Figure 1: M₀ (left) and T₁ (right) maps for three representative patients (F71, F61, F67) at the first time point. Maps are volumetric, and axial, coronal, and sagittal slices centered on the tumor section are shown. Tumor ROIs are delineated by red contours on M₀ maps. Longer T₁ can be observed in corresponding regions on the T₁ maps. In addition, T₁ maps reveal the locations of surgical cavities and extra-axial fluid collection (blue arrows) and evidence of craniotomy (green arrows). White dashed-line boxes indicate the zoomed-in region in Figure 2.

Figure 2: Closeup of T₁ maps from the axial, coronal and sagittal views on three patients shown in Figure 1. Maps are zoomed into the tumor region (delineated by white dashed lines in Figure 1), with constrained display range. The proposed method captures T₁ heterogeneity. Light green: high T₁ values, green: medium T₁ values and dark green: low T₁ values. Coefficients of variation for T₁ inside the ROIs are 10.84%, 9.96%, and 7.31% for the top, middle, and bottom rows, respectively.

Fig. 1 Study workflow. 144 patients with osteosarcoma were randomly split into training (n = 101) and test (n = 43) dataset. Images were segmented manually. Radiomics features were extracted with Pyradiomics and data in training dataset were balanced with upsampling. Radiomics model building was performed with Pearson Correlated Coefficient for dimension reduction, RFE or Relief for feature selection, and SVM or LR for classifier. The final model was evaluated using ROC analysis, radiomics score plot, calibration curve, and decision curve analysis.

Fig. 2 Clinical-radiomics combined nomogram and its utility. A nomogram (a) integrated the radiomics score and three clinical variables was constructed with an AUC of 0.838 in test dataset (b). The probability of pGR for each patient in terms of the response status in the training (c) and test dataset (d) suggested a good accuracy in response prediction. Decision curve analysis curves for clinical, radiomics and combined model in the whole dataset showed that the nomogram had a favorable clinical utility (e).

Figure 1. Image processing steps and radiomics flowchart

(a) Image acquisition of T1cwc, T1w and T2w images, registration of T1w and T2w to T1c images and intensity normalization. (b) ROI delineation for the treatment planning of GKRS by experienced neurosurgeons and radiologists on all MRI slices covering tumor regions. (c) Extraction of radiomic features from the tumor ROIs on the MRIs.

Figure 2. Representative cases and the model performance in predicting local tumor control

(a) T1c images and reconstructed 3D tumor models. (b) The SVM scores in two representative cases based on the combination of radiomic and clinical features. (c) Receiver operating characteristic curves and the area under the curves (AUCs) of three SVM classification models.

Table 2. Confusion matrices for reader scoring agreements on 99 from 410 co-identified nodes at 3T and on 159 from 601 co-identified nodes at 7T.

Figure 2. Histogram of the size of annotated nodes by Reader A, Reader T and concordant Reader A & T for both 3 and 7 Tesla USPIO-enhanced MRI of patients with prostate cancer.

Figure 1. Flowchart of the whole process.

Figure 3. a: The average ROC curves of a representative 5-fold stratified cross-validation from radiomic analysis (red line) and coventional mean APTw signals (blue line). b: The pooled ROC curves (mean ± standard deviation) from 100 runs of the 5-fold stratified cross-validation, in which the red curves referred to results from radiomic analysis and blue curves denoted results from conventional mean APTw signals.

Figure 2. DMI of [²H₉]-Cho in rats with RG2 glioma. A, B) Contrast-enhanced T₁w MRI, C-G) DMI maps acquired during a ~36 min IV infusion of [²H₉]-Cho, showing the high signal in the tumor lesion in contrast to the surrounding normal brain. D-H), DMI maps acquired 20-24 hrs after a ~36 min IV infusion of [²H₉]-Cho; note that panel A-D are from the same animal scanned on consecutive days. DMI amplitude based on peak integral of tCho signal, in a.u., and normalized to the highest value. I, J) ²H MR spectra from a voxel in the tumor, acquired during and after ²H-choline infusion, respectively.

Figure 3. High resolution ²H NMR. Spectra acquired in metabolite extracts from excised tumor tissue, harvested immediately at the end (top), and 24 hrs after a 36 min infusion of [1,1,2,2-²H₄]-Cho (bottom), Note the lack of free Cho after 24 hrs, indicating active metabolism of the blood-borne ²H-labeled Cho. Cho: choline, PC: phosphocholine, GPC, glycerophosphocholine. Note that spectra are from samples of different weights, and thus peak amplitudes are not necessarily quantitatively comparable.

Figure 2. Left: Representative MRS voxel selection on a T1WI post-contrast image, tumor voxels (red), peritumoral (green), and contralateral normal voxels (blue) were selected for analyses.

Right: mI/c-Cr ratios at baseline in the tumor, (T) periphery (P), and contralateral (C) volumes of interest (VOI). Box plots indicate quartiles. Mean and standard deviations of mI/c-Cr in each VOI with t-tests indicating each VOI pair is significantly different.

Figure 3. Mean and standard deviations of mI/c-Cr classified by OS9. the asterisk (*) indicate p <0.05.

Lactate production from [U-²H]pyruvate is localized to the tumor region in vivo. (A) Anatomical T2-weighted MRI of a mouse bearing an orthotopic BT88 tumor xenograft. Representative ²H-MR spectra from contralateral normal brain (B) and tumor (C) voxels at the first time point after injection of [U-²H]pyruvate in a mouse bearing an orthotopic BT88 tumor. Metabolic heatmap of the SNR of lactate (D) and the ratio of lactate to post-injection HDO (E) in mouse bearing an orthotopic BT88 tumor. The tumor is delineated by white line.

Lactate production from [U-²H]pyruvate is higher in orthotopic glioma-bearing mice in vivo. Lactate signal normalized to the ratio of post-injection HDO at each time point to pre-injection HDO is higher in mice bearing orthotopic glioma xenografts relative to tumor-free controls. ** refers to p<0.01, *** refers to p<0.001.

Figure 4: Representative axial T2-weighted images of control BT257 (A) and SF10417 (B) LGG-bearing mice and corresponding in vivo ¹H MRS spectra (black) acquired from 2x2x2mm³ voxel marked in blue as well as the LCModel fit used to quantify the spectra (red) (A, B). Quantification of 2HG, Glu and GLX in control, AG-881- and BAY-1436032-treated tumor voxel at D7 and D15 in BT257 (C) and SF10417 (D) LGG-bearing mice.

Figure 2: Representative axial T2-weighted images of control, AG-881 and BAY-1436032-treated BT257 (A) and SF10417 (B) LGG-bearing mice at D0, D7 and D15. Temporal evolution of average tumor volume shown as a percentage of D0 in BT257 (C) and SF10417 (D) LGG models.

Figure 1 ²H MR spectroscopic measurements of labeled fumarate, malate and water concentrations in A11 (A, B, E,F) and U87 (C,D,G,H) tumors. Tumor spectra were acquired before and 7 days after targeted radio-chemotherapy (10 Gy in total, temozolamide 100 mg/kg). (E – H) Sum of 12 ²H spectra recorded over 60 minutes. The [2,3-²H₂]fumarate injection (1 g/kg) started 5 min after the start of acquisition of the first spectrum. The peaks were fitted individually using prior knowledge.

Figure 2 Metabolite concentration maps derived from dynamic 3D CSI images summed over 60 min of signal acquisition following fumarate injection into A11 tumor-bearing mice. The color code represents concentration in mM derived from the ratios of the peak intensities in the malate and fumarate maps to peak intensities in an initial HDO map and corrected for the number of ²H labels per molecule and signal saturation. (A,D) T₂-weighted axial slices from a reference ¹H image. Concentration maps of (B) fumarate and (C) malate pre-treatment; (E) fumarate and (F) malate 7 days post-treatment.

Fig. 3 HP [1-¹³C] Pyruvate MRI showed decreased Lac/Pyr ratio after PEGPH20 treatment.

HP [1-¹³C] Pyruvate MRI was performed in BxPC3-HAS3 tumor bearing mice treated with control buffer or PEGPH20 before and after treatment. A, Representative kinetics of [1-¹³C] pyruvate and [1-¹³C] lactate and its anatomical ¹H image. B, Lac/Pyr ratio of each tumor are shown by treatment group. Individual values are shown. C, Lac/Pyr ratio change from the baseline (pre-treatment) of each group.

Fig. 1 EPR imaging showed increased pO₂ after PEGPH20 treatment.

EPR imaging was performed in BxPC3-HAS3 tumor bearing mice treated with control buffer or PEGPH20 before and after treatment. A, Representative oxygen map obtained by EPR imaging and T2-weighted anatomical image. B, Histograms of pO₂ distribution within the tumor of each group. C, The mean pO₂ of each tumor shown by treatment group. D, pO₂ change (%) from the baseline (pre-treatment) of each group.

Permeability of Gd-DPTA into SU.86.86 and MIA Paca-2 tumors before treatment on Day 1 and 1 hour after treatment on Day 5 evaluated by DCE-MRI. (A) (B) (D) (E)Kinetics of the Gd-DTPA incorporation into SU.86.86 or MIA Paca-2 tumors treated with combination of evofosfamide and GEM or vehicle. (C) Comparison of Ktrans in combination of evofosfamide and GEM and control treated tumor in SU86.86 tumor (*P=0.041). (F) Comparison of Ktrans in combination of evofosfamide and GEM and control treated tumor in MIA Paca-2 tumor (***P<0.001).

Blood volume (%) changes calculated using MRI with the blood pooling T2 contrast agent USPIO. (A) (B) (C) (D) The images of blood volume on Day 1 and Day 5 in SU.86.86 tumors. (E) Comparison of blood volume on Day 5 relative to Day1 in SU.86.86 tumors (**P=0.004). (F) (G) (H) (I) The images of blood volume on Day 1 and Day 5 in MIA Paca-2 tumors. (E) Comparison of blood volume on Day 5 relative to Day1 in MIA Paca-2 tumors (*P=0.036).

Schematic illustration of experimental flow. Schematic illustration of SPIO@NC synthesis and experiments in mice in vitro and in vivo.

Fig. 4 (A)(B)Representative flow cytometric analysis displaying the absolute percentage of (A) M1 (MHC-II+) and M2 (CD206+) macrophages in CD45+CD11b+ tumor cells, and (B) CD8+ T cells and CD4+ T cells in CD45+CD3+ tumor cells from mice with different treatments at 7 days. (n = 5). (C) Representative immunofluorescence imaging of Treg cells (CD4+FoxP3+) in tumor from mice received various treatments on 7th days. Scale bar, 30μm. (D)(E) Real pictures of tumors(D) and (E) after dissection from mice received different treatments.

Figure 4. . MR images of metastatic mice models with ProCA32.CXCR4 administration. A. Comparison of MRI images of metastatic mice models including M20-09-196, OMM2.3, and OCM1 before and after administration of ProCA32.CXCR4. B. Zoom-in view of the metastases from M20-09-196, OMM2.3, and OCM1 mouse models; MRI signal-noise-ratio (SNR) of metastases following the ProCA32.CXCR4 administration.

Figure 2. A Detection of UM liver metastasis using ProCA32.collagen (top), non-targted ProCA32 (middle) and Eovist (bottom) at 7T. Detected liver metastasis are verified by H&E and Sirius red staining (B) and UM markers HMB45 and S100 staining (C). The tumor liver contrast to noise ratio is further increased by inversion recovery pulse sequence showed in the figure on right.

Figure 3. Representative HP ¹³C MRS of (A) vehicle and (B) ascorbate-treated KRAS PDX model of pancreatic cancer. Metabolic conversion of DHA to ascorbate in pancreatic cancer PDXs changes after 1 week treatment of twice daily ascorbate.

Figure 2. Tumor growth curves for patient derived xenograft models of KRAS and BRCA driven cancer. After 2 weeks of tumor implantation, mice were randomly divided into two groups. One group was treated with freshly prepared vitamin C in 400 μL of PBS (4 g/kg) twice a day via IP injection (KRAS, n= 7; BRCA, n = 7). Control group mice were treated with PBS using the same twice a day dose (KRAS, n = 7; BRCA, n = 7). Tumor growth change for the KRAS PDX tumors is significant after 21 days and significant after 4 days for the BRCA PDX tumors.

Figure 4: T1-T2-weighted images and T1 T2 maps from conventional methods and from Multitasking acquired on an HFD mouse at week 2 with ProCA.collagen1 injection. Multitasking consistently showed improved image quality. The tumor were labeled by red dashed boundary on T2W TSE. Tumor showed higher T1 and T2 values compared to normal liver parenchymal, which was consistent with conventional methods and Multitasking. Some breathing artifacts on conventional images were labeled by white arrows

Figure 3: T1-T2-weighted images and T1 T2 maps from conventional methods and from Multitasking acquired on an LFD mouse at week 1 with Eovist injection. In each imaging session, conventional image series took about 45 to 50 minutes in total, while Multitasking can generate T1-T2-weighted images and T1 T2 maps in one single 10-min scan. Moreover, Multitasking produced improved image quality in both weighted images and quantitative maps. Some breathing artifacts on conventional images were labeled by white arrows

DMI data collected at the indicated stages following the intravenous administration of ²H_6,6’-glucose to a pancreatic cancer mouse. Metabolic maps of ²H_6,6’-glucose (A) and its metabolic products ²H_3,3’-lactate (B) and ²H-water (C) are here shown as absolute concentration colormaps. The anatomical ¹H image on top of which all ²H data are shown is depicted in (D). Concentrations for glucose and water (E) and for the lactate (F) are shown for the entire time series.

²H NMR spectra arising at the indicated times after an intravenous ²H_6,6’-glucose administration into a pancreatic cancer mouse model. (B) and (C) show organ-specific ²H spectra extracted from the CSI data at the indicated sites. Signals for ²H_6,6’-glucose and its metabolic products water and lactate are indicated by letters G, W and L respectively.

Representative mouse composite magnitude images on Day 0, 3, and 6 after intratumoral injection of 5.3 x 10⁵ NK cells into syngeneic EL4 lymphomas. Images were scaled against their own noise to place into units of pixel SNR. The ¹⁹F signal mitigating from labeled cells is detectible within the tumor in all three images. White arrow indicates the PFPE reference vial of known spin density.

In vivo NK cell quantification results indicate the calculated number of PFPE-red labeled GFP+ NK cells within the tumor volume of each mouse. Difference in time points are not statistically significant.

Figure 2: Illustrative examples of baseline (a) and post-injection (b) AACID maps; baseline (c) and post-injection (d) acidoCEST maps. All maps are overlaid on top of the corresponding T₂-weighted MRI. The tumour region is delineated with the red-dotted line, and the peri-tumour region is contoured with the light-blue-dotted line using the T₂-weighted MRI.

Figure 1: An illustrative example of the FDG-PET image of the last frame (SUV, a), time-activity curve from the tumour region (b) delineated with red-dotted line, and the arterial input function (AIF) from the left ventricle of the heart delineated with white-dotted line (c, d). The axial SUV map is overlaid on top of the corresponding T₂-weighted MRI.

Figure 1. Representative ¹H MRS spectra of control and treated NHAIDH1mut cells.

Figure 3. (A) shows fluxed from the hyperpolarized [1-¹³C]α-ketoglutarate to glutamate between control (blue) and the BAY-1436032 treated (orange) cells. A significant increase in glutamate was observed in the BAY-1436032 cells (B)

Figure 4: Summed ³¹P spectra from the muscle (left) and tumor ROI (right) with the corresponding fitted signal (red line) acquired with protocol 2 (Fig.3). Acquisition parameters: TR=300ms, α=45°, Δf=6060 Hz, 1024 time points, postprocessing with a 40-Hz Gaussian filter in time domain. The following resonances were resolved: PCr, ATP, P_i, PC, and Phosphoethanolamine (PE). Note the different scales of the y-axis in both spectra.

Figure 3: Transversal map of the fitted PCr amplitude overlaid on the morphological ¹H image in the slice covering the tumor. Data was acquired with protocol 2: matrix size=10x10, FOV=(25x25)mm², Hamming-weighted k-space averaging with 400 central averages, postprocessing with zerofilling-factor 2. The tumor ROI (blue) was drawn on the marginal part of the tumor to reduce signal contamination from muscle. The muscle ROI (green) includes the same number of voxels as the tumor ROI.

Figure 1. Anatomical T₂-weighted (T₂w) MRI and parametric maps of G_d, G_l, |G*| and Y for representative 4T1, 4T1/HAS3 and MDA-MB-231 LM2-4 tumours prior to and 24 hours after treatment with PEGPH20 (1 mg/kg). The tumour is delineated by a white dashed line.

Figure 3. Anatomical T₂-weighted (T₂w) MRI and aligned tissue sections stained with H&E, HTI-601 (hyaluronan/HA) and picrosirius red (collagen I & III). Representative saline and PEGPH20 treated tumours are shown for each tumour model (4T1, 4T1/HAS3 and MDA-MB-231 LM2-4). The PEGPH20 treated tumours are the same as those shown in Figure 1. The whole section images highlight the close matching of the histology with MRI. Representative high-power images (20x) are inset next to each whole tumour section.

Figure 4. ¹³C Isotopomer modeling of [U-¹³C] glucose-labeled of PDX047 and TSC047 using tcaCALC software. (PDH - Pyruvate dehydrogenase, PK-Pyruvate Kinase, Y-denotes anaplerotic denotes, Ypc - Pyruvate Carboxylase and Ys- anaplerosis via succinyl CoA, glutaminolysis)

Figure 3: Comparison of (A) steady state concentration and (B) Fractional enrichment of glutamate among PDX model, while (C) & (D) represent steady state concentration and Fractional enrichment of glutamate, respectively among TSCs model. *p<0.05, **p<0.005, ***p< 0.0005, n=3 for all groups except TSC072 (n=2)

Figure 3: Visualization of the data.

Table 1: Performance metrics of the default models in the modules in Normal User part (*Due to class imbalance, synthetic data was generated with ADASYN)

Figure 3. T₂w images and paramagnetic maps of baseline R₂*, USPIO-induced ΔR₂* and fractional blood volume (fBV) from representative athymic nude mice pre and post 48h treatment with vehicle, 80mg/kg DZB daily or 50mg/kg vatalanib twice daily.

Figure 4. Quantification of baseline R₂* and fractional blood volume (fBV) in all mice pre and post 48h treatment with vehicle, 80mg/kg DZB daily or 50mg/kg vatalanib twice daily. DZB and vatalanib induced significant reductions in fBV (p<0.05, 2-way repeated measures ANOVA).

Fig 1: Model fitting using A. extended Tofts and B. SSM in PCNSL. The same is shown for GBM in C and D. The original AIF curve is shown in yellow and its biexponential fit shown in black. The tissue uptake is shown in blue with its fitting shown in red.

Fig 2: K^trans maps from the extended Tofts model from a patient with PCNSL (A), Gr III Glioma (B), GBM (C) and Metastasis (D). Box plots demonstrating the K^trans (min^-1, E), v_e (F) and v_p (G) values using extended Tofts model demonstrate highest values in metastasis, while PCNSL patients demonstrate lower values.

(a) A patch of DCE data has been rearranged to 2-dimensional matrix X = (n t), where n is the number of voxel (9 in our design) and t is the number of temporal frames. (b) Schematic diagram for the deep learning network. The total of twelve 2D-convolutional layers were connected in series and each convolutional layer is linked with Rectified Linear Unit activation function. After two layers, a skipping connection was made and filters were concatenated together, just like in the Residual Network

Receiver Operating Characteristic (ROC) curve for the estimated F_p from the PKM analysis of the clinical data. F_p estimation from 3 approaches using (1)the case-specific AIF(C_a), (2)the population-averaged AIF, and (3)the predicted Cp with the population-averaged AIF(C_p+C_a,pop) was used to assess the diagnostic performance in predicting the malignancy of breast cancer. F_p estimation from our proposed model yielded the highest AUC.

Figure 1: Representative 1H MR spectra acquired from aqueous phase tissue extracts of spleen obtained from normal mice and cachectic (Pa04C ) and non-cachectic (Panc1) tumor bearing mice.

Figure 2: Lung, liver, heart, kidney and spleen weights in control mice (black) as compared with Panc1 (green) and Pa04C (red) tumor‐bearing mice. In all panels, means of cohorts are shown as rectangles bar. P-value less that ≤ 0.05 are considered statistically significant.

Illustration showing the procedure of generating oversampled proton magnetic resonance spectroscopy (¹H-MRS) from the raw ¹H-MRS by using Wavelet OverSampling (WvOS). Abbreviations: SNR, signal-to-noise ratios; ψ, wavelet basis.

Boxplots showing the balanced classification accuracy for the 1.5T cohort of childhood brain tumours derived through linear discriminant analysis (A, C) or supper vector machine (B, D) with leave-one-out (A-B) or six-fold (C-D) cross validation.

Our proposed Resnet-based convolutional neural network architecture applied to classify five tissue types in diffuse glioma.

T1-w image of a patient with a diffuse glioma (a). Chemical Shift Image (CSI) of the patient (b). Rotated overlaid image of CSI grid on T1-w image (c). color coded grid of CSI in order to exact localization of the signals (d).

Fig-2:Processing pipeline implemented for radiomics analysis and classification of tumor subgroups

Fig-5: Feature importances obtained using SVM coefficient scores on multimodal feature set

Figure 1: Processing pipeline for radiomics analysis and classification for histone H3K27M mutation

Figure 4: Box plot for top 10 most important features obtained using random forest classifier

Example VFA and MOLLI T1 maps for 6 subjects (displaying a comparable slice). The field-of-view of the images has been cropped for ease of display.

(A-B) The mean T1 in the tumour ROI of all subjects. The error bars represent the standard deviation of T1 values within the tumour ROI. (C) The mean VFA T1, normalized by the respective mean MOLLI T1 for each subject to illustrate the relative change in T1 clearly.

Figure 1 A 60-year-old male with localized lesions PCa.ROI was manually placed on the local lesion of DCE map (A). The Ktrans, Kep and Ve maps of TM were shown(B-D).

Figure 2.ROC curves show performangces of Ktrans individually and combination of age and PSA to differentiate between oligometastatic and widely metastasis PCa.

Fig. 3. The structure of ANN model

Fig. 4 Diagnostic performance of ANN model for identifying the CK-19 status.

Fig 1, male, 56 years old with gastric cancer, (A) image acquired by TSE-HASTE with a subjective score of 5 points; (B) image acquired by conventional HASTE sequence sequence with a subjective score of 5 points; (C) image acquired by TSE- BLADE sequence with a subjective score of 3 points; (D) image acquired by conventional TSE sequence with a subjective score of 3 points.

Fig 2, male, 63 years old with gastric cancer with irregular breathing rhythm, (A) image acquired by TSE-HASTE with a subjective score of 5 points; (B) image acquired by conventional HASTE sequence sequence with a subjective score of 4 points; (C) image acquired by TSE- BLADE sequence with a subjective score of 2 points; (D) image acquired by conventional TSE sequence with a subjective score of 1 points.

Figure 1. Receiver operating curves (ROC) for three models using radiomics and results of deep learning and human prediction. DL1: deep learning model using a single T2w slice; DL2: deep learning model using a T2w slice and its tumor and kidney masks. ccLS 4/5 and ccLS 3/4/5: clear cell likelihood score 4 and 5, or 3, 4 and 5 from human readers based on multiparametric MRI.

Table 2

Figure 1 A 52-year-old male patient with the rectal tubular adenocarcinoma. T2 image (a), Ktrans image (b), Kep image (c) and Ve image (d) were showed.

Figure2 Comparison of ROC Curves of the parameters with differences between the Ki67 high/low differentiation,ki67 analyzes the texture analysis of the three post-processing images of Ktrans, Ve, and Kev. After the statistically different parameters are combined, the combined diagnostic performance is obtained,the AUC was 0.923, with the sensitivity of 90.5% and the specificity of 50%.

Table 1 Baseline characteristic of included studies

Fig. 3 Forest plot of single studies for the pooled DOR and a represents EP vs non-EP, b MB vs non-MB, c PA vs non-PA.

Table 1. The testing dataset results. Mean Dice, PPV, sensitivity, and Hausdorff are shown in the table. EN is enhancing tumor core. WT is the whole tumor part. And the TC is tumor core.

Figure 2. A typical segmentation example with ground truth and predicted labels on 2D MRI slices. The whole tumor (WT) part contains all visible labels (all green, red and yellow labels), the tumor core (TC) part contains red and yellow labels, and the enhancing tumor core (ET) class is shown in yellow.

Figure 4. ADC changes in muscle tissue (top) and Bland-Altman plot of ADC repeats differences at muscle.

Figure 3. Bland-Altman plot describing median ADC repeatability. Each marker/color represents a baseline repeat difference at each tumor. RC obtained per tumor is used to distinguish true changes in that tumor. In case the patient does not have more than one repeat at baseline, the full data RC given by the dashed lines is used instead.

table

table

Figure 1. T2-Flair, MP2RAGE, and ATP were acquired from Subject 9. The images in the top row is from the higher grade (WHO Grade IV) lesion. The results of the lower grade (WHO Grade II) lesion are shown in the bottom row. Larger APT values can be found in the high-grade region. Region of interest (ROI) was defined within the glioma lesion, and the statistical analysis was performed based on the ROIs.

Figure 2. The group statistic using APT-CEST signal intensities obtained from all three grades of glioma were shown. (A) The APT-CEST signal intensity of high-grade (III and IV) glioma is significantly higher than low-grade (II) glioma (p-value <0.05). There was a significantly lower APT-CEST signal in WHO grade II compare to WHO grade IV (p-value <0.05). (B) The IDH mutation status in patient group was classified by WHO.

Table 1: The p values obtained using a Mann-Whitney U test for assessing the CRLB differences between several mutational subgroups of gliomas. (*p<0.002, Bonferroni multiple comparison correction)

Table 2: The classification results of the models for the detection of IDH and TERTp mutations based on different CRLB values. ‘No-FS’ means that all the metabolites were included in the classification.

Figure2. Magnetic resonance images showing a 57-year-old male patient with rectal cancer. The tumor is isointense on T1-weighted MR image, mildly hyperintense on T2-weighted MR image, hyperintense on DWI (b = 1000 s/mm²) , and isointense on ATP image (d).

Table1. Correlations Between Mean ATP SI(%) with Clinicopathologic Characteristics.

Figure 1. MR spectroscopic data with some of the LCModel results of grade I (a) and grade II (b) meningioma.

Table 1. The metabolite peak intensity differences between different grades of meningiomas and the p-values of a Mann-Whitney U test.

Figure 1. a: Isodose curves of a treatment plan of a LGG patient with a Left Frontal tumor on Prowess Panther 5.5; b: An example of MRS metabolite peaks of a voxel on the CC, fused with T2W MRI on a 1.5 Tesla Siemens Magnetom Aera scanner.

Figure 2. Plots of the mean percantage difference in the mean value of NAA/Cr and Cho/Cralong with the ACE and MoCA Scores at the 4th week of RT, 1-month, 3-month and 6-month post-RT, compared to the previous timepoints and baseline pre-RT values. The star markers show significant differences to the base line, while the plus ones present the significant differences between any parameters to its previous time point.

Subject-averaged ³¹P spectra (black lines) of individual brain tissue types, with corresponding standard deviation across subjects (gray shading). For white matter (A), 15 ROIs were averaged (6 from volunteers, 9 from patients), and 3 ROIs for gray matter (B). For edema (C), 9 ROIs were averaged (from all patients), and 7 ROIs for Gd-contrast enhancement (D, from all high-grade glioma patients).

Example data processing for a GDCE ROI in a high-grade glioma patient. The high-resolution ROI (green) drawn on ¹H images is mapped onto the interpolated ³¹P MRSI grid (matrix = 80×96×64, ³¹P intensity map is shown). Within the resulting low-resolution ROI (blue), phase-/frequency-aligned spectra are summed up to yield the ROI-averaged spectrum. PCr, phosphocreatine; ATP, adenosine-5’-triphosphate; NAD(H), nicotinamide dinucleotide, UDPG, uridine disphosphoglucose; (G)PE, (glycero)phosphoethanolamine; (G)PC, (glycero)phosphocholine; P_i, inorganic phosphate.

Table 2. The performance metrics of the 1D-CNN models on test and validation set.

Figure 2. Example short TE PRESS 1H-MRS data for (a) IDH-mut&TERTp-mut, (b) IDH-mut&TERTp-wt, (c) TERTp-only, and (d) IDH-wt&TERTp-wt gliomas.

Fig 3 - Direct comparison figure - LGG vs HGG

We can see ratio maps of different metabolites to Creatine ratios and the 3T contrast-enhancing clinical images, as well as the 7T T1 images. The LGG patient on top shows a different metabolite pattern in the region of the contrast-enhancing region, compared to the HGG patient below.

Fig 2 - Boxplots tumor ROI vs WM ROI

Most of the metabolites show significant differences in median voxel values of the tumor hotspot region vs. NAWM (see. A). On the B side, we can see a sample hotspot map of tCho with applied cutoff of 0.36 and the 7T T1 magnitude.