Computer-assisted surgical planning and intraoperative guidance in fetal surgery: a systematic review.

Fetal surgery has become a clinical reality, with interventions for twin-to-twin transfusion syndrome (TTTS) and spina bifida demonstrated to improve outcome. Fetal imaging is evolving, with the use of 3D ultrasound and fetal MRI becoming more common in clinical practise. Medical imaging analysis is also changing, with technology being developed to assist surgeons by creating 3D virtual models that improve understanding of complex anatomy, and prove powerful tools in surgical planning and intraoperative guidance. We introduce the concept of computer-assisted surgical planning, and present the results of a systematic review of image reconstruction for fetal surgical planning that identified six articles using such technology. Indications from other specialities suggest a benefit of surgical planning and guidance to improve outcomes. There is therefore an urgent need to develop fetal-specific technology in order to improve fetal surgical outcome.

DeepIGeoS: A Deep Interactive Geodesic Framework for Medical Image Segmentation.

Accurate medical image segmentation is essential for diagnosis, surgical planning and many other applications. Convolutional Neural Networks (CNNs) have become the state-of-the-art automatic segmentation methods. However, fully automatic results may still need to be refined to become accurate and robust enough for clinical use. We propose a deep learning-based interactive segmentation method to improve the results obtained by an automatic CNN and to reduce user interactions during refinement for higher accuracy. We use one CNN to obtain an initial automatic segmentation, on which user interactions are added to indicate mis-segmentations. Another CNN takes as input the user interactions with the initial segmentation and gives a refined result. We propose to combine user interactions with CNNs through geodesic distance transforms, and propose a resolution-preserving network that gives a better dense prediction. In addition, we integrate user interactions as hard constraints into a back-propagatable Conditional Random Field. We validated the proposed framework in the context of 2D placenta segmentation from fetal MRI and 3D brain tumor segmentation from FLAIR images. Experimental results show our method achieves a large improvement from automatic CNNs, and obtains comparable and even higher accuracy with fewer user interventions and less time compared with traditional interactive methods.

Direct, live imaging of cortical spreading depression and anoxic depolarisation using a fluorescent, voltage-sensitive dye.

Perilesion depolarisations, whether transient anoxic depolarisation (AD) or spreading depression (SD), occur in stroke models and in patients with acute brain ischaemia, but their contribution to lesion progression remains unclear. As these phenomena correspond to waves of cellular depolarisation, we have developed a technique for their live imaging with a fluorescent voltage-sensitive (VS) dye (RH-1838). Method development and validation were performed in two different preparations: chicken retina, to avoid any vascular interference; and cranial window exposing the cortical surface of anaesthetised rats. Spreading depression was produced by high-K medium, and AD by complete terminal ischaemia in rats. After dye loading, the preparation was illuminated at its excitation wavelength and fluorescence changes were recorded sequentially with a charge-coupled device camera. No light was recorded when the VS dye was omitted, ruling out the contribution of any endogenous fluorophore. With both preparations, the changes in VS dye fluorescence with SD were analogous to those of the DC (direct current) potential recorded with glass electrodes. Although some blood quenching of the emitted light was identified, the VS dye signatures of SD had a good signal-to-noise ratio and were reproducible. The changes in VS dye fluorescence associated with AD were more complex because of additional interferents, especially transient brain swelling with subsequent shrinkage. However, the kinetics of the AD-associated changes in VS dye fluorescence was also analogous to that of the DC potential. In conclusion, this method provides the imaging equivalent of electrical extracellular DC potential recording, with the SD and AD negative shifts translating directly to fluorescence increase.

Interactive Medical Image Segmentation Using Deep Learning With Image-Specific Fine Tuning.

Convolutional neural networks (CNNs) have achieved state-of-the-art performance for automatic medical image segmentation. However, they have not demonstrated sufficiently accurate and robust results for clinical use. In addition, they are limited by the lack of image-specific adaptation and the lack of generalizability to previously unseen object classes (a.k.a. zero-shot learning). To address these problems, we propose a novel deep learning-based interactive segmentation framework by incorporating CNNs into a bounding box and scribble-based segmentation pipeline. We propose image-specific fine tuning to make a CNN model adaptive to a specific test image, which can be either unsupervised (without additional user interactions) or supervised (with additional scribbles). We also propose a weighted loss function considering network and interaction-based uncertainty for the fine tuning. We applied this framework to two applications: 2-D segmentation of multiple organs from fetal magnetic resonance (MR) slices, where only two types of these organs were annotated for training and 3-D segmentation of brain tumor core (excluding edema) and whole brain tumor (including edema) from different MR sequences, where only the tumor core in one MR sequence was annotated for training. Experimental results show that: 1) our model is more robust to segment previously unseen objects than state-of-the-art CNNs; 2) image-specific fine tuning with the proposed weighted loss function significantly improves segmentation accuracy; and 3) our method leads to accurate results with fewer user interactions and less user time than traditional interactive segmentation methods.

Three-Dimensional Ultrasonic Needle Tip Tracking with a Fiber-Optic Ultrasound Receiver.

Ultrasound is frequently used for guiding minimally invasive procedures, but visualizing medical devices is often challenging with this imaging modality. When visualization is lost, the medical device can cause trauma to critical tissue structures. Here, a method to track the needle tip during ultrasound image-guided procedures is presented. This method involves the use of a fiber-optic ultrasound receiver that is affixed within the cannula of a medical needle to communicate ultrasonically with the external ultrasound probe. This custom probe comprises a central transducer element array and side element arrays. In addition to conventional two-dimensional (2D) B-mode ultrasound imaging provided by the central array, three-dimensional (3D) needle tip tracking is provided by the side arrays. For B-mode ultrasound imaging, a standard transmit-receive sequence with electronic beamforming is performed. For ultrasonic tracking, Golay-coded ultrasound transmissions from the 4 side arrays are received by the hydrophone sensor, and subsequently the received signals are decoded to identify the needle tip's spatial location with respect to the ultrasound imaging probe. As a preliminary validation of this method, insertions of the needle/hydrophone pair were performed in clinically realistic contexts. This novel ultrasound imaging/tracking method is compatible with current clinical workflow, and it provides reliable device tracking during in-plane and out-of-plane needle insertions.

GIFT-Cloud: A data sharing and collaboration platform for medical imaging research.

OBJECTIVES: Clinical imaging data are essential for developing research software for computer-aided diagnosis, treatment planning and image-guided surgery, yet existing systems are poorly suited for data sharing between healthcare and academia: research systems rarely provide an integrated approach for data exchange with clinicians; hospital systems are focused towards clinical patient care with limited access for external researchers; and safe haven environments are not well suited to algorithm development. We have established GIFT-Cloud, a data and medical image sharing platform, to meet the needs of GIFT-Surg, an international research collaboration that is developing novel imaging methods for fetal surgery. GIFT-Cloud also has general applicability to other areas of imaging research. METHODS: GIFT-Cloud builds upon well-established cross-platform technologies. The Server provides secure anonymised data storage, direct web-based data access and a REST API for integrating external software. The Uploader provides automated on-site anonymisation, encryption and data upload. Gateways provide a seamless process for uploading medical data from clinical systems to the research server. RESULTS: GIFT-Cloud has been implemented in a multi-centre study for fetal medicine research. We present a case study of placental segmentation for pre-operative surgical planning, showing how GIFT-Cloud underpins the research and integrates with the clinical workflow. CONCLUSIONS: GIFT-Cloud simplifies the transfer of imaging data from clinical to research institutions, facilitating the development and validation of medical research software and the sharing of results back to the clinical partners. GIFT-Cloud supports collaboration between multiple healthcare and research institutions while satisfying the demands of patient confidentiality, data security and data ownership.

Imaging the human placental microcirculation with micro-focus computed tomography: Optimisation of tissue preparation and image acquisition.

Micro-CT provides 3D volume imaging with spatial resolution at the micrometre scale. We investigated the optimal human placenta tissue preparation (contrast agent, perfusion pressure, perfusion location and perfusion vessel) and imaging (energy, target material, exposure time and frames) parameters. Microfil (Flow Tech, Carver, MA) produced better fill than Barium sulphate (84.1%(±11.5%)vs70.4%(±18.02%) p = 0.01). Perfusion via umbilical artery produced better fill than via chorionic vessels (83.8%(±17.7%)vs78.0%(±21.9%), p < 0.05), or via umbilical vein (83.8%(±16.4%)vs69.8%(±20.3%), p < 0.01). Imaging at 50 keV with a molybdenum target produced the best contrast to noise ratio. We propose this method to enable quantification and comparison of the human fetoplacental vascular tree.

Slic-Seg: A minimally interactive segmentation of the placenta from sparse and motion-corrupted fetal MRI in multiple views.

Segmentation of the placenta from fetal MRI is challenging due to sparse acquisition, inter-slice motion, and the widely varying position and shape of the placenta between pregnant women. We propose a minimally interactive framework that combines multiple volumes acquired in different views to obtain accurate segmentation of the placenta. In the first phase, a minimally interactive slice-by-slice propagation method called Slic-Seg is used to obtain an initial segmentation from a single motion-corrupted sparse volume image. It combines high-level features, online Random Forests and Conditional Random Fields, and only needs user interactions in a single slice. In the second phase, to take advantage of the complementary resolution in multiple volumes acquired in different views, we further propose a probability-based 4D Graph Cuts method to refine the initial segmentations using inter-slice and inter-image consistency. We used our minimally interactive framework to examine the placentas of 16 mid-gestation patients from MRI acquired in axial and sagittal views respectively. The results show the proposed method has 1) a good performance even in cases where sparse scribbles provided by the user lead to poor results with the competitive propagation approaches; 2) a good interactivity with low intra- and inter-operator variability; 3) higher accuracy than state-of-the-art interactive segmentation methods; and 4) an improved accuracy due to the co-segmentation based refinement, which outperforms single volume or intensity-based Graph Cuts.

Performance characteristics of an interventional multispectral photoacoustic imaging system for guiding minimally invasive procedures.

Precise device guidance is important for interventional procedures in many different clinical fields including fetal medicine, regional anesthesia, interventional pain management, and interventional oncology. While ultrasound is widely used in clinical practice for real-time guidance, the image contrast that it provides can be insufficient for visualizing tissue structures such as blood vessels, nerves, and tumors. This study was centered on the development of a photoacoustic imaging system for interventional procedures that delivered excitation light in the ranges of 750 to 900 nm and 1150 to 1300 nm, with an optical fiber positioned in a needle cannula. Coregistered B-mode ultrasound images were obtained. The system, which was based on a commercial ultrasound imaging scanner, has an axial resolution in the vicinity of 100 µm and a submillimeter, depth-dependent lateral resolution. Using a tissue phantom and 800 nm excitation light, a simulated blood vessel could be visualized at a maximum distance of 15 mm from the needle tip. Spectroscopic contrast for hemoglobin and lipids was observed with ex vivo tissue samples, with photoacoustic signal maxima consistent with the respective optical absorption spectra. The potential for further optimization of the system is discussed.