Search

Segment Anything Model (SAM) has revolutionized the way of segmentation. However, SAM's performance may decline when applied to tasks involving domains that differ from natural images. Nonetheless, by employing fine-tuning techniques, SAM exhibits promising capabilities in specific domains, such as medicine and planetary science. Notably, there is a lack of research on the application of SAM to sonar imaging. In this paper, we aim to address this gap by conducting a comprehensive investigation of SAM's performance on sonar images. Specifically, we evaluate SAM using various settings on sonar images. Additionally, we fine-tune SAM using effective methods both with prompts and for semantic segmentation, thereby expanding its applicability to tasks requiring automated segmentation. Experimental results demonstrate a significant improvement in the performance of the fine-tuned SAM., Comment: 12 pages, 3 figures

Three-dimensional (3D) images, such as CT, MRI, and PET, are common in medical imaging applications and important in clinical diagnosis. Semantic ambiguity is a typical feature of many medical image labels. It can be caused by many factors, such as the imaging properties, pathological anatomy, and the weak representation of the binary masks, which brings challenges to accurate 3D segmentation. In 2D medical images, using soft masks instead of binary masks generated by image matting to characterize lesions can provide rich semantic information, describe the structural characteristics of lesions more comprehensively, and thus benefit the subsequent diagnoses and analyses. In this work, we introduce image matting into the 3D scenes to describe the lesions in 3D medical images. The study of image matting in 3D modality is limited, and there is no high-quality annotated dataset related to 3D matting, therefore slowing down the development of data-driven deep-learning-based methods. To address this issue, we constructed the first 3D medical matting dataset and convincingly verified the validity of the dataset through quality control and downstream experiments in lung nodules classification. We then adapt the four selected state-of-the-art 2D image matting algorithms to 3D scenes and further customize the methods for CT images. Also, we propose the first end-to-end deep 3D matting network and implement a solid 3D medical image matting benchmark, which will be released to encourage further research., 12 pages, 7 figures

Additional file 2: Tabls S2. The gene number and DAVID enrichment score in each module.

Additional file 3: Fig. S1. Module preservation between even partitionings of datasets.

Additional file 1: Tabls S1. The number of genes in each module.

It is difficult to accurately label ambiguous and complex shaped targets manually by binary masks. The weakness of binary mask under-expression is highlighted in medical image segmentation, where blurring is prevalent. In the case of multiple annotations, reaching a consensus for clinicians by binary masks is more challenging. Moreover, these uncertain areas are related to the lesions' structure and may contain anatomical information beneficial to diagnosis. However, current studies on uncertainty mainly focus on the uncertainty in model training and data labels. None of them investigate the influence of the ambiguous nature of the lesion itself.Inspired by image matting, this paper introduces alpha matte as a soft mask to represent uncertain areas in medical scenes and accordingly puts forward a new uncertainty quantification method to fill the gap of uncertainty research for lesion structure. In this work, we introduce a new architecture to generate binary masks and alpha mattes in a multitasking framework, which outperforms all state-of-the-art matting algorithms compared. The proposed uncertainty map is able to highlight the ambiguous regions and a novel multitasking loss weighting strategy we presented can improve performance further and demonstrate their concrete benefits. To fully-evaluate the effectiveness of our proposed method, we first labelled three medical datasets with alpha matte to address the shortage of available matting datasets in medical scenes and prove the alpha matte to be a more efficient labeling method than a binary mask from both qualitative and quantitative aspects., Comment: Jounal version of our MICCAI 2021 paper, in review

Additional file 1: Kaplan-Meier curve of module (a)M6, (b)M16, (c)G1, (d)G12,(e)W13 and (f)Q7.

Knowledge of the situation of the topography of a plain in a country is an important issue for planning rural fitting out and investigate geology layers and hydrology state. This study is based on the analysis of the topography leading to explore the physical state of the environment by using different structures. The methodology used was consisted to perform statistical analysis by linear regression, to study density and structural dynamics as well a characterization of the ground by performing the 3D. As results, the analysis procedure over contour lines and TIN structures showed significant dynamics of the physical surface. The dynamics of these oscillations of ground gave place to characterize the geological, hydrological and topographical state of the region. In the geological concept, the 3D mode of contour lines structure has showed the determination of the thicknesses of the outcrop layers. However, the TIN structure clearly defined the level of degradation of environment as well the draining axes constituting the hydrological network.

Water management in agricultural areas requires knowledge of flow directions as well as lost water recuperation areas. The Inner Delta of Niger threatened by desertification, despite its great agricultural potential, loses an enormous quantity of runoff water every year. In this paper, the study focuses on the detection of flow directions and areas susceptible to be flooded. For this purpose, two new approaches have been applied, namely “geometrical interval and profile interpolation” to identify flow directions and flood zones. The flow directions and flood zones of the inner Delta of Niger were obtained from SRTM image (Shuttle Radar Topography Mission) using ArcMap software. The obtained results from the surfaces classification technique “geometrical Interval” have permitted to identify the lowest areas between 258 and 264 m altitude occupying 58% of the total area. Thereafter, the areas favorable for profile tracing are identified through the stretched colour tone. The flow paths leading to these flood zones are detected by the “longitudinal profiles” technique showing downward slope profiles. The whole area dominated by rocky outcrops and by small scattered hills, is endowed of lowlands including a large slit of southwest to northeast direction.

The problem that the underwater robot is winded by fishing nets is the most concerned. At present, there are still no effective methods which can detect fishing nets in real time to avoid obstacles. As fishing net images taken with traditional detection technique are blurry and have low definition, it is hard to identify fishing nets. In view of this phenomenon, we proposed a new underwater detection method in this paper and used underwater laser scanning system to collect clear fishing net images. Due to the high cost of data acquisition and the single amount of experiment data, we took advantage of a deep generative network to amplify original data. In fishing nets detection network, we used a residual network as new base net to deepen the network layers, then combined the method of regression with some region suggestion to detect fishing nets, which can improve the accuracy of fishing nets detection without affecting the real-time performance. Finally, the contrast experiment of laser fishing net images is carried out to verify the effectiveness of the proposed method.

In view of the unclear outline of the object characteristics and the complexity of the background noise in sonar images, an object detection algorithm is proposed based on pixel hierarchy. According to the characteristics of depth map, "depth information" is inserted into different layers of the sonar image to describe the “depth” of each layer. Due to the higher gray values of the objects, and the lower gray values of the background, the sonar image is layered according to different gray-levels. The high gray values of sonar image are divided into several areas and then the areas are marked. Using the high gray value characteristics of the object pixels in each image layers, a coefficient for characterizing the possibility of containing the object pixels in each image layer is defined so that the final object segmentation is made easier. Experimental results show that the object segmentation algorithm is accurate and fast.

In order to ensure the real-time property of the virtual surgery system and improve the model rendering speed, we propose an acceleration collision detection algorithm based on invisible separated surface. Meanwhile, to reduce the local deformation and the loss of important information caused by virtual cutting maximally in the medical preoperative diagnosis and surgery planning, we propose a surface model cutting algorithm on the basis of removing the intersecting grid cell. Under the premise of ensuring the system's real-time property, the algorithm deals the triangle mesh with a series of operation like subdividing, projection and so on. The experimental results show that the collision detection method has improved the action speed of virtual equipment touching with the model surface. The cutting algorithm can be implemented easily and retain the original information better, which is expected to help to improve the accuracy of the diagnosis judged by the doctor according to the patient's condition.

For the linear discrete-time time-invariant system with unknown noise statistics, a new estimator is presented in this paper, that is using multisensor which has the same measurement matrix to measure, same measurements form the white noise sequences. Using the correlated functions matrix of these sequences, the measurements noise variances Ri of the subsystems can be estimated, and then the input variances Qw can be estimated by the moving average (MA) innovation model. It is proved that the parameters estimation converges to the real parameter with probability 1. In this paper, the global optimality weighted measurement fusion Kalman filter is introduced first, and then the new estimator is presented, and finally the self-tuning weighted measurement fusion Kalman filter is shown. A simulation example for a tracking system with 3-sensor shows its effectiveness.

MIS (minimum invasive surgery) helps a lot for surgeons and patients and reduces the operating time and patients' pains. But lack of tactile sense makes surgeons hard to do the operation. They must be very cautious to insert the catheter. With the aim to get the lost tactile information, scientists and engineers have developed several mechanisms and tactile sensors to help surgeons to do the operation. In this paper, a novel tactile sensor based on PVDF has been developed to measure the pressure on side wall of the catheter which was not paid enough attention in previous. With this sensor, surgeons can "feel" the contact force between the side wall of the catheter and blood vessel. Thus, medical accidents can be avoided greatly by utilizing this sensor in the surgery. The principle, interface circuit and calibrating experiment of the developed sensor was presented in this paper. The result of experiments shows that our tactile sensor has a very sensitive response and suits for minimum invasive surgery.

The location of the vehicle plate is the core and most difficult technique in the vehicle plate auto-recognition system. In this paper, a novel algorithm of searching color point-pairs in locating vehicle plate is proposed. On the other hand, to decrease the time used to search the color point-pairs, the edge detection technique of image processing is used to pre-locate the areas the color point-pairs belong to. Through the experiment, it is proved that the algorithm proposed is valid and the results are satisfiable.