Fig 4. Hyperpolarized [1-13C]Pyruvate Magnetic Resonance Spectroscopic Imaging (MRSI) of the mice with NCI-N87 tumors (indicated by red circles) show significantly decreased lactate to pyruvate ratio (L/P) on day 4 of afatinib treatment compared to baseline, while the L/P ratio was not significantly changed in control group.

Fig 5. Treatment of the mice with sensitive NCI-N87 and resistant SNU16 tumors with afatinib showed no significant change in tumors size over 4 days of treatment in either group. Continued treatment in 3 weeks showed significant decrease in the size of NCI-N87 and significant growth of SNU16 tumors. Our in vivo HP-13C MRSI results were confirmed by immunohistochemistry evaluation of the tumor tissues; there is decreased anaerobic glycolysis marker, LDHA, decreased proliferation marker, Ki67, and increased apoptosis marker, Caspase3 in treated tumors compared to control.

Figure 1. Representative hyperpolarized ¹³C data from a patient with pancreatic ductal adenocarcinoma in the tail of the pancreas. The top row shows the metabolite AUC (area under the curve) for pyruvate and downstream metabolites. Normalizing to pyruvate removes signal variation from perfusion and coil shading, providing quantitative information on conversion to lactate and alanine (bottom row). The alanine/lactate ratio provides a relative measure of metabolism and was ~2-fold lower (0.46 vs 0.79) in the tumor (red arrow) than in the normal appearing pancreas (white arrows).

Figure 2. Metabolic ¹³C data from two slices in a patient with pancreatic ductal adenocarcinoma in the body of the pancreas, with extension into the retroperitoneum. Higher lactate/pyruvate and lower alanine/pyruvate was observed in the tumor slice (A, red arrows), while the opposite was seen in the normal appearing pancreas slice (B, white arrows). The alanine/lactate ratio provides a relative measure of metabolism and was 21-fold lower (0.02 vs 0.42) in the tumor than in the normal appearing pancreas.

Fig. 2. A 49-year-old man with a distal common bile duct carcinoma. A. Magnetic resonance cholangiopancreatography shows marked bile ductular dilatation with distal common bile duct stenosis. Primary pancreatic duct dilatation was not noted. B. A T1-weighted MR image shows a hypointense nodule (arrow) in the periampullary region. C. Conventional single-shot echo-planar imaging (c-EPI) shows no abnormalities in the periampullary region. D. Zoomed diffusion-weighted EPI (z-EPI) shows increased lesion signal intensity with better delineation compared with C.

Table 4. Diagnostic accuracy score comparisons between the MRCP and c-EPI DWI combined set as well as between the MRCP and z-EPI DWI combined set

Figure 1: MRI scans of 1 patient at baseline and post chemotherapy. The primary tumour is manually delineated and indicated in red. The diffusion increased with 23%, perfusion fraction and pseudodiffusion decreased with 32% and 58% respectively.

Figure 3: Plot of IVIM parameters at baseline and post chemotherapy. Only D shows a significant increase (p<0.001). The diffusion of the tumour increased after the chemotherapy.

Counter clockwise from upper left: Axial cut of specimen with alginate demonstrating mass (yellow triangle), SMV (blue arrow), CBD (*), PD (green arrow) (red boxes define histology sections A-C); Corresponding T2w axial slice with spiculated irregular mass with areas of hyper-intensity and loss of fat plane between mass and duodenum (**); Corresponding T1w axial slice demonstrating hypo-intense spiculated mass; Histology section A: tumor (red circle), SMV and PD; B: tumor (red circle) with invasion into duodenal muscularis propria (**); C: CBD with tumor invasion (red circle).

Counter clockwise from top left: Radiologic histologic correlation device (RHCD) with MRI visible grid depicting 3-dimensional axis and axial-orientation sectioning grooves (white arrows); human pancreaticoduodenectomy specimen within alginate (blue) in the RHCD; specimen submersion in alginate maintains orientation for imaging and pathologic processing; 3D grid axis as depicted in MR images allows for radiologic-histologic co-localization.

Kaplan–Meier 3-year survival curves for PDACs with high and low kurtosis values. Patients with high kurtosis exhibited lower survival rates than those with low kurtosis (P < 0.001: 1-year survival rate, 75.2% vs. 100%: 3-year survival rate, 14.7% vs. 100%).

Cox Proportional Hazard Analysis of Prognostic Factors for Overall Survival

Figure 1: DWI b600 images of the pancreas using the SENSE reconstruction (a) and the L1-regularized SENSE reconstruction (b) in a healthy pancreas. The central noise artifact seen with SENSE (yellow arrowheads) overlapping the pancreas (black star) is significantly reduced in the L1-regularized SENSE.

Figure 2: (a) The scatter-plot shows the 2D embedding of the neural network’s hidden layer output, to illustrate well separated clustering of control (green), benign (blue), and malignant (red) samples with very little overlap. The clustering performance provides a visual understanding of the high precision accuracy of discrimination obtained in the final output results. (b) Receiver Operating Characteristics (ROC) curves show the sensitivity and specificity performance of the neural-network, with the area under the curve (AUC) for all three classifications above 0.95.

Figure 3: Confusion Matrix result of cancer plasma prediction. The green diagonal boxes show the correct predictions in each class and red boxes indicate misclassifications. The numbers in each box correspond to the number of samples (and their percentage of the total data). The right-most column shows the precision for each predicted class (in green). The bottom-row shows prediction accuracy for each class (in green) and the bottom-right corner box shows the overall accuracy (in green) and error rate (in red). Cancer plasma classification resulted in an 95.2% correct prediction.

Figure 1 A 63 male patient with lymph node metastasis in rectal cancer. T2 image (a), DWI image (b), and APTw-T2w fusion image(c) were showed.A 65 male patient without lymph node metastasis in rectal cancer. T2 image (d), DWI image (e), and APT image(f).

Figure 1 A 63 male patient with lymph node metastasis in rectal cancer. T2 image (a), DWI image (b), and APTw-T2w fusion image(c) were showed.A 65 male patient without lymph node metastasis in rectal cancer. T2 image (d), DWI image (e), and APT image(f).

Title: Axial view of mesorectum using MRI

Legend: Axial view of anterior, posterior, right lateral and left lateral mesorectal fat thickness at the level of 5 cm from the anal verge

Kaplan–Meier curve analysis of prognostic stratification and predictive capability. Prognostic classification of Disease Free Survival using the mesorectal fat thickness (P <.001).

FIGURE 1: A 67-year-old male with high fibrosis PDAC. ①-③ T1WI, T2WI, arterial phase images, Pancreatic tumors show low T1WI signal, slightly higher signal on the T2WI, the arterial phase mild enhancement; ④-⑦ D*, D, f, ADC parameters mappings, 64.5 μm²/ms, 1.16 μm²/ms, 18.03%, 1.2 μm²/ms, respectively. ⑧-⑨ pathological images of Sirius red and CD34 staining, Fibrosis level is 47%, and MVD is 5.6%.

Figure 5. Correlation Between DWI Parameters and Histopathology Features. (A) The correlation between D values and fibrosis. (B) The correlation between f values and fibrosis. (C) The correlation between D* and MVD.

Figure 1. ROC analysis for discriminating malignant tumors from benign lesions

Table 2. Results of ADCmin and SUVmax and the combination of both

Relationship between tumor volume and different MR parameters.

Figure 2: a. Choline metabolite region of representative ¹H MR spectra obtained from the aqueous phase of Panc 1 and Pa04C cells. GPC; glycerophosphocholine, PC; phosphocholine, Cho: choline Cells were collected after culturing in Gln +/- medium for 72h. b. Choline metabolite levels in arbitrary unit (A.U.) obtained from ¹H MR spectra of the aqueous phase of Panc 1 and Pa04C cells. Values represent mean ± SEM, ** P ≤ 0.01, * P ≤ 0.05 (n=3).

Figure 1: Panc 1, Pa04C, Pa09C and Pa20C cell numbers with or without Gln depletion. ~10⁵ cells were seeded in 6 well plates in triplicates. Cells were cultured in Gln +/- medium for 72h. Cell numbers were quantified with an automated cell counter at the end of the time point. Values represent mean ± SEM, ** P ≤ 0.01, * P ≤ 0.05 (n=3).

Table 1. Clinical Characteristics of SPTP and PENT

Table 2. Two observers measure lesions the consistency analysis of various parameters

Comparison of 3T and 14T MR values in a tumor model of pancreatic cancer.

Figure 3. Representative 1H MR spectra obtained from aqueous phase extracts of Pa04C_EV and Pa04C_SLC1A5 pancreatic cancer cells and tumors.

Figure 2. Downregulation of SLC1A5 significantly delayed Pa04C tumor growth.

Figure1. MRI and EUS images of T1-T3 esophageal cancer. T1: A-C, T2WI BLADE(A), Enhanced T1WI radial VIBE(B), and EUS(C). The tumor located in the mucosa and submucosa and protrudes into the lumen without muscularis propria invasion. T2: D-F, T2WI BLADE(D), Enhanced T1WI radial VIBE(E), and EUS(F). The tumor breaks through the submucosa and is limited to the muscularis propria. T3:G-I, T2WI BLADE(G), Enhanced T1WI radial VIBE(H), and EUS(I). The tumor is confined to the adventitia.

Table 1. Comparison between preoperative MRI and postoperative pathological T staging

Comparison of BF from pCASL, pharmacokinetic parameters from DCE-MRI, and ADC from DWI between responders and non-responders

Figure 1 A 77-year-old male patient with the rectal tubular adenocarcinoma. T2 image (a), T2 mapping image (b), and DWI image (c) were showed. A 63-year-old male patient with rectal non-tubular adenocarcinoma. X image (d), T2 mapping image (e), and X image (f) were showed.

Figure2 T2 values of rectal tubular adenocarcinoma (group1) and non-tubular adenocarcinoma (group2) with a significant difference observed (p =0.041).

Fig.1 A61-year-old male patient with rectal cancer. T2W image (A), APTw-T2WI fusion image (B), small solid samples methodimage (C) and single slice ROI methodimage (D) showed the lesion in rectal.

Table 1 The consistency of APTw parametersby two radiologists.

Table 3. Comparison of T2 values in rectal cancer and healthy rectal wall(A); Comparison of T2 values in rectal cancer before and after chemotherapy(B)

Figure 2. The AUC of T2 mapping values to distinguish rectal cancer before and after chemotherapy was 0.774 with sensitivity of 0.588 and specificity of 0.900 and the feasible threshold was 81.738 (A). The AUC of T2 mapping values to distinguish normal rectal wall and rectal cancer was 1 with sensitivity of 0.971 and specificity of 0.929 and the feasible threshold was 66.183 (B).

Figure 2. A 65-year-old male rectal cancer patient without vascular invasion. T2W (1a), T2W with ROI s(1b), DWI (1c), T2 mapping (1d) , FA (1e) , MD (1f), Da (1g), Dr (1h) , MK (1i) map. Three ROIs of rectal cancer were showed on T2W image. Average T2, FA, MD, Da , Dr, MK values of the three ROI were 83.85ms, 0.251, 0.705um2 ms−1, 0.725 um2 ms−1, 0.612 um2 ms−1, 0.419.

Figure 1. A 74-year-old female rectal cancer patient with vascular invasion. T2W (1a), T2W with ROI s(1b), DWI (1c), T2 mapping (1d) , FA (1e) , MD (1f), Da (1g), Dr (1h) , MK (1i) map. Three ROIs of rectal cancer were showed on T2W image. Average T2, FA, MD, Da , Dr , MK values of the three ROI were 80.47ms, 0.302, 1.013 um2 ms−1, 1.313 um2 ms−1, 0.866 um2 ms−1, 2.400.

Figure 3. High b-value DWI and apparent diffusion coefficient (ADC) for noisy (NEX=4), denoised DnCNN (NEX=4), and reference (NEX=16). DnCNN denoises the original NEX=4 images and provides results that are closer to the reference (NEX=16). The arrow is pointing to the tumor location.

Figure 2. Application to the DnCNN to a test case. The inputs are the high b-value noisy image to be denoised and the low b-value image used for guidance. The estimated residual noise is subtracted from the noisy image to generate the denoised image.

Figure 1: Example case showing the segmentation performance from the DeepLab V3 network. The image mosaic shows the T2-weighted image-series [A], the predictions as probability maps [B] (voxel-wise ranging from 0.5 to 1 as indicated by the color bar), and performance maps [C] classified as true negative, false positive, and false negative as specified by the color code.

Table 1: Detection and segmentation performance

Figure 2. The ROC curves for two different methods.

Figure 1. The intra-observer analysis of the segmentation.

A case of multi parameter combined model

Table 2 Performance of models.

Abbreviations: ACC, accuracy; AUC, the mean area under the ROC curve; PPV, positive predictive value; NPV, negative predictive value (NPV); DLC, deep learning-clinical model; RC, radiomics-clinical model.

Figure 3 Flow chart and structure of deep learning-clinical model (DLC).

ROIs are processed as heterogeneity in shapes (HS), heterogeneity in voxel values (HVV) and overall appearance (OA), which represents the characteristics of ROI, shape of ROI, ROI + perROI (5-pixel-circum surrounding ROI) respectively. Six ROIs are analyzed using six ResNet⁹ models and C3D⁸ models respectively. Average of scores are ensemble score, two for ResNet⁹ and another two for C3D⁸. Outputs of the four deep learning models are combined with clinical features to predict the final result using XGBoost¹¹.

Fig. 2 Percentage change of ADC over time for the 4 patient groups. Patients with no or minimal regression (class 1-3) show an strong increase in ADC over the time, while patients with strong regression (class4) only showed a minimal increase followed by a slight decrease of ADC over time.

Fig. 1 T1- (left), T2-weighted (center) and ADC (right) image of patient with advanced rectum carcinoma.

Figure 3. Pre-contrast T1 map, post-contrast T1 map and ∆T1 map of three patients with hilar cholangiocarcinoma (HCCA). (a) Maps of one patient (female, age 62 years) acquired after bilateral drainage; (b) Maps of one patient (female, age 64 years) acquired after bilateral drainage and left portal vein embolization (PVE); (c) Maps of one patient (female, age 67 years) without drainage and PVE.

Figure 2. Pre-contrast (a) and 20 min post-contrast (b) T1-weighted images at different inversion times (TI) reconstructed from GOAL-SNAP sequence of one patient (female, 62 years) with hilar cholangiocarcinoma (HCCA).

Fig 1: Patient, male, 57 years old, ICC in the caudate lobe of liver. A-C: shows the T1W image (hypointense), T2W image (hyperintense), and DWI image (hyperintense) respectively. D-F: shows ADC image of DWI, D, and FA image of DTI respectively. The ADC, D, and FA values were 1.155×10-9m2/s, 1.465×10-9m2/s, and 0.571.

Fig 2: Patient, male, 62 years old, HCC in the right lobe of liver. A-C: shows the T1W image (hypointense), T2W image (hyperintense), and DWI image (hyperintense) respectively. D-F: shows ADC image of DWI, D, and FA image of DTI respectively. The ADC, D and FA values were 1.350×10-9m2/s, 1.655×10-9m2/s, and 0.399.

Figure 2. The MR images of one patient with high grade HCC, positive status of MVI, GPC-3, CK-7 and CK-19.

Figure 5. ROC curves evaluation. The predictive power of logistic regression based models for diagnosing patients with (a) high grade, (b) positive GPC-3, (c) positive CK-19 and (d) positive CK-7, respectively.

Figure 2. A 62 year-old male with poorly-differentiated HCC in the right lobe of the liver. (a) T2WI image; (b) phase map; (c) tumor was delineated around the edge of the tumor; (d) AS software recognized quantitatively and automatically ITSS ratio by reading phase maps. ITSS recognized were covered in green.

Table 2. Summary of mean ITSS values between well-moderately and poorly-differentiated HCC groups

Table 1. Comparison of IVIM-DWI quantitative parameters between CK19-positive and CK19-negative HCC groups

Figure 3. ROC curve for the discrimination of CK19-positive and CK19-negative HCCs.

Figure2:A 66 year-old male withICC in the right lobe of the liver. (a)DKI image; (b) phase map; (c) tumor was delineated around the edge of the tumor; (d) AS software recognized quantitatively and automatically ITSS ratio by reading phase maps. ITSS recognized were covered in green.

Figure1 workflow. (a) Input phase map that needs to be de-artifacted. (b) The median filtering result of A. (c) A and B are subtracted and threshold processed to obtain high pixel value and low pixel value bands. (d) The adjacent regions of high pixel value and low pixel value are extracted and expanded in the acquisition plane. This result is considered as the artifact region. (e) The result of recomputing the pixel values in the artifact region. The method is to calculate the average value of 26 adjacent non-artifact region pixels of each target pixel. (f) The result of removing the artifact.

Box plot showing the DKI derived metrics of peritumoral zones from progressing and pseudo-progressing groups

DKI derived metrics in true progressing and pseudo-progressing lesions

Table 2. Predictive performance of IVIM and ESWAN parameters between MVI-positive and MVI-negative groups

Table 3. Difference between AUCs of IVIM, ESWAN and Combine.P1 the difference between IVIM and ESWAN, P2 the difference between IVIM and Combine, P3 the difference between ESWAN and Combine.

Table 1. Detailed MR scanning parameters

Figure 3. ROC curve for discrimination of DPHCC and non-DPHCC.

value of R2*

FAT FRACTION

Figure 1. a) 2H MR spectra of subcutaneous tumor after injection of 2H9 choline. b)Time curves of HOD and 2H9 choline signal integrals c) T2 weighted MRI of subcutaneous tumor. d) 2H9 choline heat map overlaid on T2 weighted MRI

Figure 2 a) DMI of phantom with two tubes, one filled with 2H9 choline and the other with [6,6 2H2]glucose. The glucose and choline signals are clearly separated in the 2H spectrum b) Left: 2H spectrum of renal tumor after IV bolus infusion of 2H9 choline and [6,6 2H2]glucose combined in the tail vein of a mouse. Right DMI maps of deuterated glucose, choline and lactate overlaid on T2 MRI

Figure 1. Surgically confirmed moderately differentiated HCC in a 59-year-old man with microvascular invasion found at pathology. HCC with microvascular invasion (hematoxylin-eosin stain, Original magnification 200×) (A). R2* map shows a hepatic mass(59mm×61mm) in S6 with elevated R2* around the tumor (B).

Figure 2. Surgically confirmed poorly differentiated HCC in a 55-year-old man with microvascular invasion found at pathology. R2* of CSE-MRI shows a hepatic mass(75mm×81mm) in S7/8 with elevated R2* around the tumor (A). The tumors in S7/8 show high signal intensity on FS -T2WI (B).The tumors in S7/8 show APHE and peritumoral corona enhancement and on early arterial phase and washout on the portal vein phase(C and D). HCC with microvascular invasion (hematoxylin-eosin stain, Original magnification 200×) (E).

Figure 1. 2D region of interest (2D-ROI) and volume of interest (VOI). T2-weighted imaging (T2WI) (A) and arterial phase contrast-enhanced T1-weighted imaging (T1WI) (B), and R2* map (C) show liver metastasis (yellow line) confirmed by histology in a 59-year-old woman with lung cancer. T2WI (E) and arterial phase contrast-enhanced T1WI (F), and R2* map (G) show a live hemangioma (yellow line) in in a 59-year-old woman. (D) 2D-ROI was drawn on the section showing the maximal tumor dimension. (H) VOI was placed covering the entire tumor volume on R2* map.

Figure 4. Receiver operating characteristic curve analysis of the two positioning methods in differentiating between malignant group and benign group. 2D region of interest (2D-ROI) and volume of interest (VOI) methods yielded the similar results.

The ROC curve for SVM classifier

Figure 2. Typical examples of ADC value measurement in the right femur.

Figure 1. Spearman linear correlation test in 68 patients who underwent magnetic resonance imaging examination after treatment. There was a negative correlation between the apparent diffusion coefficient (ADC) value and the degree of bone marrow infiltration in the right ilium (r=-0.829, P<0.001).

Comparison of baseline MR parameters between the two response groups

Spearman correlation analysis between response depth and MRI parameters

Figure 2: Coronal view of mouse tibia displaying the VOI drawn to segment the bone marrow at the baseline. a) PDFF map, b) ADC map, c) MTR map, and d) R2* map.

Figure 4: Slice by slice profile of quantitative MRI metrics along the tibia VOI for a test and retest time point.

Figure.1: a-d）A 35-year-old male with adenocarcinoma of the upper left lobe. In these images, a is PET image，b is D pseudo colored map，c is f pseudo colored map, d is D* pseudo colored map.

Figure.2: The SUVmax, D, D*, and f values of the squamous carcinoma and adenocarcinoma group. (a) D = (1.29±0.38)×10^-3 mm²/s, (0.95±0.20)×10^-3 mm²/s ; (b) f = (29.91±16.64)%, (18.67±12.15)% ; (c) D* = (64.12±67.91)×10^-3 mm²/s, (21.28±22.19)×10^-3 mm²/s ; (d) SUVmax = (6.55±4.26), (12.56±5.85).

Figure 2. A) T2- and Diffusion-weighted images in the representative TAL + IRN + TMZ treated and untreated mice. Anatomical images indicate the increase in tumor size of the untreated mouse while the tumor is barely visible in the treatment group. B) Measured tumor volume from the multi-slice T2-weighted MRI, and C) measured tumor ROI ADC values in both groups.

Figure 3. Longitudinal ¹H MR spectra from a representative TAL+IRN+TMZ treated mouse and an untreated mouse showing distinct macromolecule peak patterns in the two groups.