Fast PET Scan Tumor Segmentation Using Superpixels, Principal Component Analysis and K-Means Clustering

A New Tool for Extracting Static and Dynamic Parameters from [18F]F-DOPA PET/CT in Pediatric Gliomas

Background/Objectives: PET imaging with [18F]F-DOPA has demonstrated high potential for the evaluation and management of pediatric brain gliomas. Manual extraction of PET parameters is time-consuming, lacks reproducibility, and varies with operator experience. Methods: In this study, we tested whether a semi-automated image processing framework could overcome these limitations. Pediatric patients with available static and/or dynamic [18F]F-DOPA PET studies were evaluated retrospectively. We developed a Python software to automate clinical index calculations, including preprocessing to delineate tumor volumes from structural MRI, accounting for lesions with low [18F]F-DOPA uptake. A total of 73 subjects with treatment-naïve low- and high-grade gliomas, who underwent brain MRI within two weeks of [18F]F-DOPA PET, were included and analyzed. Static analysis was conducted on all subjects, while dynamic analysis was performed on 32 patients. Results: For 68 subjects, the Intraclass Correlation Coefficient for T/S between manual and ground truth segmentation was 0.91. Using our tool, ICC improved to 0.94. Our method demonstrated good reproducibility in extracting static tumor-to-striatum ratio (p = 0.357); however, significant differences were observed in tumor slope (p < 0.05). No significant differences were found in time-to-peak (p = 0.167) and striatum slope (p = 0.36). Conclusions: Our framework aids in analyzing [18F]F-DOPA PET images of pediatric brain tumors by automating clinical score extraction, simplifying segmentation and Time Activity Curve extraction, reducing user variability, and enhancing reproducibility.

PAPNet: Convolutional network for pancreatic cyst segmentation

BACKGROUND

Automatic segmentation of the pancreas and its tumor region is a prerequisite for computer-aided diagnosis.

OBJECTIVE

In this study, we focus on the segmentation of pancreatic cysts in abdominal computed tomography (CT) scan, which is challenging and has the clinical auxiliary diagnostic significance due to the variability of location and shape of pancreatic cysts.

METHODS

We propose a convolutional neural network architecture for segmentation of pancreatic cysts, which is called pyramid attention and pooling on convolutional neural network (PAPNet). In PAPNet, we propose a new atrous pyramid attention module to extract high-level features at different scales, and a spatial pyramid pooling module to fuse contextual spatial information, which effectively improves the segmentation performance.

RESULTS

The model was trained and tested using 1,346 CT slice images obtained from 107 patients with the pathologically confirmed pancreatic cancer. The mean dice similarity coefficient (DSC) and mean Jaccard index (JI) achieved using the 5-fold cross-validation method are 84.53% and 75.81%, respectively.

CONCLUSIONS

The experimental results demonstrate that the proposed new method in this study enables to achieve effective results of pancreatic cyst segmentation.

Cancerous and Non-Cancerous Brain MRI Classification Method Based on Convolutional Neural Network and Log-Polar Transformation

Magnetic resonance imaging (MRI) offers visual representations of the interior of a body for clinical analysis and medical intervention. The MRI process is subjected to a variety of image processing and machine learning approaches to identify, diagnose, and classify brain diseases as well as detect abnormalities. In this paper, we propose an improved classification method for distinguishing cancerous and noncancerous tumors from brain MRI images by using Log Polar Transformation (LPT) and convolutional neural networks (CNN). The LPT has been applied for feature extraction of rotation and scaling of distorted images, while the integration of CNN introduces a machine learning approach for the tumor classification of distorted images. The dataset was formed with images of seven different brain diseases, and the training set was formed by applying CNN with the extracted features. The proposed method is then evaluated in comparison to state-of-the-art algorithms, showing a definite improvement of the former. The obtained results show that the machine learning approach offers better classification with a success rate of about 96% in both plain brain MR images and rotation- and scale-invariant brain MR images. This work also successfully classified T-1 and T-2 weighted images of neoplastic and degenerative brain diseases. The obtained accuracy is perfected by several kernel procedures, while the combined performance of the two wavelet transformations and a strong dataset make our method robust and efficient. Since no earlier study on machine learning approaches with rotated and scaled brain MRI has come to our attention, it is expected that our proposed method introduces a new paradigm in this research field.

Object-Based Convolutional Neural Networks for Cloud and Snow Detection in High-Resolution Multispectral Imagers

Cloud and snow detection is one of the most significant tasks for remote sensing image processing. However, it is a challenging task to distinguish between clouds and snow in high-resolution multispectral images due to their similar spectral distributions. The shortwave infrared band (SWIR, e.g., Sentinel-2A 1.55–1.75 µm band) is widely applied to the detection of snow and clouds. However, high-resolution multispectral images have a lack of SWIR, and such traditional methods are no longer practical. To solve this problem, a novel convolutional neural network (CNN) to classify cloud and snow on an object level is proposed in this paper. Specifically, a novel CNN structure capable of learning cloud and snow multiscale semantic features from high-resolution multispectral imagery is presented. In order to solve the shortcoming of “salt-and-pepper” in pixel level predictions, we extend a simple linear iterative clustering algorithm for segmenting high-resolution multispectral images and generating superpixels. Results demonstrated that the new proposed method can with better precision separate the cloud and snow in the high-resolution image, and results are more accurate and robust compared to the other methods.