VOILA: Complexity-Aware Universal Segmentation of CT Images by Voxel Interacting with Language

In this study, we present WebCeph2k, an extensive and diverse cephalometric landmark localization dataset that surpasses previous benchmark datasets in terms of number of landmark annotations. This diverse cephalometric landmarks dataset has significant value in medical imaging research. Existing studies predominantly focus on datasets obtained from a single medical center and provider, which offers a limited number of landmarks and a limited diversity of cephalograms, resulting in models that exhibit low robustness and generalization when applied to more diverse datasets. The clinical application of cephalometry is hampered by significant localization errors in landmark localization models, in addition to the inadequacy of existing datasets’ landmarks for clinical cephalometric diagnosis. The limited generalization ability and the occurrence of “overfitting” in deep learning models are mainly caused by the small size in the dataset. In the medical field, the inclusion of large and diverse datasets can greatly improve the generalization and performance of landmark localization models. This paper presents our WebCeph2k dataset from 9 medical centers, covering 9 different imaging devices, which surpasses the only publicly available ISBI2015 dataset in terms of sample size and number of landmarks. In addition, this study employs a low computational cost methodology to achieve optimal landmarks localization: 1) ROI regions of X-ray images are derived by exploiting the prior distribution of the data, 2) the model computational cost is reduced by adopting a spatial-depth transformation strategy, 3) the standard heatmap decoding method is optimized by integrating a compensation strategy. The results show that the proposed method not only achieves competitive localization results to other state-of-the-art approaches, but also offers a reduction of the model computational cost, resulting in faster inference. Consequently, this research offers valuable prospects in the field of general-purpose medical landmark localization methods. We also find that our proposed dataset is more complex and challenging than the ISBI dataset. The dataset and code are available at https://github.com/switch626/WebCeph2k.

Computed tomography (CT) is an important imaging technique used in medical analysis of the internal structure of the human body. Previously, image segmentation methods were required after acquiring reconstructed CT images to obtain segmented CT images which made it susceptible to errors from both reconstruction and segmentation algorithms. However, this paper introduces a new approach using an advanced quantum optimization algorithm called quadratic unconstrained binary optimization (QUBO) for CT image segmentation. This algorithm allows CT image reconstruction and segmentation to be performed simultaneously. This algorithm segments CT images by minimizing the difference between a sinogram in a superposition state with qubits, obtained using the mathematical projection including the Radon transform, and the experimentally acquired sinogram from X-ray images for various angles. Furthermore, we leveraged X-ray mass attenuation coefficients to reduce the number of logical qubits required for our quantum optimization algorithm, and we employed D-Wave’s hybrid solver to solve the optimization problem. We compared the segmentation results of our algorithm with those of classical algorithms using X-ray images of actual tooth samples to validate the results of our algorithm. The comparison revealed that, after undergoing appropriate image post-processing, our algorithm’s segmentation results matched those of classical algorithms that perform segmentation after reconstruction, except for some pixels at the boundary. We expect that the new quantum optimization CT algorithm will bring about great advancements in medical imaging.

Computed tomography (CT) is extensively used for accurate visualization and segmentation of organs and lesions. While deep learning models such as convolutional neural networks (CNNs) and vision transformers (ViTs) have significantly improved CT image analysis, their performance often declines when applied to diverse, real-world clinical data. Although foundation models offer a broader and more adaptable solution, their potential is limited due to the challenge of obtaining large-scale, voxel-level annotations for medical images. In response to these challenges, prompting-based models using visual or text prompts have emerged. Visual-prompting methods, such as the Segment Anything Model (SAM), still require significant manual input and can introduce ambiguity when applied to clinical scenarios. Instead, foundation models that use text prompts offer a more versatile and clinically relevant approach. Notably, current text-prompt models, such as the CLIP-Driven Universal Model, are limited to text prompts already encountered during training and struggle to process the complex and diverse scenarios of real-world clinical applications. Instead of fine-tuning models trained from natural imaging, we propose OpenVocabCT, a vision-language model pretrained on large-scale 3D CT images for universal text-driven segmentation. Using the large-scale CT-RATE dataset, we decompose the diagnostic reports into fine-grained, organ-level descriptions using large language models for multi-granular contrastive learning. We evaluate our OpenVocabCT on downstream segmentation tasks across 14 public datasets and 1 institutional dataset for organ and tumor segmentation, demonstrating the superior performance of our model compared to existing methods. All code, datasets, and models will be publicly released at https://github.com/ricklisz/OpenVocabCT.

Image Guided Radiation Therapy for Prostate IMRT: Rectum Volume Changes and Dosimetric Considerations

In recent years, the machine learning field has been inundated with a variety of deep learning methods. Different deep learning model types, including recurrent neural networks (RNNs), convolutional neural networks (CNNs), adversarial neural networks (ANNs), and autoencoders, are successfully tackling challenging computer vision problems including image detection and segmentation in an unconstrained environment. Although image segmentation has received a lot of interest, there have been several new deep learning methods discovered with regard to object detection and recognition. An academic review of deep learning image segmentation methods is presented in this article. In this study, the major goal is to offer a sensible comprehension of the basic approaches that have already made a substantial contribution to the domain of image segmentation throughout the years. The article describes the existing state of image segmentation, and goes on to make the argument that deep learning has revolutionized this field. Afterwards, segmentation algorithms have been scientifically classified and optimized, each with their own special contribution. With a variety of informative narratives, the reader may be able to understand the internal workings of these processes more quickly.

We develop a novel method for image segmentation of 3D confocal microscopy images of emerging hematopoietic stem cells. The method is based on the theory of persistent homology and uses an optimal threshold to select the most persistent cycles in the persistence diagram. This enables the segmentation of the image's most contrasted and representative shapes. Coupling this segmentation method with a meshing algorithm, we define a pipeline for 3D reconstruction of confocal volumes. Compared to related methods, this approach improves shape segmentation, is more ergonomic to automatize, and has fewer parameters. We apply it to the segmentation of membranes, at subcellular resolution, of cells involved in the endothelial-to-hematopoietic transition (EHT) in the zebrafish embryos.

Image-guided brachytherapy for cervical cancer: A Canadian Brachytherapy Group survey

Horizontal and vertical search artificial bee colony for image segmentation of COVID-19 X-ray images

Evolutionary algorithms are used in many engineering applications for optimization of problems that are often difficult to solve using conventional methods. One such problem is image segmentation. This task is used for object (contour) extraction from images to create sensible representation of the image. There are many image segmentation and optimization methods. This work is focused on selected evolutionary optimization methods. Namely, particle swarm optimization, genetic algorithm, and differential evolution. Our image segmentation method is inspired in algorithm known as k-means. The optimization function from k-means algorithm is replaced by evolutionary technique. We compare original k-means algorithm with evolutionary approaches and we show that our evolutionary approaches easily outperform the classical approach.

Research on infrared image enhancement and segmentation of power equipment based on partial differential equation

In this study, we aimed to develop a new automated method for kidney volume measurement in children using ultrasonography (US) with image pre-processing and hybrid learning and to formulate an equation to calculate the expected kidney volume. The volumes of 282 kidneys (141 subjects, <19 years old) with normal function and structure were measured using US. The volumes of 58 kidneys in 29 subjects who underwent US and computed tomography (CT) were determined by image segmentation and compared to those calculated by the conventional ellipsoidal method and CT using intraclass correlation coefficients (ICCs). An expected kidney volume equation was developed using multivariate regression analysis. Manual image segmentation was automated using hybrid learning to calculate the kidney volume. The ICCs for volume determined by image segmentation and ellipsoidal method were significantly different, while that for volume calculated by hybrid learning was significantly higher than that for ellipsoidal method. Volume determined by image segmentation was significantly correlated with weight, body surface area, and height. Expected kidney volume was calculated as (2.22 × weight (kg) + 0.252 × height (cm) + 5.138). This method will be valuable in establishing an age-matched normal kidney growth chart through the accumulation and analysis of large-scale data.

Anatomical changes can occur during course of head-and-neck (H and N) radiotherapy like tumor shrinkage, decreased edema and/or weight loss. This can lead to discrepancies in planned and delivered dose increasing the dose to organs at risk. A study was conducted to determine the volumetric and dosimetric changes with the help of repeat computed tomography (CT) and replanning for selected H and N cancer patients treated with IMRT plans to see for these effects. In 15 patients with primary H and N cancer, a repeat CT scan after 3rd week of radiotherapy was done when it was clinically indicated and then two plans were generated on repeat CT scan, actual plan (AP) planned on repeat CT scan, and hybrid plan (HP), which was generated by applying the first intensity-modulated radiation therapy (IMRT) plan (including monitoring units) to the images of second CT scan. Both plans (AP and HP) on repeat CT scan were compared for volumetric and dosimetric parameter. The mean variation in volumes between CT and repeat CT were 44.32 cc, 82.2 cc, and 149.83 cc for gross tumor volume (GTV), clinical target volumes (CTV), and planning target volume (PTV), respectively. Mean conformity index and homogeneity index was 0.68 and 1.07, respectively for AP and 0.5 and 1.16, respectively for HP. Mean D95 and D99 of PTV was 97.92% (standard deviation, SD 2.32) and 93.4% (SD 3.75), respectively for AP and 92.8% (SD 3.83) and 82.8% (SD 8.0), respectively for HP. Increase in mean doses to right parotid, left parotid, spine, and brainstem were 5.56 Gy (Dmean), 3.28 Gy (Dmean), 1.25 Gy (Dmax), and 3.88 Gy (Dmax), respectively in HP compared to AP. Repeat CT and replanning reduces the chance of discrepancies in delivered dose due to volume changes and also improves coverage to target volume and further reduces dose to organ at risk.

Radiotherapy dose calculation requires accurate Computed Tomography (CT) imaging while tissue delineation may necessitate the use of contrast agents (CA). Acquiring these two sets is a common practice in radiotherapy. This study aims to evaluate the effect of CA on the dose calculations. Two hundred and twenty-six volumetric modulated arc therapy (VMAT) patients that had planning CT with contrast (CCT) and non-contrast CT (NCCT) of different cancer sites (e.g., brain, head, and neck (H&amp;N), chest, abdomen, and pelvis) were evaluated. Treatment plans were recalculated using CCT, then compared to NCCT. The variation in Hounsfield units (HU) and dose distributions for critical structures and target volumes were analyzed using mean HU, mean and maximum relative dose values, D2%, D98%, and 3D gamma analysis. HU variations were statistically significant for most structures. However, this was not clinically significant as the difference in mean HU values was within 30 HU for soft tissue and 50 HU for lungs. Variation in target volumes’ D2% and D98% were insignificant for all sites except brain and nasopharynx. Dose maximum differences were within 2% for the majority of critical structures and target volumes. 3D gamma analysis results revealed that majority of plans satisfied the 2% and 2 mm criteria. CCT may be acquired for VMAT radiotherapy planning purposes instead of NCCT, since there is no clinically significant difference in dose calculations based on either image set.

Image segmentation and edge detection methods have proven to be very useful in a variety of fields, including tissue volume estimation, medical diagnosis, anatomical structure analysis, treatment planning, and so on. Due to several problems, image segmentation remains a contentious subject. To begin with, the majority of image segmentation methods are problem-based. Second, the highly comparable grey levels and textures in medical photos limit most edge detection algorithms. This project aims to provide a framework for extracting several items of interest from CT scans while also utilizing previous information. In a multi-agent context, our technique leverages the characteristics of each agent.Each local agent divides the input image into sub-images and attempts, using a priori knowledge, to label each pixel as a specific location. During this process, the local agent stamps each cell in the subimago individually. To generate the final segmented image, the moderator agent examines the results of all agents' activities.

Clustering-based segmentation algorithm is one of the common methods in image segmentation. However, how to avoid the influence of the initialization and the singular value is a problem we have to face. For overcoming this problem, we proposed a new image segmentation technology which is based on the AIC criterion and not affected by the singular values. The simulation results show that the method has good performance on stability and accuracy. Image segmentation is the key step in image analysis of computer vision and image processing field. It has been widely applied in various areas such as in the process of automating production, sensing images and biomedical images. The main goal of image segmentation is to simplify an image into segments that have a strong correlation with objects in the real world. The existing segmentation algorithm can be generally divided into three classes: edge detection, region- based segmentation and data clustering. Edge detection method relies on the edges of the object with background and the other objects. For the images with noise, it is difficult to separate the boundary and the target background (1). Region- based segmentation method segments an image into several meaningful sub-domains that are non-overlapping and same- nature. But it may lead to the over-segmentation, and has to combine with other methods to use (2). Data clustering method is based on the whole image and considers the distance between each data. Advantages of data clustering method are low complexity and easy to implement. Disadvantages are sensitive to the noise and selection of the initial centroids (3). Data clustering is one of the common techniques in image segmentation. Its purpose is to cluster pixels into several parts (regions) according to the feature of image. Based on the properties of clustering algorithm, researchers have proposed various image segmentation algorithm, such as based on hierarchical clustering, based on K-means clustering, based on fuzzy clustering, based on mean shift, and so on (3-5). Clustering analysis is an unsupervised process of partitioning a data set into subsets of similar data objects. The elementary principle is as far as possible to increase the difference between categories and reduce the difference within the category. In the process of image segmentation based on data clustering, thus the objective is that the similar pixels are divided into the same regions and the non-similar pixels are divided into different regions. However, Two difficulties we must face are that how to determine the cluster centers and avoid influence of noise. In this paper, a new image segmentation algorithm is proposed based on the AIC criterion. The new segmentation technology can achieve an ideal performance and not affected by the initialization of centroids and the singular values. The simulation results verify the efficiency of proposed algorithm.