in left temporal-insular region, in follow-up. In RM study in January 2011 (A) showed an increase of CBV in the right above-trigonal areas (A3): the single-voxel spectroscopic, in the same area showed an increase of metabolic ratio suggestive for tumor progression (A4). Spectroscopic and perfusion alterations preceded by 4 months (B) the evidence of disease progression in standard MRI (B1 and B2).

Figure 10.

ROC curve in high-grade gliomas.

Discussion

Conventional MRI is the examination of choice in the evaluation of brain tumours; given its multiplanar and multiparametric capabilities and its best contrast resolution, in the absence of invasiveness, MRI allows the lesion to be identified, the exact intra-axial or extra-axial localisation to be defined, and a hypothesis of nature to be proposed.

However, it has a limited sensibility and specificity in the definition of both the type and grading of gliomas. Furthermore, standard MRI does not always allow a precise definition of the margins of brain intra-axial tumours or does not permit the tumour to be distinguished from oedema and/or treatment effects (recurrence, radionecrosis, pseudo-response and pseudo-progression).

Thanks to metabolic, structural and haemodynamic information, the new advanced MRI techniques can reduce diagnostic limitations of standard MRI, allowing even more precise and exhaustive diagnosis for the neurosurgeon, the oncologist and the radiation oncologist. The same, in fact, allows us to assess the average close pathophysiological aspects of the tumour, as the microcirculation by perfusion and cellular metabolism by spectroscopy, providing a correct diagnosis, a distinction between HGGs and LGGs, and a reliable follow-up after surgery. In the latter condition, in particular, new advanced MRI techniques allow us to identify the transformation from low to high grade. They also allow us to distinguish, after radiotherapy, the recurrence of the tumour by radiation necrosis or misleading conditions such as pseudo-progression or pseudo-response.

Unfortunately, with conventional MRI, tumour recurrences often have similar neuroradiological features of pseudo-progression or radionecrosis. Standard T2-weighted sequence and gadolinium-enhanced T1-weighted sequences have insufficient specificity to differentiate between recurrence, radionecrosis and pseudo-progression, due to their similar neuroimaging patterns characterised by contrast enhancing lesions surrounded by oedema.¹⁶ In many situations, changes in enhancement do not correlate with response, for example in the case of pseudo-progression, in which an increase in contrast uptake does not reflect tumour progression, or pseudo-response, in which a decrease in contrast enhancement does not reflect tumour regression in patients treated with antiangiogenetic agents.¹⁷ Therefore, in the follow-up of patients with brain gliomas, the use of advanced MRI techniques allows us to assess more accurately the response to therapy.¹⁸

Although we included only one case of pseudo-progression and one of pseudo-response, perfusion was extremely useful for the correct differential diagnosis, according to literature data.

According to the literature, cerebral perfusion in MRI is an important marker of microvascular density and, thus, of the histological grade of a brain tumour. In fact, with the increase of tumour grading, there is an increase in both the microvascular density and in the neoangiogenesis, which are essential histological criteria in order to determine the level of the biological aggressiveness of gliomas. The higher these two parameters are, the higher is the aggressiveness of the neoplasm.^19,20

All these parameters translate into an increase of the rCBV, which is considered one of the major factors in predicting neoplastic aggressiveness.^21,22

In the literature there are many studies using dynamic susceptibility contrast MRI, as we do, that report lower perfusion values in treatment-related changes compared to recurrent tumours in patients with HGGs.^21–23 So, in the follow-up of patients with HGGs, perfusion MRI allows us to identify any residual of the tumour, to anticipate the development of tumour recurrence or to distinguish post-treatment effects. It also allows us to make a differential diagnosis between pseudo-progression or pseudo-response and true progression.

The LGGs have a very slow growth (some years) but they can change at any given time, becoming a HGG. In these patients it is very important that the MRI follow-up is with perfusion MRI because it will allow us to evaluate changes in the angiogenesis (rCBV), therefore allowing a quick intervention, before the morphological changes can take place and become evident on standard MRI.

Different perfusion studies use different cut-off values to differentiate between low and high-grade tumours, with 1.75 as the lowest value;^24–26 we decided to use a comparatively high threshold as a cut-off (2.0) to obtain a really indicative value of high-grade transformation, also because we included patients with oligodendroglioma; in fact, the literature shows that oligodendroglioma has a higher perfusion value than other low-grade tumours.

On the other hand, magnetic resonance spectroscopy (MRS) provides information regarding the metabolic characteristics of the gliomas and, in combination with perfusion and conventional MRI, increases the number of correct diagnoses. It gives quantitative parameters of the membrane turnover and consequently of the cellular proliferation (Cho), the energetic homeostasis (Cre), the integrity of the neuronal cells (NAA) and the cellular necrosis (Lip/Lac).²⁷

Metabolites are quantified in relation to creatine, that is used as a benchmark given its stable concentration. In the first diagnosis of cerebral tumour, spectroscopy is essential because it helps to make a histological diagnosis (for example in the diagnosis of oligodendroglioma, which typically presents with elevated perfusion values even if it is low grade). Spectroscopy will furthermore be important in the follow-up of patients who have had brain surgery, because it will allow us to analyse the areas with altered residual signal for a correct characterisation of them (gliosis or residual tumour).

In the follow-up of patients operated for LGGs, the increase of both Cho/Cre and Cho/NAA ratios of the residual tumour induced the suspicion of a transformation to HGGs; also, the appearance of the Lip/Lac peak supports the diagnostic hypothesis because it is observed only in cases of progression to high-grade tumour (Figure 1).

In our study, according to the cut-off values chosen for each functional parameter of spectroscopy and perfusion in order to assess tumour progression, we found that in LGG the combination of these two advanced imaging techniques seem the best diagnostic tool to identify tumour transformation to HGG with an average anticipation of 8 months compared to standard MRI; neither of the two advanced MRI diagnostic techniques have a diagnostic accuracy equal to their combination.

In HGGs, however, the combination of these two techniques did not significantly increase the diagnostic accuracy, but perfusion MRI was the best predictor of tumour progression, with a median anticipation of the diagnosis of 6 months compared to the standard MRI.

The apparent discrepancy emerged from our findings that in LGGs a combination of both advanced MRI techniques give the best diagnostic accuracy, while in HGGs only perfusion MRI can be justified taking into account the following two aspects.

On one hand, in fact, was the apparent easier diagnosis of tumour progression in the case of HGGs, compared to the progression of the tumour in the LGG group. Not only was this in terms of the greater frequency of recurrence according to the natural history, but also for the highest values of rCBV typically observed at the time of relapse, as also seen in our experience and expressed by the average value of rCBV at the relapsing time, which is highest in the HGG group in progression (Figure 5).

In the case of LGGs, in our experience we find a greater level of diagnostic difficulty. The diagnostic hypothesis of transformation to high grade needs the integration of more MRI functional parameters, usually occurring with a more gradual rCBV increase (Figure 1), which therefore, for the correct diagnosis, must be supported from the simultaneous changes of metabolic parameters by spectroscopy.

On the other hand, based on our experience, it is possible to affirm that spectroscopy is very sensitive and susceptible to technical artifacts, which are affected by heterogeneity of the region of interest. The heterogeneity is higher in HGGs than LGGs and will be even more in patients surgically treated for HGGs (haemorrhage, surgical clips, adjacent bony structures, cerebrospinal fluid, etc.), where artifacts from anatomical inhomogeneity or ferromagnetic ones will further limit the diagnostic capabilities of MRS in post-surgical follow-up,²⁸ as occurred in nine HGG patients whose spectroscopy was not diagnostic. In patients with LGGs, however, the ROI will be more homogeneous and less prone to artifacts from surgery, and then MRS may offer more reliable results.

Although we have tried to make the samples as similar as possible in the subsequent follow-ups, one of the limitations of this technique is the possible difficulty in positioning the VOI and the spectroscopy grid that may not be perfectly reproduced.

Therefore, taking into account the apparently easier diagnostics in the case of recurrence in patients with HGGs and also the spectroscopic limits, especially in HGG-operated patients, two different methods of diagnostics can be taken. The latter can be studied with only the sequence of perfusion in addition to the standard MRI.

The possibility of reducing the acquisition time of the MRI examination may be useful in HGG patients who may be less compliant compared to LGG patients, due to the more negative surgical outcomes for clinical decline due to the progression of the disease. In patients with LGGs the correct diagnostic approach should include the acquisition of perfusion and spectroscopy in addition to the standard MRI.

Our experience in the follow-up confirms the findings of recent literature for the diagnosis;²⁹ according to which, the rCBV alone has a huge impact on the evaluation of tumour grading at the first observation, but a more accurate classification of the tumour will require the combination of all imaging parameters (standard MRI, DWI, spectroscopy and perfusion).

In fact, although in case of LGG the greater diagnostic accuracy results from the combination of both advanced techniques, perfusion alone presents values of sensitivity and specificity (77.27% and 92.59%) greater than spectroscopy alone (68.2% and 88.9%).

In conclusion, according to the literature, our study confirms that the use of MRS and perfusion improves the correct diagnosis compared to standard MRI, and can anticipate the evidence of tumour progression during follow-ups.

Furthermore, in patients with LGGs, the findings of alteration in at least one of the functional parameters studied (perfusion and spectroscopy) may suggest a tendency towards tumoral progression, even though the best diagnostic accuracy results from the combination of both techniques. In HGGs the best diagnostic technique predictive of progression seems to be perfusion MRI; therefore, it may be proposed as the only advanced method to be performed in the follow-up of these patients in addition to standard examination.

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Conflict of interest

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.