Late lesion growth following endovascular therapy: is 24 hours too early to assess acute infarct size including the effects of secondary injury?

Introduction Stroke lesion volume on MRI or CT provides objective evidence of tissue injury as a consequence of ischemic stroke. Measurement of "final" lesion volume at 24hr following endovascular therapy (post-EVT) has been used in multiple studies as a surrogate for clinical outcome. However, despite successful recanalization, a significant proportion of patients do not experience favorable clinical outcome. The goals of this study were to quantify lesion growth during the first week after treatment, identify early predictors, and explore the association with clinical outcome. Methods This is a prospective study of stroke patients at two centers who met the following criteria: i) anterior large vessel occlusion (LVO) acute ischemic stroke, ii) attempted EVT, and iii) had 3T MRI post-EVT at 24hr and 5-day. We defined "Early" and "Late" lesion growth as ≥10mL lesion growth between baseline and 24hr DWI, and between 24hr DWI and 5-day FLAIR, respectively. Complete reperfusion was defined as >90% reduction of the volume of tissue with perfusion delay (Tmax>6sec) between pre-EVT and 24hr post-EVT. Favorable clinical outcome was defined as modified Rankin scale (mRS) of 0-2 at 30 or 90 days. Results One hundred twelve patients met study criteria with median age 67 years, 56% female, median admit NIHSS 19, 54% received IV or IA thrombolysis, 66% with M1 occlusion, and median baseline DWI volume 21.2mL. Successful recanalization was achieved in 87% and 68% had complete reperfusion, with an overall favorable clinical outcome rate of 53%. Nearly two thirds (65%) of the patients did not have Late lesion growth with a median volume change of -0.3mL between 24hr and 5-days and an associated high rate of favorable clinical outcome (64%). However, ~1/3 of patients (35%) did have significant Late lesion growth despite successful recanalization (87%: 46% mTICI 2b/ 41% mTICI 3). Late lesion growth patients had a 27.4mL change in Late lesion volume and 30.1mL change in Early lesion volume. These patients had an increased hemorrhagic transformation rate of 68% with only 1 in 3 patients having favorable clinical outcome. Late lesion growth was independently associated with incomplete reperfusion, hemorrhagic transformation, and unfavorable outcome. Conclusion Approximately 1 out of 3 patients had Late lesion growth following EVT, with a favorable clinical outcome occurring in only 1 out of 3 of these patients. Most patients with no Early lesion growth had no Late lesion growth. Identification of patients with Late lesion growth could be critical to guide clinical management and inform prognosis post-EVT. Additionally, it can serve as an imaging biomarker for the development of adjunctive therapies to mitigate reperfusion injury.

Fully automated patellofemoral MRI segmentation using holistically nested networks: Implications for evaluating patellofemoral osteoarthritis, pain, injury, pathology, and adolescent development.

PURPOSE: Our clinical understanding of the relationship between 3D bone morphology and knee osteoarthritis, as well as our ability to investigate potential causative factors of osteoarthritis, has been hampered by the time-intensive nature of manually segmenting bone from MR images. Thus, we aim to develop and validate a fully automated deep learning framework for segmenting the patella and distal femur cortex, in both adults and actively growing adolescents. METHODS: Data from 93 subjects, obtained from on institutional review board-approved protocol, formed the study database. 3D sagittal gradient recalled echo and gradient recalled echo with fat saturation images and manual models of the outer cortex were available for 86 femurs and 90 patellae. A deep-learning-based 2D holistically nested network (HNN) architecture was developed to automatically segment the patella and distal femur using both single (sagittal, uniplanar) and 3 cardinal plane (triplanar) methodologies. Errors in the surface-to-surface distances and the Dice coefficient were the primary measures used to quantitatively evaluate segmentation accuracy using a 9-fold cross-validation. RESULTS: Average absolute errors for segmenting both the patella and femur were 0.33 mm. The Dice coefficients were 97% and 94% for the femur and patella. The uniplanar, relative to the triplanar, methodology produced slightly superior segmentation. Neither the presence of active growth plates nor pathology influenced segmentation accuracy. CONCLUSION: The proposed HNN with multi-feature architecture provides a fully automatic technique capable of delineating the often indistinct interfaces between the bone and other joint structures with an accuracy better than nearly all other techniques presented previously, even when active growth plates are present.

Automatic quadriceps and patellae segmentation of MRI with cascaded U2 -Net and SASSNet deep learning model.

PURPOSE: Automatic muscle segmentation is critical for advancing our understanding of human physiology, biomechanics, and musculoskeletal pathologies, as it allows for timely exploration of large multi-dimensional image sets. Segmentation models are rarely developed/validated for the pediatric model. As such, autosegmentation is not available to explore how muscle architectural changes during development and how disease/pathology affects the developing musculoskeletal system. Thus, we aimed to develop and validate an end-to-end, fully automated, deep learning model for accurate segmentation of the rectus femoris and vastus lateral, medialis, and intermedialis using a pediatric database. METHODS: We developed a two-stage cascaded deep learning model in a coarse-to-fine manner. In the first stage, the U2 -Net roughly detects the muscle subcompartment region. Then, in the second stage, the shape-aware 3D semantic segmentation method SASSNet refines the cropped target regions to generate the more finer and accurate segmentation masks. We utilized multifeature image maps in both stages to stabilize performance and validated their use with an ablation study. The second-stage SASSNet was independently run and evaluated with three different cropped region resolutions: the original image resolution, and images downsampled 2× and 4× (high, mid, and low). The relationship between image resolution and segmentation accuracy was explored. In addition, the patella was included as a comparator to past work. We evaluated segmentation accuracy using leave-one-out testing on a database of 3D MR images (0.43 × 0.43 × 2 mm) from 40 pediatric participants (age 15.3 ± 1.9 years, 55.8 ± 11.8 kg, 164.2 ± 7.9 cm, 38F/2 M). RESULTS: The mid-resolution second stage produced the best results for the vastus medialis, rectus femoris, and patella (Dice similarity coefficient = 95.0%, 95.1%, 93.7%), whereas the low-resolution second stage produced the best results for the vastus lateralis and vastus intermedialis (DSC = 94.5% and 93.7%). In comparing the low- to mid-resolution cases, the vasti intermedialis, vastus medialis, rectus femoris, and patella produced significant differences (p = 0.0015, p = 0.0101, p < 0.0001, p = 0.0003) and the vasti lateralis did not (p = 0.2177). The high-resolution stage 2 had significantly lower accuracy (1.0 to 4.4 dice percentage points) compared to both the mid- and low-resolution routines (p value ranged from < 0.001 to 0.04). The one exception was the rectus femoris, where there was no difference between the low- and high-resolution cases. The ablation study demonstrated that the multifeature is more reliable than the single feature. CONCLUSIONS: Our successful implementation of this two-stage segmentation pipeline provides a critical tool for expanding pediatric muscle physiology and clinical research. With a relatively small and variable dataset, our fully automatic segmentation technique produces accuracies that matched or exceeded the current state of the art. The two-stage segmentation avoids memory issues and excessive run times by using a first stage focused on cropping out unnecessary data. The excellent Dice similarity coefficients improve upon previous template-based automatic and semiautomatic methodologies targeting the leg musculature. More importantly, with a naturally variable dataset (size, shape, etc.), the proposed model demonstrates slightly improved accuracies, compared to previous neural networks methods.

Fully automated prostate whole gland and central gland segmentation on MRI using holistically nested networks with short connections.

Accurate and automated prostate whole gland and central gland segmentations on MR images are essential for aiding any prostate cancer diagnosis system. Our work presents a 2-D orthogonal deep learning method to automatically segment the whole prostate and central gland from T2-weighted axial-only MR images. The proposed method can generate high-density 3-D surfaces from low-resolution ( z axis) MR images. In the past, most methods have focused on axial images alone, e.g., 2-D based segmentation of the prostate from each 2-D slice. Those methods suffer the problems of over-segmenting or under-segmenting the prostate at apex and base, which adds a major contribution for errors. The proposed method leverages the orthogonal context to effectively reduce the apex and base segmentation ambiguities. It also overcomes jittering or stair-step surface artifacts when constructing a 3-D surface from 2-D segmentation or direct 3-D segmentation approaches, such as 3-D U-Net. The experimental results demonstrate that the proposed method achieves 92.4 % ± 3 % Dice similarity coefficient (DSC) for prostate and DSC of 90.1 % ± 4.6 % for central gland without trimming any ending contours at apex and base. The experiments illustrate the feasibility and robustness of the 2-D-based holistically nested networks with short connections method for MR prostate and central gland segmentation. The proposed method achieves segmentation results on par with the current literature.

Volumetric neuroimage analysis extensions for the MIPAV software package.

We describe a new collection of publicly available software tools for performing quantitative neuroimage analysis. The tools perform semi-automatic brain extraction, tissue classification, Talairach alignment, and atlas-based measurements within a user-friendly graphical environment. They are implemented as plug-ins for MIPAV, a freely available medical image processing software package from the National Institutes of Health. Because the plug-ins and MIPAV are implemented in Java, both can be utilized on nearly any operating system platform. In addition to the software plug-ins, we have also released a digital version of the Talairach atlas that can be used to perform regional volumetric analyses. Several studies are conducted applying the new tools to simulated and real neuroimaging data sets.

Pre-operative estimation of esophageal tumor metabolic length in FDG-PET images with surgical pathology confirmation.

OBJECTIVE: The aim of this study was to evaluate a 3D tumor segmentation method for fluorodeoxyglucose positron emission tomography (FDG-PET) in the context of noninvasive estimation of tumor metabolic length (Lm), as it correlates with surgical pathology and phantom results. METHODS: Thirty-four patients (7 women, 27 men) with esophageal cancer were retrospectively evaluated. All patients underwent FDG-PET-computed tomography (CT) imaging following endoscopic ultrasound (EUS). Seventeen patients had esophagectomy after PET/CT, without prior neoadjuvant therapy. Tumor length was assessed by EUS (Le, n=31) and histopathology (Lp, n=17). Images were evaluated quantitatively with a 3D threshold-based region-growing program (Medical Image Processing Analysis and Visualization). Lm, total metabolic volume (Vm), maximum standardized uptake value (SUVmax), and average SUV (SUVa) over the entire tumor were obtained for several threshold values (mean activity in the liver plus 0-, 1-, 2-, 3-, and 4-SD of the activity in the liver). RESULTS: Lm showed a good correlation with Lp for all thresholds (best correlation for Lm(2-SD), r=0.74, P<0.001). A positive nonsignificant correlation was observed between Lp and Le (r=0.30, P=0.29). Lm(2-SD) correlated well with Le (r=0.71, P<0.001). Good correlations were also observed between Lm(2-SD) and Vm(2-SD) (r=0.89, P<0.001) and SUVa(2-SD) (r=0.38, P<0.05). Vm(2-SD) also had a significant correlation with Lp (r=0.61, P<0.05) and Le (r=0.57, P<0.001). CONCLUSIONS: FDG-PET-derived tumor metabolic length of untreated esophageal carcinomas correlates well with surgical pathology results, and provides preliminary evidence that noninvasive delineation of the superior and inferior extent of viable tumor involvement might be feasible using computer-generated metabolic length measurements.

Automatic magnetic resonance prostate segmentation by deep learning with holistically nested networks.

Accurate automatic segmentation of the prostate in magnetic resonance images (MRI) is a challenging task due to the high variability of prostate anatomic structure. Artifacts such as noise and similar signal intensity of tissues around the prostate boundary inhibit traditional segmentation methods from achieving high accuracy. We investigate both patch-based and holistic (image-to-image) deep-learning methods for segmentation of the prostate. First, we introduce a patch-based convolutional network that aims to refine the prostate contour which provides an initialization. Second, we propose a method for end-to-end prostate segmentation by integrating holistically nested edge detection with fully convolutional networks. Holistically nested networks (HNN) automatically learn a hierarchical representation that can improve prostate boundary detection. Quantitative evaluation is performed on the MRI scans of 250 patients in fivefold cross-validation. The proposed enhanced HNN model achieves a mean ± standard deviation. A Dice similarity coefficient (DSC) of [Formula: see text] and a mean Jaccard similarity coefficient (IoU) of [Formula: see text] are used to calculate without trimming any end slices. The proposed holistic model significantly ([Formula: see text]) outperforms a patch-based AlexNet model by 9% in DSC and 13% in IoU. Overall, the method achieves state-of-the-art performance as compared with other MRI prostate segmentation methods in the literature.

GPU based multi-histogram volume navigation for virtual bronchoscopy.

An interactive navigation system for virtual bronchoscopy is presented, which is based solely on GPU based high performance multi-histogram volume rendering.

Dual-view plane illumination microscopy for rapid and spatially isotropic imaging.

We describe the construction and use of a compact dual-view inverted selective plane illumination microscope (diSPIM) for time-lapse volumetric (4D) imaging of living samples at subcellular resolution. Our protocol enables a biologist with some prior microscopy experience to assemble a diSPIM from commercially available parts, to align optics and test system performance, to prepare samples, and to control hardware and data processing with our software. Unlike existing light sheet microscopy protocols, our method does not require the sample to be embedded in agarose; instead, samples are prepared conventionally on glass coverslips. Tissue culture cells and Caenorhabditis elegans embryos are used as examples in this protocol; successful implementation of the protocol results in isotropic resolution and acquisition speeds up to several volumes per s on these samples. Assembling and verifying diSPIM performance takes ∼6 d, sample preparation and data acquisition take up to 5 d and postprocessing takes 3-8 h, depending on the size of the data.

Atlas based AAM and SVM model for fully automatic MRI prostate segmentation.

Automatic prostate segmentation in MR images is a challenging task due to inter-patient prostate shape and texture variability, and the lack of a clear prostate boundary. We propose a supervised learning framework that combines the atlas based AAM and SVM model to achieve a relatively high segmentation result of the prostate boundary. The performance of the segmentation is evaluated with cross validation on 40 MR image datasets, yielding an average segmentation accuracy near 90%.