Thermography as a Method for Bedside Monitoring of Infantile Hemangiomas.

Infantile hemangiomas occur in 3 to 10% of infants. To predict the clinical course and counsel on treatment, it is crucial to accurately determine the hemangiomas' extension, volume, and location. However, this can represent a challenge because hemangiomas may present irregular patterns or be covered by hair, or their depth may be difficult to estimate. Diagnosis is commonly made by clinical inspection and palpation, with physicians basing their diagnoses on visual characteristics such as area, texture, and color. Doppler ultrasonography or magnetic resonance imaging are normally used to estimate depth or to confirm difficult assessments. This paper presents an alternative diagnosis tool-thermography-as a useful, immediate means of carrying out accurate hemangioma examinations. We conducted a study analyzing infantile hemangiomas with a custom thermographic system. In the first phase of the study, 55 hemangiomas of previously diagnosed patients were analyzed with a thermal camera over several sessions. An average temperature variation before and after treatment of -0.19 °C was measured. In the second phase, we selected nine patients and assessed their evolution over nine months by analyzing their thermographic images and implementing dedicated image processing algorithms. In all cases, we found that the thermal image analysis concurred with the independent diagnoses of two dermatologists. We concluded that a higher temperature inside the tumor in the follow-up was indicative of an undesirable evolution.

Joint segmentation of bones and muscles using an intensity and histogram-based energy minimization approach.

BACKGROUND AND OBJECTIVES: The segmentation of muscle and bone structures in CT is of interest to physicians and surgeons for surgical planning, disease diagnosis and/or the analysis of fractures or bone/muscle densities. Recently, the issue has been addressed in many research works. However, most studies have focused on only one of the two tissues and on the segmentation of one particular bone or muscle. This work addresses the segmentation of muscle and bone structures in 3D CT volumes. METHODS: The proposed bone and muscle segmentation algorithm is based on a three-label convex relaxation approach. The main novelty is that the proposed energy function to be minimized includes distance to histogram models of bone and muscle structures combined with gray-level information. RESULTS: 27 CT volumes corresponding to different sections from 20 different patients were manually segmented and used as ground-truth for training and evaluation purposes. Different metrics (Dice index, Jaccard index, Sensitivity, Specificity, Positive Predictive Value, accuracy and computational cost) were computed and compared with those used in some state-of-the art algorithms. The proposed algorithm outperformed the other methods, obtaining a Dice coefficient of 0.88 ±â€¯0.14, a Jaccard index of 0.80 ±â€¯0.19, a Sensitivity of 0.94 ±â€¯0.15 and a Specificity of 0.95 ±â€¯0.04 for bone segmentation, and 0.78 ±â€¯0.12, 0.65 ±â€¯0.16, 0.94 ±â€¯0.04 and 0.95 ±â€¯0.04 for muscle tissue. CONCLUSIONS: A fast, generalized method has been presented for segmenting muscle and bone structures in 3D CT volumes using a multilabel continuous convex relaxation approach. The results obtained show that the proposed algorithm outperforms some state-of-the art methods. The algorithm will help physicians and surgeons in surgical planning, disease diagnosis and/or the analysis of fractures or bone/muscle densities.

Three-dimensional segmentation of retroperitoneal masses using continuous convex relaxation and accumulated gradient distance for radiotherapy planning.

An innovative algorithm has been developed for the segmentation of retroperitoneal tumors in 3D radiological images. This algorithm makes it possible for radiation oncologists and surgeons semiautomatically to select tumors for possible future radiation treatment and surgery. It is based on continuous convex relaxation methodology, the main novelty being the introduction of accumulated gradient distance, with intensity and gradient information being incorporated into the segmentation process. The algorithm was used to segment 26 CT image volumes. The results were compared with manual contouring of the same tumors. The proposed algorithm achieved 90 % sensitivity, 100 % specificity and 84 % positive predictive value, obtaining a mean distance to the closest point of 3.20 pixels. The algorithm's dependence on the initial manual contour was also analyzed, with results showing that the algorithm substantially reduced the variability of the manual segmentation carried out by different specialists. The algorithm was also compared with four benchmark algorithms (thresholding, edge-based level-set, region-based level-set and continuous max-flow with two labels). To the best of our knowledge, this is the first time the segmentation of retroperitoneal tumors for radiotherapy planning has been addressed.

Linearized multidimensional earth-mover's-distance gradient flows.

This paper presents the first framework capable of performing active contour segmentation using Earth Mover's Distance (EMD) to measure dissimilarity between multidimensional feature distributions. EMD is the best known and understood cross-bin histogram distance measure, and as such it allows for meaningful comparisons between distributions, unlike bin-to-bin measures that only account for discrepancies on a bin-to-bin basis. Because EMD is obtained with linear programming techniques, its differential structure with respect to variations in bin weights as the active contour evolves is expressed through sensitivity analysis. Euler-Lagrange equations are then derived from the computed sensitivity at every iteration to produce gradient descent flows. We validate our approach with color image segmentation, in comparison with state-of-the-art Bhattacharyya (bin-to-bin) and 1D EMD (cross-bin) active contours. Some unique advantages of cross-bin comparison are highlighted in our segmentation results: better perceptual value and increased robustness with respect to the initialization.

Mapping from frame-driven to frame-free event-driven vision systems by low-rate rate coding and coincidence processing--application to feedforward ConvNets.

Event-driven visual sensors have attracted interest from a number of different research communities. They provide visual information in quite a different way from conventional video systems consisting of sequences of still images rendered at a given "frame rate." Event-driven vision sensors take inspiration from biology. Each pixel sends out an event (spike) when it senses something meaningful is happening, without any notion of a frame. A special type of event-driven sensor is the so-called dynamic vision sensor (DVS) where each pixel computes relative changes of light or "temporal contrast." The sensor output consists of a continuous flow of pixel events that represent the moving objects in the scene. Pixel events become available with microsecond delays with respect to "reality." These events can be processed "as they flow" by a cascade of event (convolution) processors. As a result, input and output event flows are practically coincident in time, and objects can be recognized as soon as the sensor provides enough meaningful events. In this paper, we present a methodology for mapping from a properly trained neural network in a conventional frame-driven representation to an event-driven representation. The method is illustrated by studying event-driven convolutional neural networks (ConvNet) trained to recognize rotating human silhouettes or high speed poker card symbols. The event-driven ConvNet is fed with recordings obtained from a real DVS camera. The event-driven ConvNet is simulated with a dedicated event-driven simulator and consists of a number of event-driven processing modules, the characteristics of which are obtained from individually manufactured hardware modules.

Fast vision through frameless event-based sensing and convolutional processing: application to texture recognition.

Address-event representation (AER) is an emergent hardware technology which shows a high potential for providing in the near future a solid technological substrate for emulating brain-like processing structures. When used for vision, AER sensors and processors are not restricted to capturing and processing still image frames, as in commercial frame-based video technology, but sense and process visual information in a pixel-level event-based frameless manner. As a result, vision processing is practically simultaneous to vision sensing, since there is no need to wait for sensing full frames. Also, only meaningful information is sensed, communicated, and processed. Of special interest for brain-like vision processing are some already reported AER convolutional chips, which have revealed a very high computational throughput as well as the possibility of assembling large convolutional neural networks in a modular fashion. It is expected that in a near future we may witness the appearance of large scale convolutional neural networks with hundreds or thousands of individual modules. In the meantime, some research is needed to investigate how to assemble and configure such large scale convolutional networks for specific applications. In this paper, we analyze AER spiking convolutional neural networks for texture recognition hardware applications. Based on the performance figures of already available individual AER convolution chips, we emulate large scale networks using a custom made event-based behavioral simulator. We have developed a new event-based processing architecture that emulates with AER hardware Manjunath's frame-based feature recognition software algorithm, and have analyzed its performance using our behavioral simulator. Recognition rate performance is not degraded. However, regarding speed, we show that recognition can be achieved before an equivalent frame is fully sensed and transmitted.