Multi-contrast magnetic resonance image reconstruction

Magnetic resonance imaging (MRI) has an important feature that it provides multiple images with different contrasts for complementary diagnostic information. However, a large amount of data is needed for multi-contrast images depiction, and thus, the scan is time-consuming. Many methods based on parallel magnetic resonance imaging (pMRI) and compressed sensing (CS) are applied to accelerate multi-contrast MR imaging. Nevertheless, the image reconstructed by sophisticated pMRI methods contains residual aliasing artifact that degrades the quality of the image when the acceleration factor is high. Other methods based on CS always suffer the regularization parameter-selecting problem. To address these issues, a new method is presented for joint multi-contrast image reconstruction and coil sensitivity estimation. The coil sensitivities can be shared during the reconstruction due to the identity of coil sensitivity profiles of different contrast images for imaging stationary tissues. The proposed method uses the coil sensitivities as sharable information during the reconstruction to improve the reconstruction quality. As a result, the residual aliasing artifact can be effectively removed in the reconstructed multi-contrast images even if the acceleration factor is high. Besides, as there is no regularization term in the proposed method, the troublesome regularization parameter selection in the CS can also be avoided. Results from multi-contrast in vivo experiments demonstrated that multi-contrast images can be jointly reconstructed by the proposed method with effective removal of the residual aliasing artifact at a high acceleration factor.

Magnetic resonance imaging (MRI) tasks often involve multiple contrasts. Recently, numerous deep learning-based multi-contrast MRI super-resolution (SR) and reconstruction methods have been proposed to explore the complementary information from the multi-contrast images. However, these methods either construct parameter-sharing networks or manually design fusion rules, failing to accurately model the correlations between multi-contrast images and lacking certain interpretations. In this paper, we propose a multi-contrast convolutional dictionary (MC-CDic) model under the guidance of the optimization algorithm with a well-designed data fidelity term. Specifically, we bulid an observation model for the multi-contrast MR images to explicitly model the multi-contrast images as common features and unique features. In this way, only the useful information in the reference image can be transferred to the target image, while the inconsistent information will be ignored. We employ the proximal gradient algorithm to optimize the model and unroll the iterative steps into a deep CDic model. Especially, the proximal operators are replaced by learnable ResNet. In addition, multi-scale dictionaries are introduced to further improve the model performance. We test our MC-CDic model on multi-contrast MRI SR and reconstruction tasks. Experimental results demonstrate the superior performance of the proposed MC-CDic model against existing SOTA methods. Code is available at https://github.com/lpcccc-cv/MC-CDic.

The ability to manipulate image contrast and thus to obtain complementary information is one of the main advantages of MRI. Motion consistency within the whole data set is a key point in the context of multi contrast imaging. In cardiac and abdominal MRI, the acquisition strategy uses multiple breath-holds and often relies on acceleration methods that inherently suffer from a signal to-noise ratio loss. The aim of this work is to propose a free-breathing multi-contrast acquisition and reconstruction workflow to improve image quality and the subsequent data analysis. We extended a previously proposed motion-compensated image reconstruction method for multi-contrast imaging. Shared information throughout the imaging protocol is now exploited by the image reconstruction in the form of an additional constraint based on image gradient sparsity. This constraint helps to minimize the amount of data needed for efficient non-rigid motion correction. T₁and T₂weighted images were reconstructed from free-breathing acquisitions in 4 healthy volunteers and in a phantom. The impact of multi-contrast motion correction was evaluated in a phantom in terms of precision and accuracy of T₁and T₂quantification. In the phantom, the proposed method achieved an accuracy of 97.5 % on the quantified parameters against 88.0 % before motion correction. In volunteers, motion inconsistency in T₁and T₂quantification were noticeably reduced within 5 min of free-breathing acquisition. An efficient, free-breathing, multi-contrast imaging method has been demonstrated that does not require prior assumptions about contrast and that is applicable to a wide range of examinations.

The authors investigated the magnetic resonance appearance of hepatocellular carcinoma using a 1.5-Tesla magnet. Twenty-four patients with pathologically proven hepatocellular carcinoma had magnetic resonance imaging (MRI) studies, which were retrospectively reviewed. All patients were imaged with at least two of the following techniques: (1) T1-weighted (T1W), (2) T1-weighted with Gd-DTPA enhancement (T1W-E), (3) T2-weighted (T2W), (4) proton density (PD), and (5) gradient-recalled echoes (GRE). T1W images were equal to T2W images for tumor detection using a grading system. T1W images were slightly better than T2W images for the total number of lesions detected. The other pulsing techniques (PD, T1W-E, and GRE) detected fewer lesions. Eight cases of hepatocellular carcinoma (33%) had nonhomogeneous increased signal intensity on both T1W and T2W images. The authors conclude that T1W images are equal to T2W images for detection of hepatocellular carcinoma. The authors also conclude that 33% of hepatocellular carcinomas have an imaging pattern with increased signal intensity on both T1W and T2W images. This pattern is atypical for most other hepatic masses and hence can be used to suggest the mass is hepatocellular carcinoma.

A6-year-old male castrated Greyhound was presented to the William R. Pritchard Veterinary Medical Teaching Hospital, University of California, Davis, with a 4-week history of apparent spinal pain and paraparesis. There was no improvement despite treatment consisting of oral carprofen (unknown dose) for 10 days followed by a tapering course of oral prednisone for 6 days before referral (1.2–0.2 mg/kg/d). Abnormalities on physical examination included 5% dehydration, a large flaccid bladder that was easily expressed, moderate conjunctivitis, and mild generalized patchy alopecia. On neurological examination, the dog was ambulatory with marked paraparesis and absent conscious proprioception in both pelvic limbs. The patellar, gastrocnemius, and pelvic limb withdrawal reflexes were decreased bilaterally, the cutaneous trunci reflex was absent caudal to L4, and the perineal reflex and anal tone were absent. The tail was flaccid with intact superficial pain sensation. The dog was painful with dorsal palpation of the lumbar spine. The neurological abnormalities were consistent with an L4–Cd myelopathy. A CBC and serum biochemistry had mild abnormalities, interpreted as clinically insignificant. Urinalysis indicated a specific gravity of 1.048 and marked proteinuria (4+ on sulfosalicylic acid method). Thoracic radiographs and abdominal ultrasound examination indicated no clinically relevant findings. After vertebral column radiographs, where no abnormalities were seen, two 22 G 3.5-in. spinal needles were placed, 1 at the L4–5 and 1 at the L5–6 intervertebral disc spaces, but cerebrospinal fluid (CSF) could not be obtained. CSF was collected from the cerebellomedullary cistern. A diagnostic myelogram was not performed and, therefore, magnetic resonance (MR) images of the lumbar spinal cord followed by those of the brain were acquired with a 1.5-T scanner.a T1-weighted (T1W) images and T2-weighted (T2W) images included the L3 vertebra to the caudal vertebrae and the entire brain to the C3 vertebra. The spinal cord was enlarged and slightly elevated and dorsoventrally flattened from the L4 through the S2 vertebrae. On the T2W images, there was intramedullary hyperintensity of the lumbar and sacral segments of the spinal cord and the cauda equina, with partial loss of epidural fat signal and gray and white matter definition in the mid lumber region (Fig 1A). After IV contrast administration,b there was moderate parenchymal contrast enhancement and mild diffuse meningeal enhancement from the cranial lumbar region to the conus medullaris (Fig 1B and C). On the postcontrast T1W images of the brain, there was diffuse meningeal enhancement, most evident along the tentorium cerebelli and the falx cerebri (not shown). (A) T2-weighted sagittal magnetic resonance images (MRI) from dog 1. Note the intramedullary hyperintensity compared with spinal cord gray matter, loss of epidural fat signal, and gray/white matter definition. (B) T1-weighted sagittal MRI and (C) T1-weighted sagittal MRI postcontrast from dog 1. There is moderate parenchymal contrast enhancement and mild diffuse meningeal enhancement (arrow) from L4 to the conus medullaris. The CSF had a markedly increased nucleated cell count (660 cells/μL; reference range, ≤ 3 cells/μL) and a moderate increase in the protein concentration (72 mg/dL; reference range, ≤ 25 mg/dL). The differential cell count was 15% neutrophils, 12% lymphocytes, 68% large mononuclear cells, and 5% eosinophils. On cytologic evaluation, most of the large mononuclear cells were considered typical macrophages; however, some of the cells were extremely large and variably vacuolated with large immature nuclei (Fig 2). Within this small subpopulation of large immature cells, numerous mitoses were noted (2, 3). There were occasional large multinucleate tumor cells, with a marked anisokaryosis of the component nuclei and nuclear fragmentation (Fig 3). Rare bizarre multinucleate tumor cell variants also were present (Fig 4). The interpretation was marked, mixed, predominantly histiocytic inflammation with underling neoplasia, most likely histiocytic sarcoma. Cerebrospinal fluid cytospin slide, dog 1. There is a marked, predominantly large mononuclear pleocytosis. Note the scattered, extremely large, single round cells with large immature nuclei. Within this small subpopulation of large immature cells, there are numerous mitoses. Wright–Giemsa stain, scale bar = 50 μm. Cerebrospinal fluid cytospin slide, dog 1. Occasional large multinucleate tumor cells are present that have marked anisokaryosis of component nuclei and nuclear fragmentation. Wright–Giemsa stain, scale bar = 30 μm. Cerebrospinal fluid cytospin slide, dog 1. Rare bizarre multinucleate tumor cell variants are present that also have marked anisokaryosis of component nuclei and nuclear fragmentation. Wright–Giemsa stain, scale bar = 30 μm. Treatment with an anti-inflammatory dosage of dexamethasone sodium phosphate IV (0.22 mg/kg/d) was initiated immediately, but the dog deteriorated neurologically over the next 36 hours after anesthesia and became nonambulatory with marked apparent spinal pain. Immunophenotyping of the cells in the CSF was carried out by staining CSF cytospin slides using canine-specific monoclonal antibodies against CD1c, CD11b, CD11c, CD11d, CD3, CD79a, CD34, pancytokeratin, vimentin, and a negative control using previously described methods.1 The large immature cells strongly expressed CD1c (Fig 5) and CD11c. The neutrophils expressed CD11c. As expected, all leukocytes expressed vimentin. The diagnosis was histiocytic sarcoma with secondary marked, mixed inflammation. Based on the deteriorating neurologic status and the poor prognosis associated with histiocytic sarcoma, the dog was euthanized and submitted for necropsy. Cerebrospinal fluid cytospin slide, dog 1. The large immature single round tumor cells have a variable, predominantly cytoplasmic membrane, expression of CD1c. Immunocytochemistry stain, AEC substrate, hematoxylin counterstain, scale bar = 30 μm. At necropsy, clinically relevant findings on gross and histological examination were restricted to the central nervous system (CNS). The meninges of the brain and the spinal cord were diffusely opaque. On the cut surface of the spinal cord, the subarachnoid space was diffusely filled by an infiltrative, pale tan mass that entrapped the nerve rootlets from L1 to the cauda equina and invaded the spinal cord from L4 to the cauda equina. Tissues were immersed in 10% neutral buffered formalin and routinely processed. Five micrometer paraffin-embedded sections were stained with hematoxylin and eosin (H&E) for routine histologic assessment. Immunohistochemistry was performed on sections of the spinal cord using canine leukocyte-specific antibodies for which the sources, reactivity, and staining protocol have been documented previously.2, 3 Histologically, dense sheets of pleomorphic, neoplastic round cells were present within the subarachnoid space of the spinal cord and brain, often obscuring the leptomeninges and multifocally invading the dura mater and subpial neuropil. The most extensive invasion involved the L4–6 and caudal spinal cord segments, in which neoplastic cells extended to and within the central canal. The neoplasm tapered from the thoracic to the cervical spinal cord, markedly expanded again around the medulla oblongata, cerebellum, and thalamus, and then tapered toward the cerebral hemispheres. It invaded along the subarachnoid space around the blood vessels within the superficial neuropil, outlined the cerebellar folia, and occasionally invaded the subpial neuropil. The affected nerve rootlets and the neuropil of the spinal cord and brain contained extensive edema and Wallerian degeneration. Consistent with the histiocytic origin demonstrated by immunocytochemistry of the CSF, neoplastic cells were broadly immunoreactive for CD18, inconsistently immunoreactive for CD45, and not immunoreactive for all other leukocyte antibodies (CD3, CD11d, CD79a, and pancytokeratin). A 9-year-old male castrated Pembroke Welsh Corgi was presented to the William R. Pritchard Veterinary Medical Teaching Hospital, University of California, Davis, for an acute onset of obtundation, head tilt, and difficulty in walking. The dog had received nonsteroidal anti-inflammatory treatment (unknown drug and dose) before presentation. Neurologic examination indicated moderate obtundation, tetraparesis, and mild generalized ataxia with a tendency to walk against the wall on the left side. Inconsistent menace response in the right eye, decreased response to nasal stimulation in the left naris, a mild head tilt to the right, and miotic pupils were noted. Proprioceptive positioning was evaluated by paw replacement reaction, and was absent in the right thoracic and pelvic limbs, and intact on the left. Neuroanatomic localization was consistent with multifocal brain disease. A CBC, serum biochemistry, urinalysis, thoracic radiographs, and abdominal ultrasound examination were unremarkable except for a mildly enlarged and heterogeneously echogenic right pancreatic limb interpreted as most consistent with previous pancreatitis. The dog was anesthetized and MR imaging of the brain was performed, including pre- and postcontrast T1W, T2W, proton density, and fluid-attenuated inversion recovery (FLAIR) images. A mass was identified, which appeared to arise from the lateral margin of the hippocampus, with an extensive component within the lateral ventricles bilaterally. The mass was a hypointense to gray matter on T1W images (Fig 6A) and isointense to slightly hyperintense on T2W images. On T2W images and FLAIR sequences, bilateral periventricular and peritumoral edema was observed (Fig 6B). The mass enhanced homogeneously with contrast administration (Fig 6C and D). (A) T1-weighted transverse magnetic resonance imaging (MRI) at the level of the zygomatic process of the temporal bone of dog 2. Note the bilateral hypointense mass associated with the lateral margin of the hippocampus. (B) Transverse fluid-attenuated inversion recovery image. Note the bilateral periventricular and peritumoral edema. Transverse (C) and dorsal plane (D) T1-weighted postcontrast images. Note the uniform, bilateral contrast enhancement of the mass with asymmetrical bilateral extension into the ventricular lamina. Cisternal CSF had a protein concentration of 135mg/dL (reference range, ≤ 25 mg/dL), 220 RBC/μL (reference range, 0 RBC/μL), and a markedly increased nucleated cell count of 1,550/μL (reference range, ≤ 3 cells/μL), with a differential cell count of 1% lymphocytes, 19% neutrophils, and 80% macrophages. On cytological evaluation, macrophages predominated, with nondegenerate neutrophils in lower numbers. A population of individual giant cells with rounded cytoplasmic margins was also present. These cells had a moderate to high N : C ratio and abundant, homogeneous, medium blue cytoplasm. Rare cells were erythrophagic. Nuclei were oval to reniform with lacy, open chromatin and were somewhat eccentrically placed within the cytoplasm. Binucleated and multinucleated cells were seen occasionally and featured marked anisocytosis and anisokaryosis. Numerous mitotic figures were noted within this population of large, immature round cells. The diagnosis was malignant neoplasia with associated marked mixed inflammation and mild prior hemorrhage. Additional CSF cytospin slides were made and assessed with antibodies against CD1c, CD11b, CD11c, cytokeratin (pancytokeratin and AE1/AE3), and GFAP along with a negative control. The neoplastic cells were not immunoreactive for cytokeratin or GFAP, moderately immunoreactive for CD1c, weakly to moderately immunoreactive for CD11c, and variably and weakly immunoreactive for CD11b. The diagnosis was histiocytic sarcoma with associated marked mixed inflammation. The dog recovered from anesthesia and was treated with an anti-inflammatory dosage of dexamethasone sodium phosphate (0.2 mg/kg/d) IV and lactated ringers solution with 20 mEq KCl/L was added IV (6.7 mL/kg/h). Twelve hours later, the dog was markedly obtunded. Because of rapid deterioration in neurological status and the poor prognosis associated with histiocytic sarcoma, the dog was euthanized. Clinically relevant gross necropsy findings were limited to the CNS. After fixation in 10% neutral buffered formalin, the brain was sectioned frontally and a soft tan mass that was bilaterally within the lateral ventricles and third ventricle was observed. It extensively invaded and effaced the dentate gyri and corpus callosum. Tissues were fixed in formalin, embedded in paraffin, sectioned at 5 μm, and stained with H&E. A section of the brain that included the dentate gyrus was immunohistochemically stained using a panel of leukocyte markers, as previously described in this report. On histopathologic examination, densely cellular sheets of neoplastic round cells, supported by moderate fibrovascular connective tissue, extensively effaced the dentate gyri and the corpus callosum. Few to numerous neoplastic round cells were scattered throughout the subarachnoid space and there was multifocal invasion of the subpial neuropil. The neoplastic cells were strongly immunoreactive with antibodies directed against CD45 and CD18. A subset of the neoplastic cell population, most notably within the meninges, was CD11d immunopositive. The neoplastic cell population exhibited no immunoreactivity with antibodies directed against CD3, CD79a, or CD20. Based on the cellular morphology and results of the immunohistochemical staining of the paraffin-embedded brain tissue, the mass was diagnosed as a histiocytic sarcoma. These results confirmed the diagnosis made antemortem by cytological and immunocytochemical evaluation of the CSF. The neoplasm was considered a primary CNS histiocytic sarcoma because there was no gross or histological evidence of neoplasia outside of the CNS. Histiocytic disorders are common in dogs and can be divided into reactive histiocytoses, cutaneous histiocytoma, localized histiocytic sarcoma, and disseminated histiocytic sarcoma (dHS)/malignant histiocytosis (mHS). dHS and mHS are histiocytic cell malignancies that either arise in 1 location with secondary widespread dissemination (dHS) or in many locations simultaneously (mHS).2, 3 These 2 diseases may be difficult to differentiate and, therefore, the terms frequently are used interchangeably. Dogs with dHS/mHS often have clinical signs of systemic disease because the internal organs are affected.3 Predilection sites include spleen, liver, lymph nodes, lungs, and bone marrow.4, 5 Bernese Mountain Dogs, Rottweilers, and Golden, Labrador, and Flatcoat Retrievers are predisposed to this disease,3, 6-18 whereas other breeds are affected sporadically.5, 19-22 Although many reports describe dHS/mHS affecting the CNS,4, 16, 20, 21 localized CNS histiocytic sarcoma has been infrequently reported.19, 20, 22 Three reports describe localized histiocytic sarcoma confined to the CNS of dogs, sparing other visceral organs,19, 20, 22 but bone marrow was not evaluated in these dogs. All 3 dogs were diagnosed with localized HS postmortem, 2 dogs had the CSF evaluated.20, 22 In 1 dog with diffuse leptomeningeal histiocytic sarcoma, the CSF nucleated cell count was 152 cells/μL, with a differential cell count of 50% neutrophils and 50% mononuclear cells and a total protein of 95 mg/dL. No atypical cells were recognized and CSF immunphenotyping was not performed.20 The other dog had a parenchymal brain mass and the cisternal CSF had a total nucleated cell count of 520 cells/μL, with a differential cell count of 7% neutrophils, 17% lymphocytes, and 76% monocytoid cells. The total protein (pyrogallol red methodology) was 51.8 mg/dL. The cytological interpretation was severe monocytoid—predominant, mixed cell pleocytosis, with cellular atypia suggestive of neoplasia.22 Postmortem, archived, unstained, air-dried, cytocentrifuged slides were immunophenotyped. The monocytoid cells were immunoreactive for CD1c and CD11d and were not immunoreactive for T and B cell markers (CD3 and CD79a). The CSF immunocytochemistry results supported the postmortem diagnosis of HS.22 The MR imaging findings were similar in both dogs reported here. In the first dog, the primary tumor was located in the L4–Cd segments whereas in the second dog, the primary tumor arose from the anatomical margins of the hippocampus with an extensive component in both lateral ventricles. In both cases, the tumors were contrast enhancing on T1W images, with diffuse meningeal enhancement in the area of the tumor as well as in other locations within the CNS. In the second dog, peritumoral edema was present. To our knowledge, this is the first report of antemortem diagnosis of primary CNS HS using CSF analysis. Similar to the dogs in previous reports,20, 22 the dogs in this report had marked, mixed, predominantly mononuclear pleocytosis with a moderate increase in the CSF protein concentration. CSF cytology warranted immunophenotyping because the neoplastic cell population was highly pleomorphic, exhibited both round to stellate and spindloid morphology, and was accompanied by a high number of inflammatory leukocytes, which complicated definitive interpretation without concurrent immunophenotyping. The large immature cells in the CSF were immunoreactive for CD1c and CD11c, variably and weakly immunoreactive for CD11b, but not immunoreactive for CD11d, CD3, or CD79a. The cell morphology and immunophenotype were indicative of histiocytic sarcoma of a dendritic cell origin.2, 3 In a report assessing histiocytic sarcomas of a dendritic cell origin in 39 dogs, the neoplastic cells were consistently immunoreactive for CD1c, CD11c, CD18, and CD45. The cell of origin in these tumors is believed to be an interstitial myeloid dendritic antigen-presenting cell. Postmortem immunohistochemistry performed on the CNS tissue from the dogs reported here confirmed the antemortem diagnosis. Neoplastic cells were strongly and diffusely immunoreactive for CD18, strongly and variably immunoreactive for CD45, and not immunoreactive for CD3, CD79a, and CD20. The MR image characteristics and CSF analysis, including immunocytochemical immunophenotyping and postmortem findings, confirmed the presence of a primary HS with metastasis within the CNS, possibly tracking along the meninges. Dogs with localized CNS HS may have a marked CSF pleocytosis with moderately increased protein concentration. When large, discrete atypical cells are present, immunophenotyping of CSF cells may be used to definitively confirm the histiocytic lineage of the atypical cells and hence diagnose HS. In the 2 dogs reported here, the MR findings of a primary tumor with diffuse T1W meningeal enhancement, CSF analysis, and CSF immunophenotyping permitted an antemortem diagnosis of HS, which was confirmed on postmortem evaluation. aGeneral Electric Signa LX, Milwaukee, WI bMagnevist (Gadopentate dimeglumine 469.01 mg/mL), Berlex Laboratories, Wayne, NJ

MR arthrography in chondromalacia patellae diagnosis on a low-field open magnet system

To evaluate existing methods of standardization by exploiting the well-known tissue characterizing property of magnetization transfer ratio (MTR) values obtained from MT imaging, and compare the tissue characterizability of standardized T2, proton density (PD), and T1 images against the MTR images. Image intensity standardization is a postprocessing method that was designed to correct for acquisition-to-acquisition signal intensity variations (nonstandardness) inherent in magnetic resonance (MR) images. The main idea of this technique is to deform the volume image histogram of each study to match a standard histogram, and to utilize the resulting transformations to map the image intensities into a standard scale. The method has been shown to produce a significant gain in similarity of resulting images and to achieve numeric tissue characterization. In this work we compared PD-, T2-, and T1-weighted images before and after standardization with the corresponding MT images for 10 patient MRI studies of the brain, in terms of the normalized median values on the corresponding image histograms. No statistically significant difference was observed between the standardized PD-, T2-, and T1-weighted images and the corresponding MTR images. However, a statistically significant difference was found between the pre- and poststandardized PD-, T2-, and T1-weighted images, and between the prestandardized PD-, T2-, and T1-weighted images and the corresponding MTR images. These results suggest that standardized T2, PD, and T1 images and their tissue-specific intensity signatures may be useful for characterizing disease.

In magnetic resonance imaging (MRI), multi-contrast pulse sequences are routinely acquired, providing complementary information for medical diagnosis. Compared with the single-contrast MR image reconstruction, the multi-contrast MR imaging could further accelerate data acquisition and improve reconstruction quality by leveraging the complementary information of multi-contrast MR images. In this paper, we propose a model-driven graph transformer (MD-GraphFormer) for fast multi-contrast MR imaging, which incorporates the physical constraints of MRI and investigates the complementary information among multi-contrast MR images using graph structure and attention mechanism. The MD-GraphFormer consists of graph-attention-based interaction modules (GAB-IM) and multi-contrast data consistency modules (MC-DCM). GAB-IM learns and interacts the features of multi-contrast MR images over the graph with nodes representing MR contrasts. MC-DCM enforces the consistency between the reconstructed multi-contrast MR images and their corresponding measurements in k-space. Extensive experiments are conducted on the collected raw uMR and SMS brain MRI datasets under different sampling patterns and sampling rates. The results demonstrate that the proposed MD-GraphFormer outperforms the previous multi-contrast MRI reconstruction methods in multi-coil imaging settings.

A multichannel statistical classifier for detecting prostate cancer was developed and validated by combining information from three different magnetic resonance (MR) methodologies: T2-weighted, T2-mapping, and line scan diffusion imaging (LSDI). From these MR sequences, four different sets of image intensities were obtained: T2-weighted (T2W) from T2-weighted imaging, Apparent Diffusion Coefficient (ADC) from LSDI, and proton density (PD) and T2 (T2 Map) from T2-mapping imaging. Manually segmented tumor labels from a radiologist, which were validated by biopsy results, served as tumor "ground truth." Textural features were extracted from the images using co-occurrence matrix (CM) and discrete cosine transform (DCT). Anatomical location of voxels was described by a cylindrical coordinate system. A statistical jack-knife approach was used to evaluate our classifiers. Single-channel maximum likelihood (ML) classifiers were based on 1 of the 4 basic image intensities. Our multichannel classifiers: support vector machine (SVM) and Fisher linear discriminant (FLD), utilized five different sets of derived features. Each classifier generated a summary statistical map that indicated tumor likelihood in the peripheral zone (PZ) of the prostate gland. To assess classifier accuracy, the average areas under the receiver operator characteristic (ROC) curves over all subjects were compared. Our best FLD classifier achieved an average ROC area of 0.839(+/-0.064), and our best SVM classifier achieved an average ROC area of 0.761(+/-0.043). The T2W ML classifier, our best single-channel classifier, only achieved an average ROC area of 0.599(+/-0.146). Compared to the best single-channel ML classifier, our best multichannel FLD and SVM classifiers have statistically superior ROC performance (P=0.0003 and 0.0017, respectively) from pairwise two-sided t-test. By integrating the information from multiple images and capturing the textural and anatomical features in tumor areas, summary statistical maps can potentially aid in image-guided prostate biopsy and assist in guiding and controlling delivery of localized therapy under image guidance.

Parallel magnetic resonance imaging (pMRI) which utilizes the redundant information from different coil sensitivities has been widely used. As the number of coils increases, algorithms that reconstruct single combined image are more suitable for pMRI. Most single combined image reconstruction need explicit functions of coil sensitivity. As a result, estimation of the coil sensitivities is important in pMRI. Nowadays, ESPIRiT(Iterative self-consistent parallel imaging reconstruction using eigenvector maps) based on the estimation of coil sensitivities is an outstanding algorithm. However, ESPIRiT has its limitation, which estimates the coil sensitivities not very correctly, while the autocalibration data is insufficient. In this paper, we focus on the pMRI algorithms based on improving the estimation of coil sensitivities and propose a new algorithm called GESPIRiT (ESPIRiT combined with GRAPPA), which is based on combining ESPIRiT with GRAPPA to obtain better estimation of coil sensitivities and final reconstructed image. Experiments are applied to demonstrate its feasibility and efficiency.

Contrastive learning, a successful form of representational learning, has shown promising results in pretraining deep learning (DL) models for downstream tasks. When working with limited annotation data, as in medical image segmentation tasks, learning domain-specific local representations can further improve the performance of DL models. In this work, we extend the contrastive learning framework to utilize domain-specific contrast information from unlabeled Magnetic Resonance (MR) images to improve the performance of downstream MR image segmentation tasks in the presence of limited labeled data. The contrast in MR images is controlled by underlying tissue properties (e.g., T1 or T2) and image acquisition parameters. We hypothesize that learning to discriminate local representations based on underlying tissue properties should improve subsequent segmentation tasks on MR images. We propose a novel constrained contrastive learning (CCL) strategy that uses tissue-specific information via a constraint map to define positive and negative local neighborhoods for contrastive learning, embedding this information in the representational space during pretraining. For a given MR contrast image, the proposed strategy uses local signal characteristics (constraint map) across a set of related multi-contrast MR images as a surrogate for underlying tissue information. We demonstrate the utility of the approach for downstream: (1) multi-organ segmentation tasks in T2-weighted images where a DL model learns T2 information with constraint maps from a set of 2D multi-echo T2-weighted images (n=101) and (2) tumor segmentation tasks in multi-parametric images from the public brain tumor segmentation (BraTS) (n=80) dataset where DL models learn T1 and T2 information from multi-parametric BraTS images. Performance is evaluated on downstream multi-label segmentation tasks with limited data in (1) T2-weighted images of the abdomen from an in-house Radial-T2 (Train/Test=30/20), (2) public Cartesian-T2 (Train/Test=6/12) dataset, and (3) multi-parametric MR images from the public brain tumor segmentation dataset (BraTS) (Train/Test=40/50). The performance of the proposed CCL strategy is compared to state-of-the-art self-supervised contrastive learning techniques. In each task, a model is also trained using all available labeled data for supervised baseline performance. The proposed CCL strategy consistently yielded improved Dice scores, Precision, and Recall metrics, and reduced HD95 values across all segmentation tasks. We also observed performance comparable to the baseline with reduced annotation effort. The t-SNE visualization of features for T2-weighted images demonstrates its ability to embed T2 information in the representational space. On the BraTS dataset, we also observed that using an appropriate multi-contrast space to learn T1+T2, T1, or T2 information during pretraining further improved the performance of tumor segmentation tasks. Learning to embed tissue-specific information that controls MR image contrast with the proposed constrained contrastive learning improved the performance of DL models on subsequent segmentation tasks compared to conventional self-supervised contrastive learning techniques. The use of such domain-specific local representations could help understand, improve performance, and mitigate the scarcity of labeled data in MR image segmentation tasks.

Our goal was to find MRI signs of use for identifying a spinal arachnoid diverticulum. Three cases of spinal arachnoid diverticula, one extradural and two intradural, were examined on a 1.5 T imager. There was obvious mass effect on the adjacent structures in one case and increased signal intensity in the diverticulum on proton density- and T2-weighted images in two cases. Signal changes due to turbulent movement of the spinal fluid inside the diverticula were seen in all cases on sagittal fast spin-echo (FSE) proton density- and T2-weighted images; it was difficult to tell whether these signal changes imply a communication or are simply FSE artefacts. On contrast-enhanced studies, all cases showed partial enhancement inside the diverticula. There thus are four signs of diverticula: mass effect, the increased signal, signal void sign and partial enhancement; the last of these, the most reliable, has never been reported before.

We proposed a method for segmentation of brain tissues—gray matter, white matter, and cerebrospinal fluid—using multi-contrast images, including a T1 map and a uniform T1-weighted image, from a magnetization-prepared 2 rapid acquisition gradient echoes (MP2RAGE) sequence at 7 Tesla. The proposed method was evaluated with respect to the processing time and the similarity of the segmented masks of brain tissues with those obtained using FSL, FreeSurfer, and SPM12. The processing time of the proposed method (28 ± 0 s) was significantly shorter than those of FSL and SPM12 (444 ± 4 s and 159 ± 2 s for FSL and SPM12, respectively). In the similarity assessment, the tissue mask of the brain obtained by the proposed method showed higher consistency with those obtained using FSL than with those obtained using SPM12. The proposed method misclassified the subcortical structures and large vessels since it is based on the intensities of multi-contrast images obtained using MP2RAGE, which uses a similar segmentation approach as FSL but is not based on a template image or a parcellated brain atlas, which are used for FreeSurfer and SPM12, respectively. However, the proposed method showed good segmentation in the cerebellum and white matter in the medial part of the brain in comparison with the other methods. Thus, because the proposed method using different contrast images of MP2RAGE sequence showed the shortest processing time and similar segmentation ability as the other methods, it may be useful for both neuroimaging research and clinical diagnosis.

BackgroundSeveral recent reports have demonstrated that the preoperative sensitivity and accuracy of identifying and locating non-palpable testes increases with the use of conventional MRI, in addition to diffusion-weighted imaging (DWI). Therefore, pre-operative prediction of the presence and location of testes using imaging techniques may guide management of intra-abdominal testis. Fowler-Stephens orchiopexy is effective for treating patients with intra-abdominal testis; however, long-term testicular function after this procedure has not been clarified. We present a case of a high intra-abdominal testis located below the kidney, and discuss the usefulness of fusion view with T2-weighted and DWI images to make a diagnosis of high intra-abdominal testis and the pathological findings to predict future fertility potential.Case presentationA 10-month-old boy was referred to the urology department for the management of non-palpable testis. We employed not only conventional MRI, but also DWI, to improve the diagnostic accuracy of non-palpable testes by MRI examination. The high-intensity mass-like structure below the kidney on the T2-weighted image and the markedly high signal intensity mass on the DWI image completely matched, which suggested that the mass below the kidney was the right testis. The patient underwent diagnostic and therapeutic laparoscopy. A testis was found under the ascending colon, 1 cm below the right kidney. We performed 2-stage Fowler-Stephens orchiopexy. The testis could be delivered to the scrotum without any tension. We examined expression patterns of the stem cell marker, undifferentiated embryonic cell transcription factor 1 (UTF1) in the testicular biopsy sample, and demonstrated that the UTF1-positive Ad spermatogonia / negative Ad spermatogonia ratio was lower in this patient than in boys his age with descended and inguinal undescended testes, indicating that spermatogonial stem cell activity may decrease remarkably in this boy.ConclusionsFusion view with T2-weighted and DWI images may be a useful diagnostic modality for high intra-abdominal testes. Fowler-Stephens orchiopexy may provide blood supply to the testis but that might not be enough to achieve spermatogenesis.

The Backward Wave Propagation (BWP) algorithm provides a means for reconstructing an image from spatial wave samples measured across an aperture at discrete transducer locations. The technique performs a spatial wave number decomposition on the aperture samples. This decomposition can be phase adjusted and retransformed back into the space domain taking into account the constraints of the wave equation. In this manner the image “scanned” by the aperture can be reconstructed using primarily Fast Fourier Transform (FFT) techniques. Use of parallel processing incorporating AMD 29500 series VLSI integrated circuits has been described previously to achieve image reconstruction at video rates with each frame derived from a single illuminating pulse. This paper explores the accuracy of the BWP image reconstruction by comparison with known interference patterns. The effects of finite word length arithmetic and noise corruption are examined. The conclusion suggests that the use of the new AMD 29325 series 32 bit floating point circuits will lead to a more robust image reconstruction. The previously reported system has been redesigned to incorporate these new circuits. These new building blocks illustrate once again the dramatic effect VLSI can have on the design of ultrasonic imaging systems.