3D Magnetic Resonance Spectroscopy Research Articles

Deep learning pipeline for quality filtering of MRSI spectra.

With the rise of novel 3D magnetic resonance spectroscopy imaging (MRSI) acquisition protocols in clinical practice, which are capable of capturing a large number of spectra from a subject's brain, there is a need for an automated preprocessing pipeline that filters out bad-quality spectra and identifies contaminated but salvageable spectra prior to the metabolite quantification step. This work introduces such a pipeline based on an ensemble of deep-learning classifiers. The dataset consists of 36,338 spectra from one healthy subject and five brain tumor patients, acquired with an EPSI variant, which implemented a novel type of spectral editing named SLOtboom-Weng (SLOW) editing on a 7T MR scanner. The spectra were labeled manually by an expert into four classes of spectral quality as follows: (i) noise, (ii) spectra greatly influenced by lipid-related artifacts (deemed not to contain clinical information), (iii) spectra containing metabolic information slightly contaminated by lipid signals, and (iv) good-quality spectra. The AI model consists of three pairs of networks, each comprising a convolutional autoencoder and a multilayer perceptron network. In the classification step, the encoding half of the autoencoder is kept as a dimensionality reduction tool, while the fully connected layers are added to its output. Each of the three pairs of networks is trained on different representations of spectra (real, imaginary, or both), aiming at robust decision-making. The final class is assigned via a majority voting scheme. The F1 scores obtained on the test dataset for the four previously defined classes are 0.96, 0.93, 0.82, and 0.90, respectively. The arguably lower value of 0.82 was reached for the least represented class of spectra mildly influenced by lipids. Not only does the proposed model minimise the required user interaction, but it also greatly reduces the computation time at the metabolite quantification step (by selecting a subset of spectra worth quantifying) and enforces the display of only clinically relevant information.

Fast high-resolution metabolic imaging of acute stroke with 3D magnetic resonance spectroscopy.

Impaired oxygen and cellular metabolism is a hallmark of ischaemic injury in acute stroke. Magnetic resonance spectroscopic imaging (MRSI) has long been recognized as a potentially powerful tool for non-invasive metabolic imaging. Nonetheless, long acquisition time, poor spatial resolution, and narrow coverage have limited its clinical application. Here we investigated the feasibility and potential clinical utility of rapid, high spatial resolution, near whole-brain 3D metabolic imaging based on a novel MRSI technology. In an 8-min scan, we simultaneously obtained 3D maps of N-acetylaspartate and lactate at a nominal spatial resolution of 2.0 × 3.0 × 3.0 mm3 with near whole-brain coverage from a cohort of 18 patients with acute ischaemic stroke. Serial structural and perfusion MRI was used to define detailed spatial maps of tissue-level outcomes against which high-resolution metabolic changes were evaluated. Within hypoperfused tissue, the lactate signal was higher in areas that ultimately infarcted compared with those that recovered (P < 0.0001). Both lactate (P < 0.0001) and N-acetylaspartate (P < 0.001) differed between infarcted and other regions. Within the areas of diffusion-weighted abnormality, lactate was lower where recovery was observed compared with elsewhere (P < 0.001). This feasibility study supports further investigation of fast high-resolution MRSI in acute stroke.

Three-dimensional proton magnetic resonance spectroscopy and diffusion-weighted imaging in the differentiation of incidental prostate carcinoma from benign prostate hyperplasia.

The present study evaluated three-dimensional proton magnetic resonance spectroscopy (MRS) and diffusion-weighted imaging (DWI) features in differentiating incidental prostate carcinoma (IPCa) and benign prostate hyperplasia (BPH) in the central gland of the prostate. The clinical and imaging data of 9 patients with IPCa, 118 patients with BPH [including those with glandular hyperplasia (GH), stromal hyperplasia (SH) and mixed hyperplasia (MH)], were retrospectively analyzed. The mean (choline + creatine)/citrate (CC/C) value of 3D MRS, the apparent diffusion coefficient (ADC) value and the minimal ADC value of DWI were compared between carcinoma and non-carcinoma tissues. The mean CC/C values were 1.04±0.28, and 1.09±0.58 in IPCa and BPH, respectively (t=-0.205, P=0.838). No significant difference in CC/C values (χ2=2.595, P=0.458) could be detected between IPCa, GH, SH and MH groups. The ADC values of the central gland only differed between IPCa (1.48±0.18) ×10-3 and GH (1.60±0.16) ×10-3 mm2/sec (P=0.037). The minimal ADC values were similar between IPCa (1.15±0.10) ×10-3 and BPH (1.14±0.11) ×10-3 mm2/sec, no significant differences could be detected between IPCa and GH (P=0.930), IPCa and SH (P=0.192), and IPCa and MH (P=0.544). Although the ADC values of the central gland of the prostate differed between IPCa and GH, the findings of the present study therefore indicate that combining 3D MRS with DWI cannot potentially improve the detection of IPCa.

Everyman's prostate phantom: kiwi-fruit substitute for human prostates at magnetic resonance imaging, diffusion-weighted imaging and magnetic resonance spectroscopy.

To apply an easy-to-assemble phantom substitute for human prostates in T2-weighted magnetic resonance imaging (T2WI), diffusion-weighted imaging (DWI) and 3D magnetic resonance spectroscopy (MRS). Kiwi fruit were fixed with gel hot and cold compress packs on two plastic nursery pots, separated by a plastic plate, and submerged in tap water inside a 1-L open-spout plastic watering can for T2WI (TR/TE 7500/101ms), DWI (5500/61ms, ADC b50-800s/mm2 map) and MRS (940/145ms) at 3.0T, with phased array surface coils. One green kiwi fruit was additionally examined with an endorectal coil. Retrospective comparison with benign peripheral zone (PZ) and transitional zone (TZ) of prostate (n = 5), Gleason 6-7a prostate cancer (n = 8) and Gleason 7b-9 prostate cancer (n = 7) validated the phantom. Mean contrast between central placenta (CP) and outer pericarp (OP, 0.346-0.349) or peripheral placenta (PP, 0.364-0.393) of kiwi fruit was similar to Gleason 7b-9 prostate cancer and PZ (0.308) in T2WI. ADC values of OP and PP (1.27 ± 0.07-1.37 ± 0.08mm2/s × 10-3) resembled PZ and TZ (1.39 ± 0.17-1.60 ± 0.24mm2/s × 10-3), while CP (0.91 ± 0.14-0.99 ± 0.10mm2/s × 10-3) resembled Gleason 7b-9 prostate cancer (1.00 ± 0.25mm2/s × 10-3). MR spectra showed peaks of citrate and myo-inositol in kiwi fruit, and citrate and "choline+creatine" in prostates. The phantom worked with an endorectal coil, too. The kiwi fruit phantom reproducibly showed zones similar to PZ, TZ and cancer in human prostates in T2WI and DWI and two metabolite peaks in MRS and appears suitable to compare different MR protocols, coil systems and scanners. • Kiwi fruit appear suitable as phantoms for human prostate in MR examinations. • Kiwi fruit show zonal anatomy like human prostates in T2-weighted MRI and DWI. • MR spectroscopy reliably shows peaks in kiwi fruit (citrate/inositol) and human prostates (citrate/choline+creatine). • The kiwi fruit phantom works both with and without an endorectal coil. • EU regulation No. 543/2011 specifies physical and biochemical properties of kiwi fruit.

Treatment Response Assessment in IDH-Mutant Glioma Patients by Noninvasive 3D Functional Spectroscopic Mapping of 2-Hydroxyglutarate.

Measurements of objective response rates are critical to evaluate new glioma therapies. The hallmark metabolic alteration in gliomas with mutant isocitrate dehydrogenase (IDH) is the overproduction of oncometabolite 2-hydroxyglutarate (2HG), which plays a key role in malignant transformation. 2HG represents an ideal biomarker to probe treatment response in IDH-mutant glioma patients, and we hypothesized a decrease in 2HG levels would be measureable by in vivo magnetic resonance spectroscopy (MRS) as a result of antitumor therapy. We report a prospective longitudinal imaging study performed in 25 IDH-mutant glioma patients receiving adjuvant radiation and chemotherapy. A newly developed 3D MRS imaging was used to noninvasively image 2HG. Paired Student t test was used to compare pre- and posttreatment tumor 2HG values. Test-retest measurements were performed to determine the threshold for 2HG functional spectroscopic maps (fSM). Univariate and multivariate regression were performed to correlate 2HG changes with Karnofsky performance score (KPS). We found that mean 2HG (2HG/Cre) levels decreased significantly (median = 48.1%; 95% confidence interval = 27.3%-56.5%;P= 0.007) in the posttreatment scan. The volume of decreased 2HG correlates (R(2)= 0.88,P= 0.002) with clinical status evaluated by KPS. We demonstrate that dynamic measurements of 2HG are feasible by 3D fSM, and the decrease of 2HG levels can monitor treatment response in patients with IDH-mutant gliomas. Our results indicate that quantitative in vivo 2HG imaging may be used for precision medicine and early response assessment in clinical trials of therapies targeting IDH-mutant gliomas.

In Vivo Detection of Choline in Ovarian Tumors Using 3D Magnetic Resonance Spectroscopy

To assess the clinical feasibility of 3-dimensional (3D) proton magnetic resonance spectroscopy (MRS) of ovarian masses at 1.5 T. We prospectively evaluated 16 patients with 23 ovarian masses using contrast-enhanced magnetic resonance imaging and 3D chemical shift imaging MRS (time of reception/time of echo = 700/135 ms, number of excitations = 6, interpolated voxel = 5 × 5 × 5 mm(3), water and fat suppression). Spectral editing consisted of water reference, filtering, zero-filling, Fourier transformation, frequency shift, automatic baseline and phase correction, and curve fitting. The volume of interest was placed to encompass both solid and cystic tumor components as well as apparently healthy pelvic tissues. The presence of a choline peak at 3.14 to 3.34 ppm was considered as a marker of malignancy. All patients underwent surgery and histopathological evaluation. Of 23 masses, 19 were malignant and the remaining 4 benign lesions were a fibrothecoma, an endometriosis, a cyst, and a cystadenofibroma. A choline peak was detected in 17/19 malignant tumors (sensitivity 89%), absent in 2 G1 tumors. It was visible in 16 solid components of 19 malignant tumors (in one of them, a choline peak was detected only in the cystic component, in 6 in both solid and cystic components). The choline peak was absent in 20/21 apparently healthy pelvic tissues, with a very low choline peak being detected in one intraperitoneal fluid collection with malignant cells at cytologic analysis; 3/4 benign tumors showed a choline peak (overall specificity 21/25 = 84%). A significant difference between the mean choline peak integral detected within the solid component and that within the cystic component was observed (P = 0.002). No correlation between the choline peak integral and the tumor size was found (r = 0.120, P = 0.615). 3D MRS of ovarian masses is clinically feasible at 1.5 T. This opens new research strategies for early diagnosis of ovarian cancer.

61 poster: Integration of 3D Magnetic Resonance Spectroscopy Maps into Treatment Planning of Glioblastoma

61 poster: Integration of 3D Magnetic Resonance Spectroscopy Maps into Treatment Planning of Glioblastoma

Combined magnetic resonance imaging and spectroscopy in the assessment of high grade prostate carcinoma in patients with elevated PSA: A single-institution experience of 356 patients

Purpose To assess the ability of combined whole-prostate magnetic resonance imaging and magnetic resonance spectroscopy imaging (MRI + MRSI) to predict the presence or absence of high grade (Gleason 4 + 3 or higher) prostate carcinoma in men with elevated PSA. Materials and methods Between March 2002 and September 2007, 356 subjects (mean serum PSA 11.5 ng/ml, range 0.4–133.0 ng/ml) were examined with fast-T2-weighted magnetic resonance imaging (MRI) and 3D-magnetic resonance spectroscopy imaging (MRSI) on a 1.5 T scanner. Prostate cancer was histopathologically proven in 220 patients (41 with high grade and 179 with lower grade cancer) and non-evidence of cancer was determined after at least 12 months (mean 21 months) clinical follow-up in 136 subjects. The sensitivity, false positive rate, and negative predictive value of MRI + MRSI were calculated using histopathology and follow-up results as reference standard. Results MRI + MRSI had a significantly higher sensitivity for high grade tumors (92.7%) than for lower grade tumors (67.6%), and was false positive in only 7.4% of patients with non-evidence of prostate cancer. For exclusion of a high grade tumor, MRI + MRSI had a negative predictive value of 98.4%. Conclusions MRI + MRSI holds great potential for predicting presence or absence of high grade tumors in men with elevated PSA. This can be important in the selection of patients for active surveillance, or in the decision to rebiopsy patients with prior negative biopsies.

A qualitative approach to combined magnetic resonance imaging and spectroscopy in the diagnosis of prostate cancer

Purpose To investigate the feasibility and diagnostic value of a whole prostate qualitative approach to combined magnetic resonance imaging and spectroscopy (MRI + MRS) in the detection of prostate cancer in patients with elevated PSA. Materials and methods Three hundred and fifty six subjects (mean serum PSA 11.47 ng/ml, range 0.40–133 ng/ml) were examined with fast-T2-weighted images (MRI) and 3D-magnetic resonance spectroscopy (MRS). Both modalities were qualitatively analyzed on a whole prostate basis by a single radiologist using a 4-point diagnostic scale. Prostate cancer was histopathologically proven in 220 patients and non-evidence of cancer was determined after at least 12 months clinical follow-up in 136 subjects. Results Receiver operating curve analysis revealed a significantly better diagnostic performance of MRI + MRS ( A z = 0.857) than MRI alone ( A z = 0.801) and MRS alone ( A z = 0.810). The sensitivity, specificity and accuracy of MRI + MRS for detection of prostate cancer were 72.3%, 92.6%, and 80.1%, respectively. Conclusions Spectral evaluation with a whole prostate qualitative approach is feasible in routine clinical practice. The combination of MRI and MRS yields superior diagnostic results than either modality alone.

Potential Value of MR Spectroscopic Imaging for the Radiosurgical Management of Patients with Recurrent High-Grade Gliomas

Previous studies have shown that metabolic information provided by 3D Magnetic Resonance Spectroscopy Imaging (MRSI) could affect the definition of target volumes for radiation treatments (RT). This study aimed to (i) investigate the effect of incorporating spectroscopic volumes as determined by MRSI on target volume definition, patient selection eligibility, and dose prescription for stereotactic radiosurgery treatment planning; (ii) correlate the spatial extent of pre-SRS spectroscopic abnormality and treatment volumes with areas of focal recurrence as defined by changes in contrast enhancement; and (iii) examine the metabolic changes following SRS to assess treatment response. Twenty-six patients treated with Gamma Knife radiosurgery for recurrent glioblastoma multiforme (GBM) were retrospectively evaluated. All patients underwent both MRI and MRSI studies prior to SRS. Follow-up MRI exams were available for all 26 patients, with MRI/MRSI available in only 15/26 patients. We observed that the initial CNI 2 contours extended beyond the pre-SRS CE in 25/26 patients ranging in volume from 0.8 cc to 18.8 cc (median 5.6 cc). The inclusion of the volume of CNI 2 extending beyond the CE would have increased the SRS target volume by 5-165% (median 23.4%). This would have necessitated decreasing the SRS prescription dose in 19/26 patients to avoid increased toxicity; the resultant treatment volume would have exceeded 20cc in five patients, thus possibly excluding those from RS treatment per our institutional practice. MRSI follow-up studies showed a decrease in Choline, stable Creatine, and increased NAA indicative of response to SRS in the majority of patients. When combined with patient survival data, metabolic information obtained during follow-up MRSI studies seemed to indicate the potential to help to distinguish necrosis from new/recurrent tumor; however, this should be further verified by biopsy studies.

3D Magnetic Resonance Spectroscopy Without Endorectal Coil in Prostate Cancer Patients Undergoing Androgen Deprivation and External Beam Radiation Therapy

Sensitivity and specificity of 3D MRS [magnetic resonance spectroscopy] for prostate studies increased significantly in the last few years causing interest for its potential role for definition of target sub-volumes at higher risk of failure following radical radiotherapy. Similarly to other tumour sites, prostate MRS can also be a non-invasive predictive factor of treatment outcome.

Prostate Cancer: Sextant Localization with MR Imaging, MR Spectroscopy, and11C-Choline PET/CT

To retrospectively compare sensitivity and specificity of magnetic resonance (MR) imaging, three-dimensional (3D) MR spectroscopy, combined MR imaging and 3D MR spectroscopy, and carbon 11 (11C)-choline positron emission tomography (PET)/computed tomography (CT) for intraprostatic tumor sextant localization, with histologic findings as reference standard. The local ethics committee on human research provided approval and a waiver of informed consent for the retrospective study. MR imaging, 3D MR spectroscopy, and 11C-choline PET/CT results were retrospectively reviewed in 26 men with biopsy-proved prostate cancer (mean age, 64 years; range, 51-75 years) who underwent radical prostatectomy. Cancer was identified as areas of nodular low signal intensity on T2-weighted MR images. At 3D MR spectroscopy, choline-plus-creatine-to-citrate and choline-to-creatine ratios were used to distinguish healthy from malignant voxels. At PET/CT, focal uptake was visually assessed, and maximum standardized uptake values (SUVs) were recorded. Agreement between 3D MR spectroscopic and PET/CT results was calculated, and ability of maximum SUV to help localize cancer was assessed with receiver operating characteristic analysis. Significant differences between positive and negative sextants with respect to mean maximum SUV were calculated with a paired t test. Sensitivity, specificity, and accuracy were, respectively, 55%, 86%, and 67% at PET/CT; 54%, 75%, and 61% at MR imaging; and 81%, 67%, and 76% at 3D MR spectroscopy. The highest sensitivity was obtained when either 3D MR spectroscopic or MR imaging results were positive (88%) at the expense of specificity (53%), while the highest specificity was obtained when results with both techniques were positive (90%) at the expense of sensitivity (48%). Concordance between 3D MR spectroscopic and PET/CT findings was slight (kappa=0.139). In localizing cancer within the prostate, comparable specificity was obtained with either 3D MR spectroscopy and MR imaging or PET/CT; however, PET/CT had lower sensitivity relative to 3D MR spectroscopy alone or combined with MR imaging.

Contribution of the MR spectroscopic imaging in the diagnosis of prostate cancer in the peripheral zone

To establish the additional value of 3D magnetic resonance spectroscopy (3D-MRS) imaging to endorectal MR imaging in the diagnosis of prostrate cancer in the peripheral zone. MR imaging and MRS imaging were performed in 79 patients with suspicion of prostate cancer on the basis of digital rectal exploration, transrectal ultrasound and PSA level. All the examinations were performed with 1.5 T MR scan using an endorectal coil (transverse and coronal FSE T2-weighted sequences, axial SE T1-weighted and PRESS 3D CSI). MR examinations have been evaluated by two Radiologists blind of the clinical data in a "per patients" analysis. MR imaging and MRS imaging findings were compared with the result of histological data from radical prostatectomy in 53 patients and biopsy in 17 patients. Nine patients (11.4%) were excluded because of serious artefacts in the MR spectrum. The reported values of sensitivity, specificity, PPV and NPV for MR imaging alone were respectively 84%, 50%, 76% and 63% (LR+ 1.7; LR- 0.3). Instead the reported values of sensitivity, specificity, PPV and NPV for the combination of MR imaging to MRS imaging were respectively 89%, 79%, 89% and 79% (LR+ 4.28; LR- 0.14). We found an incremental benefit of MRS imaging to MR imaging for tumour diagnosis although these results did not show statistically significant differences. The MRS imaging improves the accuracy of the endorectal MR imaging in the diagnosis of prostate cancer.

Histopathological validation of a three-dimensional magnetic resonance spectroscopy index as a predictor of tumor presence.

Data obtained preoperatively from three-dimensional (3D)/proton magnetic resonance (MR) spectroscopy were compared with the results of histopathological assays of tissue biopsies obtained during surgery to verify the sensitivity and specificity of a choline-containing compound-N-acetylaspartate index (CNI) used to distinguish tumor from nontumorous tissue within T2-hyperintense and contrast-enhancing lesions of patients with untreated gliomas. The information gleaned from the biopsy correlation study was used to test the hypothesis that there is metabolically active tumor in nonenhancing regions of the T2-hyperintense lesion that can be detected using MR spectroscopy. Patients suspected of harboring a glioma underwent 3D MR spectroscopy during their preoperative MR imaging examination. Surgical navigation techniques were used to record the location of tissue biopsies collected during open resection of the tumor. A receiver operating curve analysis of the CNI and histological characteristics of specimens at each biopsy location was performed to determine the optimal threshold of the CNI required to separate tumor from nontumorous tissue. Histograms of the CNIs within enhancing and nonenhancing regions of lesions appearing on MR images were generated to determine the spatial distribution of CNIs consistent with tumor. Biopsy samples containing tumor were distinguished from those containing a mixture of normal, edematous, gliotic, and necrotic tissue with 90% sensitivity and 86% specificity by using a CNI threshold of 2.5. The CNIs of nontumorous specimens were significantly different from those of biopsy specimens containing Grade II (p < 0.03), Grade III (p < 0.005), and Grade IV (p < 0.01) tumors. On average, one third to one half of the T2-hyperintense lesion outside the contrast-enhancing lesion contained CNI greater than 2.5.