Deep learning DCE-MRI parameter estimation: Application in pancreatic cancer.
Medical image analysis
Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) is an MRI technique for quantifying perfusion that can be used in clinical applications for classification of tumours and other types of diseases. Conventionally, the non-linear least squares (NLLS) methods is used for tracer-kinetic modelling of DCE data. However, despite promising results, NLLS suffers from long processing times (minutes-hours) and noisy parameter maps due to the non-convexity of the cost function. In this work, we investigated physics-informed deep neural networks for estimating physiological parameters from DCE-MRI signal-curves. Three voxel-wise temporal frameworks (FCN, LSTM, GRU) and two spatio-temporal frameworks (CNN, U-Net) were investigated. The accuracy and precision of parameter estimation by the temporal frameworks were evaluated in simulations. All networks showed higher precision than the NLLS. Specifically, the GRU showed to decrease the random error on v by a factor of 4.8 with respect to the NLLS for noise (SD) of 1/20. The accuracy was better for the prediction of the v parameter in all networks compared to the NLLS. The GRU and LSTM worked with arbitrary acquisition lengths. The GRU was selected for in vivo evaluation and compared to the spatio-temporal frameworks in 28 patients with pancreatic cancer. All neural network approaches showed less noisy parameter maps than the NLLS. The GRU had better test-retest repeatability than the NLLS for all three parameters and was able to detect one additional patient with significant changes in DCE parameters post chemo-radiotherapy. Although the U-Net and CNN had even better test-retest characteristics than the GRU, and were able to detect even more responders, they also showed potential systematic errors in the parameter maps. Therefore, we advise using our GRU framework for analysing DCE data.
10.1016/j.media.2022.102512
Screening for pancreatic cancer in high-risk individuals using MRI: optimization of scan techniques to detect small lesions.
Familial cancer
Pancreatic cancer has a dismal prognosis in the general population. However, early detection and treatment of disease in high-risk individuals can improve survival, as patients with localized disease and especially patients with lesions smaller than 10 mm show greatly improved 5-year survival rates. To achieve early detection through MRI surveillance programs, optimization of imaging is required. Advances in MRI technologies in both hardware and software over the years have enabled reliable detection of pancreatic cancer at a small size and early stage. Standardization of dedicated imaging protocols for the pancreas are still lacking. In this review we discuss state of the art scan techniques, sequences, reduction of artifacts and imaging strategies that enable early detection of lesions. Furthermore, we present the imaging features of small pancreatic cancers from a large cohort of high-risk individuals. Refinement of MRI techniques, increased scan quality and the use of artificial intelligence may further improve early detection and the prognosis of pancreatic cancer in a screening setting.
10.1007/s10689-024-00394-z
Accelerated Pancreatobiliary MRI for Pancreatic Cancer Surveillance in Patients With Pancreatic Cystic Neoplasms.
Journal of magnetic resonance imaging : JMRI
BACKGROUND:Pancreatobiliary MRI is often recommended for patients at risk of developing pancreas cancer. But the surveillance MRI protocol has not yet been widely accepted. PURPOSE:To establish an accelerated MRI protocol targeting the table time of 15 minutes for pancreatic cancer surveillance and test its performance in lesion characterization. STUDY TYPE:Prospective. POPULATION:A total of 30 participants were enrolled, who were undergoing follow-up care for intraductal papillary mucinous neoplasms or newly diagnosed pancreatic cysts (≥10 mm) and were scheduled for or had recently undergone contrast-enhanced CT (CECT). FIELD STRENGTH/SEQUENCE:A 3 T; heavily T2WI, 3D MRCP, DWI, dynamic T1WI, two-point Dixon. ASSESSMENT:In-room time and table time were measured. Seven radiologists independently reviewed image quality of MRI and then the presence of high-risk stigmata and worrisome features in addition to diagnostic confidence for accelerated MRI, CECT, and the noncontrast part of accelerated MRI (NC-MRI). STATISTICAL ANALYSIS:Fisher's exact test was used for categorical variables and either the Student's t-test or Mann-Whitney test was performed for continuous variables. The generalized estimated equation was used to compare the diagnostic performance of examinations on a per-patient basis. Interobserver agreement was evaluated via Fleiss kappa. A P value of <0.05 was considered to be statistically significant. RESULTS:The in-room time was 18.5 ± 2.6 minutes (range: 13.7-24.9) and the table time was 13.9 ± 1.9 minutes (range: 10.7-17.5). There was no significant difference between the diagnostic performances of the three examinations (pooled sensitivity: 75% for accelerated MRI and CECT, 68% for NC-MRI, P = 0.95), with the highest significant diagnostic confidence for accelerated MRI (4.2 ± 0.1). With accelerated MRI, the interobserver agreement was fair to excellent for high-risk stigmata (κ = 0.34-0.98). DATA CONCLUSION:Accelerated MRI protocol affords a table time of 15 minutes, making it potentially suitable for cancer surveillance in patients at risk of developing pancreatic cancer. EVIDENCE LEVEL:2 TECHNICAL EFFICACY STAGE: 2.
10.1002/jmri.28189
Imaging diagnosis of pancreatic cancer: a state-of-the-art review.
Lee Eun Sun,Lee Jeong Min
World journal of gastroenterology
Pancreatic cancer (PC) remains one of the deadliest cancers worldwide, and has a poor, five-year survival rate of 5%. Although complete surgical resection is the only curative therapy for pancreatic cancer, less than 20% of newly-diagnosed patients undergo surgical resection with a curative intent. Due to the lack of early symptoms and the tendency of pancreatic adenocarcinoma to invade adjacent structures or to metastasize at an early stage, many patients with pancreatic cancer already have advanced disease at the time of their diagnosis and, therefore, there is a high mortality rate. To improve the patient survival rate, early detection of PC is critical. The diagnosis of PC relies on computed tomography (CT) and/or magnetic resonance imaging (MRI) with magnetic resonance cholangiopancreatography (MRCP), or biopsy or fine-needle aspiration using endoscopic ultrasound (EUS). Although multi-detector row computed tomography currently has a major role in the evaluation of PC, MRI with MRCP facilitates better detection of tumors at an early stage by allowing a comprehensive analysis of the morphological changes of the pancreas parenchyma and pancreatic duct. The diagnosis could be improved using positron emission tomography techniques in special conditions in which CT and EUS are not completely diagnostic. It is essential for clinicians to understand the advantages and disadvantages of the various pancreatic imaging modalities in order to be able to make optimal treatment and management decisions. Our study investigates the current role and innovative techniques of pancreatic imaging focused on the detection of pancreatic cancer.
10.3748/wjg.v20.i24.7864
Solid bone tumors of the spine: Diagnostic performance of apparent diffusion coefficient measured using diffusion-weighted MRI using histology as a reference standard.
Pozzi Grazia,Albano Domenico,Messina Carmelo,Angileri Salvatore Alessio,Al-Mnayyis Asma'a,Galbusera Fabio,Luzzati Alessandro,Perrucchini Giuseppe,Scotto Gennaro,Parafioriti Antonina,Zerbi Alberto,Sconfienza Luca Maria
Journal of magnetic resonance imaging : JMRI
PURPOSE:To assess the diagnostic performance of mean apparent diffusion coefficient (mADC) in differentiating benign from malignant bone spine tumors, using histology as a reference standard. Conventional magnetic resonance imaging (MRI) sequences have good reliability in evaluating spinal bone tumors, although some features of benign and malignant cancers may overlap, making the differential diagnosis challenging. MATERIALS AND METHODS:In all, 116 patients (62 males, 54 females; mean age 59.5 ± 14.1) with biopsy-proven spinal bone tumors were studied. Field strength/sequences: 1.5T MR system; T -weighted turbo spin-echo (repetition time / echo time [TR/TE], 500/13 msec; number of excitations [NEX], 2; slice thickness, 4 mm), T -weighted turbo spin-echo (TR/TE, 4100/102 msec; NEX, 2; slice thickness, 4 mm), short tau inversion recovery (TR/TE, 4800/89 msec; NEX, 2; slice thickness, 4 mm, IT, 140 msec), axial spin-echo echo-planar diffusion-weighted imaging (DWI) (TR/TE 5200/72 msec; slice thickness 5 mm; field of view, 300; interslice gap, 1.5 mm; NEX, 6; echo-planar imaging factor, 96; no parallel imaging) with b-values of 0 and 1000 s/mm², and 3D fat-suppressed T -weighted gradient-recalled-echo (TR/TE, 500/13 msec; slice thickness, 4 mm) after administration of 0.2 ml/kg body weight gadolinum-diethylenetriamine pentaacetic acid. Two readers manually drew regions of interest on the solid portion of the lesion (hyperintense on T -weighted images, hypointense on T -weighted images, and enhanced after gadolinium administration on fat-suppressed T -weighted images) to calculate mADC. Histology was used as the reference standard. Tumors were classified into malignant primary tumors (MPT), bone metastases (BM), or benign primary tumors (BPT). Statistical tests: Nonnormality of distribution was tested with the Shapiro-Wilk test. The Kruskal-Wallis and Mann-Whitney U-test with Bonferroni correction were used. Sensitivity and specificity of the mADC values for BM, MPT, and BPT were calculated. Approximate receiver operating characteristic curves were created. Interobserver reproducibility was evaluated using the intraclass correlation coefficient (ICC). RESULTS:The mADC values of MPT (n = 35), BM (n = 65), and BPT (n = 16) were 1.00 ± 0.32 (0.59-2.10) × 10 mm /s, 1.02 ± 0.25 (0.73-1.96) × 10 mm /s, 1.31 ± 0.36 (0.83-2.14) × 10 mm /s, respectively. The mADC was significantly different between BPT and all malignant lesions (BM+MPT) (P < 0.001), BM and BPT (P = 0.008), and MPT and BPT (P = 0.008). No difference was found between BM and MPT (P = 0.999). An mADC threshold of 0.952 × 10 mm /s yielded 81.3% sensitivity, 55.0% specificity. Accuracy was 76% (95% confidence interval [CI] = 63.9%-88.1%). Interobserver reproducibility was almost perfect (ICC = 0.916; 95% CI = 0.879-0.942). CONCLUSION:DWI with mADC quantification is a reproducible tool to differentiate benign from malignant solid tumors with 76% accuracy. The mADC values of BPT were statistically higher than that of malignant tumors. However, the large overlap between cases may make mADC not helpful in a specific patient. LEVEL OF EVIDENCE:3 Technical Efficacy: Stage 2 J. Magn. Reson. Imaging 2018;47:1034-1042.
10.1002/jmri.25826
Standard-b-value vs low-b-value DWI for differentiation of benign and malignant vertebral fractures: a meta-analysis.
Luo Zhanpeng,Litao Li,Gu Suxi,Luo Xiaobo,Li Dawei,Yu Long,Ma Yuanzheng
The British journal of radiology
OBJECTIVE:To determine the comparative diagnostic performance of standard-b-value (≥500 mm(2)) vs low-b-value (<500s mm(-2)) diffusion-weighted imaging (DWI) for discriminating malignant from benign vertebral compression fractures. METHODS:12 studies with a total of 350 malignant and 312 benign vertebral fractures were included. RESULTS:The apparent diffusion coefficient (ADC) value of benign vertebral compression fractures was lower than that of malignant vertebral compression fractures (SMD = 1.81, 95% CI 0.98 to 2.64 Z = 4.27, p < 0.05). ADC value difference was more pronounced in the group of low-b-value DWI (SMD = 2.31, 95% CI 1.02 to 3.60 Z = 3.51, p < 0.05) than in the group of standard-b-value DWI (SMD = 1.38, 95% CI 0.18 to 2.59 Z = 2.25, p < 0.05). Ethnicity stratified analysis demonstrated higher ADC values in benign vertebral compression fractures in comparison to malignant tissues in both the Asian and Caucasian subgroups (Asians: SMD = 2.400, 95%CI 1.45 to approximately 3.35, p<0.05; Caucasians: SMD = 0.592, 95 % CI -0.848 to approximately 2.032, p < 0.05). And the ADC value difference was more pronounced in the Asian subgroup. CONCLUSION:ADC value appears to be a reliable method to differentiate benign from malignant fractures. Low-b-value DWI was more a valuable parameter than standard-b-value DWI for discriminating malignant from benign vertebral compression fractures. And the diffusion characteristics of the benign vertebral fractures such as osteoporosis, trauma and infection have rarely been investigated separately. ADVANCES IN KNOWLEDGE:The use of low-b-value DWI for differentiation of benign and malignant vertebral fractures is recommended.
10.1259/bjr.20150384