Search

Aim A pressure ulcer is localized injury to the skin and/or underlying tissue usually over a bony prominence, as a result of pressure, or pressure in combination with shear. Although the external forces and bony prominences differ depending on ulcer location, the way in which these anatomical differences affect pressure ulcer development is not well studied. Methods To clarify the location-dependent factors for pressure ulcer development, we focused on superficial injuries that develop over an undermining lesion, which we have termed them bilayer pressure ulcers. Because it is thought that a deep pressure ulcer is caused by ischemia at the deep lesion and a shallow pressure ulcer is caused by shear force to the superficial skin, a bilayer pressure ulcer can be considered a mixed phenotype, induced by both pressure and shear force. We retrospectively examined the frequency of bilayer pressure ulcers by location in a total of 568 pressure ulcers. Results The ratio of bilayer pressure ulcers to deep pressure ulcers staged III or more was significantly larger for pressure ulcers over the sacrum. Conclusion A new concept, the relative position between the external force and bony prominence, could explain the frequency and developmental mechanism of bilayer pressure ulcers. The external forces, shape of the bony prominence, and mobility of the soft tissue may be responsible for this concept.

Background: Significant heterogeneity exists in target volumes for postoperative stereotactic radiosurgery (SRS) for brain metastases. A set of contouring guidelines was recently published, and we investigated the impact of deviations. Methods: Patients (n = 41) undergoing single-fraction Gamma Knife SRS following surgical resection of brain metastases from 2011 to 2017 were retrospectively reviewed. SRS included the entire contrast-enhancing cavity with heterogeneity in inclusion of the surgical tract and no routine margin along the dura or clinical target volume margin. Follow-up MR imaging was fused with SRS plans to assess patterns of failure. Results: The median follow-up was 11.1 months with a median prescription of 18 Gy. There were 5 local failures: infield (n = 3, 60%), surgical tract (n = 1, 20%), and marginal > 5 mm from the resection cavity (n = 1, 20%). No marginal failures < 5 mm or dural margin failures were noted. For deep lesions (n = 13), 62% (n = 8) had the entire tract covered. The only tract recurrence was in a deep lesion without coverage of the surgical tract (n = 1/5). Conclusion: In this small preliminary experience, despite no routine inclusion of the dural tract or bone flap, no failures were noted in these locations. Omission of the surgical tract in deep lesions may increase failure rates.

Lesion detection from computed tomography (CT) scans is challenging compared to natural object detection because of two major reasons: small lesion size and small inter-class variation. Firstly, the lesions usually only occupy a small region in the CT image. The feature of such small region may not be able to provide sufficient information due to its limited spatial feature resolution. Secondly, in CT scans, the lesions are often indistinguishable from the background since the lesion and non-lesion areas may have very similar appearances. To tackle both problems, we need to enrich the feature representation and improve the feature discriminativeness. Therefore, we introduce a dual-attention mechanism to the 3D contextual lesion detection framework, including the cross-slice contextual attention to selectively aggregate the information from different slices through a soft re-sampling process. Moreover, we propose intra-slice spatial attention to focus the feature learning in the most prominent regions. Our method can be easily trained end-to-end without adding heavy overhead on the base detection network. We use DeepLesion dataset and train a universal lesion detector to detect all kinds of lesions such as liver tumors, lung nodules, and so on. The results show that our model can significantly boost the results of the baseline lesion detector (with 3D contextual information) but using much fewer slices.

Radiologists in their daily work routinely find and annotate significant abnormalities on a large number of radiology images. Such abnormalities, or lesions, have collected over years and stored in hospitals' picture archiving and communication systems. However, they are basically unsorted and lack semantic annotations like type and location. In this paper, we aim to organize and explore them by learning a deep feature representation for each lesion. A large-scale and comprehensive dataset, DeepLesion, is introduced for this task. DeepLesion contains bounding boxes and size measurements of over 32K lesions. To model their similarity relationship, we leverage multiple supervision information including types, self-supervised location coordinates and sizes. They require little manual annotation effort but describe useful attributes of the lesions. Then, a triplet network is utilized to learn lesion embeddings with a sequential sampling strategy to depict their hierarchical similarity structure. Experiments show promising qualitative and quantitative results on lesion retrieval, clustering, and classification. The learned embeddings can be further employed to build a lesion graph for various clinically useful applications. We propose algorithms for intra-patient lesion matching and missing annotation mining. Experimental results validate their effectiveness., Accepted by CVPR2018. DeepLesion url added

Volumetric lesion segmentation from computed tomography (CT) images is a powerful means to precisely assess multiple time-point lesion/tumor changes. However, because manual 3D segmentation is prohibitively time consuming, current practices rely on an imprecise surrogate called response evaluation criteria in solid tumors (RECIST). Despite their coarseness, RECIST markers are commonly found in current hospital picture and archiving systems (PACS), meaning they can provide a potentially powerful, yet extraordinarily challenging, source of weak supervision for full 3D segmentation. Toward this end, we introduce a convolutional neural network (CNN) based weakly supervised slice-propagated segmentation (WSSS) method to (1) generate the initial lesion segmentation on the axial RECIST-slice; (2) learn the data distribution on RECIST-slices; (3) extrapolate to segment the whole lesion slice by slice to finally obtain a volumetric segmentation. To validate the proposed method, we first test its performance on a fully annotated lymph node dataset, where WSSS performs comparably to its fully supervised counterparts. We then test on a comprehensive lesion dataset with 32,735 RECIST marks, where we report a mean Dice score of 92% on RECIST-marked slices and 76% on the entire 3D volumes.

The surgical management of deep brain tumors is often challenging due to the limitations of stereotactic needle biopsies and the morbidity associated with transcortical approaches. We present a novel microscopic navigational technique utilizing the Viewsite Brain Access System (VBAS) (Vycor Medical, Boca Raton, FL, USA) for resection of a deep parietal periventricular high-grade glioma as well as another glioma and a cavernoma with no related morbidity. The approach utilized a navigational tracker mounted on a microscope, which was set to the desired trajectory and depth. It allowed gentle continuous insertion of the VBAS directly to a deep lesion under continuous microscopic visualization, increasing safety by obviating the need to look up from the microscope and thus avoiding loss of trajectory. This technique has broad value for the resection of a variety of deep brain lesions.