Segmentation Step Research Articles

Lung computed tomography image segmentation based on U-Net network fused with dilated convolution

PurposeIn order to solve the problem of accurate and effective segmentation of the patient's lung computed tomography (CT) images, so as to improve the efficiency of treating lung cancer. MethodWe propose a U-Net network (DC-U-Net) fused with dilated convolution, and compare the results of segmented lung CT with DC-U-Net, Otsu and region growth. We use Intersection over Union (IOU), Dice coefficient, Precision and Recall to evaluate the performance of the three algorithms. ResultsCompared with the common segmentation algorithm Otsu and region growing, the segmented image of DC-U-Net is closer to the Ground truth. The IOU of DC-U-Net is 0.9627, and the Dice coefficient is 0.9743, which is close to 1 and much higher than the other two algorithms. ConclusionWe propose that the model can directly segment the original image automatically, and the segmentation effect is good. This model speeds up the segmentation, simplifies the steps of medical image segmentation, and provides better segmentation for subsequent lung blood vessels, trachea and other tissues.

Adaptive thresholding technique based classification of red blood cell and sickle cell using Naïve Bayes Classifier and K-nearest neighbor classifier

Detection of anomalous cells by analyzing the microscopic blood smear plays a key role in identification of various blood diseases in medical field. It becomes easier with robust segmentation steps followed by a classifier for automatic recognition of normal as well as anomalous blood cells. Sickle cell anemia is a disease that changes shape of the red blood cell (RBC) into sickle cell which can be detected by use of automatic classifier. In this reported work, a novel robust hybrid segmentation method is introduced for segmenting normal RBCs and sickle cells. The hybrid segmentation process proposed here is a combination of two segmentation methods namely fuzzy C-means and adaptive (local) thresholding. Four different adaptive thresholding methods are experimented here in combination with fuzzy C-means. The features extracted for carrying out this analysis are major axis, minor axis, aspect ratio, area, perimeter, form factor (metric value), eccentricity and solidity of each of the cells present in the sample blood smear. These eight features are used for training and testing the classifiers. For classification, two supervised classifiers, i.e. Naïve Bayes classifier and K-nearest neighbor classifier have been demonstrated on a dataset of 10 image data samples and the evaluated results are compared for each of the hybrid combinations. The comparison shows that the proposed hybrid combination of fuzzy C-means and NICK's thresholding alongwith K-nearest neighbor classifier gives the best performance in terms of the considered performance parameters. This combination gives a significant classification accuracy of 98.87% with lowest computational time.

Deep learning-based segmentation of the lung in MR-images acquired by a stack-of-spirals trajectory at ultra-short echo-times

BackgroundFunctional lung MRI techniques are usually associated with time-consuming post-processing, where manual lung segmentation represents the most cumbersome part. The aim of this study was to investigate whether deep learning-based segmentation of lung images which were scanned by a fast UTE sequence exploiting the stack-of-spirals trajectory can provide sufficiently good accuracy for the calculation of functional parameters.MethodsIn this study, lung images were acquired in 20 patients suffering from cystic fibrosis (CF) and 33 healthy volunteers, by a fast UTE sequence with a stack-of-spirals trajectory and a minimum echo-time of 0.05 ms. A convolutional neural network was then trained for semantic lung segmentation using 17,713 2D coronal slices, each paired with a label obtained from manual segmentation. Subsequently, the network was applied to 4920 independent 2D test images and results were compared to a manual segmentation using the Sørensen–Dice similarity coefficient (DSC) and the Hausdorff distance (HD). Obtained lung volumes and fractional ventilation values calculated from both segmentations were compared using Pearson’s correlation coefficient and Bland Altman analysis.To investigate generalizability to patients outside the CF collective, in particular to those exhibiting larger consolidations inside the lung, the network was additionally applied to UTE images from four patients with pneumonia and one with lung cancer.ResultsThe overall DSC for lung tissue was 0.967 ± 0.076 (mean ± standard deviation) and HD was 4.1 ± 4.4 mm. Lung volumes derived from manual and deep learning based segmentations as well as values for fractional ventilation exhibited a high overall correlation (Pearson’s correlation coefficent = 0.99 and 1.00). For the additional cohort with unseen pathologies / consolidations, mean DSC was 0.930 ± 0.083, HD = 12.9 ± 16.2 mm and the mean difference in lung volume was 0.032 ± 0.048 L.ConclusionsDeep learning-based image segmentation in stack-of-spirals based lung MRI allows for accurate estimation of lung volumes and fractional ventilation values and promises to replace the time-consuming step of manual image segmentation in the future.

3D Hand Pose Estimation in Point Cloud Using 3D Convolutional Neural Network on Egocentric Datasets

3D hand pose estimation from egocentric vision is an important study in the construction of assistance systems and modeling of robot hand in robotics. In this paper, we propose a complete method for estimating 3D hand posefrom the complex scene data obtained from the egocentric sensor. In which we propose a simple yet highly efficient pre-processing step for hand segmentation. In the estimation process, we used the Hand PointNet (HPN), V2V-PoseNet(V2V), Point-to-Point Regression PointNet (PtoP) for finetuning to estimate the 3D hand pose from the collected data obtained from the egocentric sensor, such as CVRA, FPHA (First-Person Hand Action) datasets. HPN, V2V, PtoP are thedeep networks/Convolutional Neural Networks (CNNs) for estimating 3D hand pose that uses the point cloud data of the hand. We evaluate the estimation results using the preprocessing step and do not use the pre-processing step to see the effectiveness of the proposed method. The results show that 3D distance error is increased many times compared to estimates on the hand datasets are not obstructed (the hand data obtained from surveillance cameras, are viewed from top view, front view, sides view) such as MSRA, NYU, ICVL datasets. The results are quantified, analyzed, shown on the point cloud data of CVAR dataset and projected on the color image of FPHA dataset.

Medical Image Segmentation Of Wbc Using Improved Dual Threshold Method

segmentation of medical images is a critical method for the proper identification and diagnosis of diseases. Because of the distinction between them, the segmentation of the WBCs is an important issue.The automatic diagnosis of multiple white blood cell or leukocyte disorders plays a significant role. The suggested work is an Improved Dual-Threshold approach based on a variation of white blood cell (WBC) segmentation of different color spaces. The improved dual-threshold segmentation method consists of three stages: the pre-processing stage, the segmentation step of the threshold, and the post-procurement phase. For additional handling, two images are obtained: one is a contrast-stretched grey image and the other is an H part image from the altered colour space of YCbCr in the pre-processing step. The three main steps comprising the threshold segmentation are context separation, red blood cell extraction, and the finest threshold selection.The single-threshold method in RGB colour space is involved in the process of separation of Red Blood Cells (RBC), the H channel image is used in the step of background extraction. The Golden Section search method is used for the best threshold selection item. Finally, to eliminate the noise and the unfinished
 WBCs, median filtering and arithmetical morphology are used in the post-processing stage. The image data collection for Acute Lymphoblastic Leukaemia (ALL) is used to test and segment irregular cells from test data using an overall accuracy. Compared to the current segmentation outcome, the implemented segmentation approach shows better precision performance.

Gridding and Segmentation Method for DNA Microarray Images

This article mainly explores meshing and segmentation techniques for microarray image analysis. The term "grid" refers to dividing an image into subgrids of dots and then dividing them into point detection. Most of the existing methods depend on input parameters such as the number of rows / columns, the number of points in each row / column, the size of the subarrays, etc. This article proposes a fully automatic mesh generation algorithm. This can remove any initialized parameter without any manual intervention. In the segmentation step, clustering algorithms are used because they do not consider the size and shape of the spots, do not depend on the initial state of the pixels, and do not require post-processing. In this article, a new method is proposed to estimate the initial parameters (centroid and number of clusters) required by any clustering algorithm. Qualitative and quantitative analysis shows that the algorithm can perform grid processing on microarray images well, and improves the performance of the clustering algorithm.

Approach to hand posture recognition based on hand shape features for human–robot interaction

Hand segmentation is the initial step for hand posture recognition. To reduce the effect of variable illumination in hand segmentation step, a new CbCr-I component Gaussian mixture model (GMM) is proposed to detect the skin region. The hand region is selected as a region of interest from the image using the skin detection technique based on the presented CbCr-I component GMM and a new adaptive threshold. A new hand shape distribution feature described in polar coordinates is proposed to extract hand contour features to solve the false recognition problem in some shape-based methods and effectively recognize the hand posture in cases when different hand postures have the same number of outstretched fingers. A multiclass support vector machine classifier is utilized to recognize the hand posture. Experiments were carried out on our data set to verify the feasibility of the proposed method. The results showed the effectiveness of the proposed approach compared with other methods.

Automated segmentation and quantitative analysis of optic disc and fovea in fundus images

Fundus image is widely used diagnosis method and involves the retinal tissues which can be important biomarkers for diagnosing diseases. Many studies have proposed automatic algorithms to detect the optic disc (OD) and fovea. However, they showed some limitations. Although the precise regions of retinal tissues are clinically important, most of these studies focused on the localization not the segmentation. Also, they did not sufficiently prove the clinical effectiveness of the methods using quantitative analysis. Furthermore, many of them have researched about the single retinal tissue. To compensate for these limitations, this study proposed automated segmentation method for both of the OD and fovea. In this study, the dataset was acquired from the DRIVE and Drions databases, and additional ground truth dataset was obtained from an ophthalmologist. The original fundus image was preprocessed to remove noise and enhance contrast. And the retinal vessel was segmented to use for the OD and fovea segmentation. In the OD and fovea segmentation step, a region of interest was designated based on the image features to increase the segmentation accuracy. To segment the OD, the retinal vessel was removed and substituted based on the intensity value of the four nearest non-vessel pixels. Finally, the OD and fovea regions were segmented based on the image features including intensity, shape and size. The proposed method was evaluated by quantitative analysis using eight methods. As a result, the proposed method showed high segmentation performance for the OD and fovea with accuracy of 99.18 and 99.80 % on the DRIVE database.

ANN Classification and Modified Otsu Labeling on Retinal Blood Vessels

Background: Diagnosis of ophthalmologic and cardiovascular systems most often rely on the prerequisite step of segmentation of retinal blood vessels. Analysis of vascular structures in the retinal fundus images can aid in the early screening or detection of many ophthalmological diseases like glaucoma, diabetic retinopathy, vein occlusions, hemorrhages etc. In most cases, optic nerve gets damaged causing a blind spot. In this paper, a method of blood vessel segmentation using improved SOM (iSOM) and ANN classifier is presented. Methods: Morphological operations are carried out to enhance the input image. Clustering of pixels is done using improved Kohonen Self- Organizing Map (SOM) based on texture feature wherein a new node is introduced and new learning methodology is adopted using constrained weight updation. Finally, modified Otsu method is designed to label the output neuron class as vessel and non -vessel. Results: Segmentation is tested on public image sets, High Resolution Fundus (HRF) images and DRIONS-DB databases for Accuracy, Recall rate, Precision, F-Score, AUC and JC. The results achieve an appreciable level of accuracy (~97%) as compared to other similar methods of classification. The average time taken is less in estimating the neuron class and is about 12.1 sec per image when evaluated on Intel Core i5 CPU running at 2.30 GHz coupled with 4 GB RAM. The mean squared error for the segmented images is found to be in the range of 4-5%. Conclusion: Segmentation of retinal blood vessels based on artificial neural networks employing iSOM preserves the topology consuming less time for constrained weight updation achieving better results than SOM. A new model to detect vessels can be developed by concatenating iSOMs in parallel for multi class functions.

Evaluation of statistical and Haralick texture features for lymphoma histological images classification

ABSTRACT The investigation of different types of cancer can be performed by images classification with features extracted from specific regions identified by a segmentation step. Therefore, this study presents the evaluation of texture features extracted from neoplastic nuclei for the classification of lymphomas images. The neoplastic nuclei were segmented by steps of pre and post-processing and a thresholding. Statistical and Haralick’s features extracted from wavelet and ranklet transforms were evaluated with different classifiers. The use of the statistical metrics from the wavelet transform in association with the K-nearest neighbour classifier provided the best results in most of the two-class classifications.

Estimating individual‐level plant traits at scale

Functional ecology has increasingly focused on describing ecological communities based on their traits (measurable features affecting individuals' fitness and performance). Analyzing trait distributions within and among forests could significantly improve understanding of community composition and ecosystem function. Historically, data on trait distributions are generated by (1) collecting a small number of leaves from a small number of trees, which suffers from limited sampling but produces information at the fundamental ecological unit (the individual), or (2) using remote-sensing images to infer traits, producing information continuously across large regions, but as plots (containing multiple trees of different species) or pixels, not individuals. Remote-sensing methods that identify individual trees and estimate their traits would provide the benefits of both approaches, producing continuous large-scale data linked to biological individuals. We used data from the National Ecological Observatory Network (NEON) to develop a method to scale up functional traits from 160 trees to the millions of trees within the spatial extent of two NEON sites. The pipeline consists of three stages: (1) image segmentation, to identify individual trees and estimate structural traits; (2) an ensemble of models to infer leaf mass area (LMA), nitrogen, carbon, and phosphorus content using hyperspectral signatures, and DBH from allometry; and (3) predictions for segmented crowns for the full remote-sensing footprint at the NEON sites. The R2 values on held-out test data ranged from 0.41 to 0.75 on held-out test data. The ensemble approach performed better than single partial least-squares models. Carbon performed poorly compared to other traits (R2 of 0.41). The crown segmentation step contributed the most uncertainty in the pipeline, due to over-segmentation. The pipeline produced good estimates of DBH (R2 of 0.62 on held-out data). Trait predictions for crowns performed significantly better than comparable predictions on pixels, resulting in improvement of R2 on test data of between 0.07 and 0.26. We used the pipeline to produce individual-level trait data for ~5million individual crowns, covering a total extent of ~360km2 . This large data set allows testing ecological questions on landscape scales, revealing that foliar traits are correlated with structural traits and environmental conditions.

A deep cascaded segmentation of obstructive sleep apnea-relevant organs from sagittal spine MRI

PurposeThe main purpose of this work was to develop an efficient approach for segmentation of structures that are relevant for diagnosis and treatment of obstructive sleep apnea syndrome (OSAS), namely pharynx, tongue, and soft palate, from mid-sagittal magnetic resonance imaging (MR) data. This framework will be applied to big data acquired within an on-going epidemiological study from a general population.MethodsA deep cascaded framework for subsequent segmentation of pharynx, tongue, and soft palate is presented. The pharyngeal structure was segmented first, since the airway was clearly visible in the T1-weighted sequence. Thereafter, it was used as an anatomical landmark for tongue location. Finally, the soft palate region was extracted using segmented tongue and pharynx structures and used as input for a deep network. In each segmentation step, a UNet-like architecture was applied.ResultsThe result assessment was performed qualitatively by comparing the region boundaries obtained from the expert to the framework results and quantitatively using the standard Dice coefficient metric. Additionally, cross-validation was applied to ensure that the framework performance did not depend on the specific selection of the validation set. The average Dice coefficients on the test set were 0.89pm 0.03, 0.87pm 0.02, and 0.79pm 0.08 for tongue, pharynx, and soft palate tissues, respectively. The results were similar to other approaches and consistent with expert readings.ConclusionDue to high speed and efficiency, the framework will be applied for big epidemiological data with thousands of participants acquired within the Study of Health in Pomerania as well as other epidemiological studies to provide information on the anatomical structures and aspects that constitute important risk factors to the OSAS development.

Reducing the Impact of Confounding Factors on Skin Cancer Classification via Image Segmentation: Technical Model Study.

BackgroundStudies have shown that artificial intelligence achieves similar or better performance than dermatologists in specific dermoscopic image classification tasks. However, artificial intelligence is susceptible to the influence of confounding factors within images (eg, skin markings), which can lead to false diagnoses of cancerous skin lesions. Image segmentation can remove lesion-adjacent confounding factors but greatly change the image representation.ObjectiveThe aim of this study was to compare the performance of 2 image classification workflows where images were either segmented or left unprocessed before the subsequent training and evaluation of a binary skin lesion classifier.MethodsSeparate binary skin lesion classifiers (nevus vs melanoma) were trained and evaluated on segmented and unsegmented dermoscopic images. For a more informative result, separate classifiers were trained on 2 distinct training data sets (human against machine [HAM] and International Skin Imaging Collaboration [ISIC]). Each training run was repeated 5 times. The mean performance of the 5 runs was evaluated on a multi-source test set (n=688) consisting of a holdout and an external component.ResultsOur findings showed that when trained on HAM, the segmented classifiers showed a higher overall balanced accuracy (75.6% [SD 1.1%]) than the unsegmented classifiers (66.7% [SD 3.2%]), which was significant in 4 out of 5 runs (P<.001). The overall balanced accuracy was numerically higher for the unsegmented ISIC classifiers (78.3% [SD 1.8%]) than for the segmented ISIC classifiers (77.4% [SD 1.5%]), which was significantly different in 1 out of 5 runs (P=.004).ConclusionsImage segmentation does not result in overall performance decrease but it causes the beneficial removal of lesion-adjacent confounding factors. Thus, it is a viable option to address the negative impact that confounding factors have on deep learning models in dermatology. However, the segmentation step might introduce new pitfalls, which require further investigations.

Overview of MR Image Segmentation Strategies in Neuromuscular Disorders.

Neuromuscular disorders are rare diseases for which few therapeutic strategies currently exist. Assessment of therapeutic strategies efficiency is limited by the lack of biomarkers sensitive to the slow progression of neuromuscular diseases (NMD). Magnetic resonance imaging (MRI) has emerged as a tool of choice for the development of qualitative scores for the study of NMD. The recent emergence of quantitative MRI has enabled to provide quantitative biomarkers more sensitive to the evaluation of pathological changes in muscle tissue. However, in order to extract these biomarkers from specific regions of interest, muscle segmentation is mandatory. The time-consuming aspect of manual segmentation has limited the evaluation of these biomarkers on large cohorts. In recent years, several methods have been proposed to make the segmentation step automatic or semi-automatic. The purpose of this study was to review these methods and discuss their reliability, reproducibility, and limitations in the context of NMD. A particular attention has been paid to recent deep learning methods, as they have emerged as an effective method of image segmentation in many other clinical contexts.

Fast-track virtual reality for cardiac imaging in congenital heart disease.

We sought to evaluate the appropriateness of cardiac anatomy renderings by a new virtual reality (VR) technology, entitled DIVA, directly applicable to raw magnetic resonance imaging (MRI) data without intermediate segmentation steps in comparison to standard three-dimensional (3D) rendering techniques (3D PDF and 3D printing). Differences in post-processing times were also evaluated. We reconstructed 3D (STL, 3D-PDF, and 3D printed ones) and VR models of three patients with different types of complex congenital heart disease (CHD). We then asked a senior pediatric heart surgeon to compare and grade the results obtained. All anatomical structures were well visualized in both VR and 3D PDF/printed models. Ventricular-arterial connections and their relationship with the great vessels were better visualized with the VR model (Case 2); aortic arch anatomy and details were also better visualized by the VR model (Case 3). The median post-processing time to get VR models using DIVA was 5 min in comparison to 8 h (range 8-12 h including printing time) for 3D models (PDF/printed). VR directly applied to non-segmented 3D-MRI data set is a promising technique for 3D advanced modeling in CHD. It is systematically more consistent and faster when compared to standard 3D-modeling techniques.

Segmented School Physical Activity and Weight Status in Children: Application of Compositional Data Analysis.

The purpose of this study was to apply compositional data analysis (CoDA) for the analysis of segmented school step counts and associate the school step count composition to body mass index (BMI) z-scores in a sample of children. Participants were 855 (51.8% female) children recruited from the fourth and fifth grades from four schools following a 7-h school schedule. Using piezoelectric pedometers, step count data were collected during physical education, recess, lunch, and during academic class time. A multi-level mixed effects model associated the step count composition with BMI z-scores. Compositional isotemporal substitution determined changes in BMI z-scores per reallocation of steps between pairs of school segments. A higher percentage of steps accrued during physical education (b = −0.34, 95%CI: −0.65–−0.03, p = 0.036) and recess (b = −0.47, 95%CI: −0.83–−0.11, p = 0.012), relative to other segments, was associated with lower BMI z-scores. Specifically, a 5% to 15% reallocation of steps accrued during lunchtime to either physical education or recess was associated with lower BMI z-scores, ranging from −0.07 to −0.25 standard deviation units. Focusing school-based promotion of physical activity during physical education and recess may have greater relative importance if targeted outcomes are weight-related.

Massive picture retrieval system based on big data image mining

The traditional picture retrieval system has a slow retrieval speed, poor retrieval accuracy, and a low recall when performing massive picture retrieval. In this paper, we design a massive picture retrieval system using the big data image mining technology. It is constructed with data processing layer, business logic layer and presentation layer and works through three steps of data segmentation, mining and merging. For instance, it runs the distributed file system module in a Master/Slave operation mode and designs file read and write requests according to user interaction. Next, it performs parallel computing of picture data sets based on Map Reduce module to solve the picture matching and similarity metrics and returns to the user sorted picture matching result. Then, it extracts the color and texture features of the target area to generate the final picture retrieval result. We select a large number of pictures on a big data platform as simulation test set. The results show that the system we designed has a good retrieval accuracy and a high retrieval speed, which greatly improves the recall of picture retrieval.

From a Point Cloud to a Simulation Model-Bayesian Segmentation and Entropy Based Uncertainty Estimation for 3D Modelling.

The 3D modelling of indoor environments and the generation of process simulations play an important role in factory and assembly planning. In brownfield planning cases, existing data are often outdated and incomplete especially for older plants, which were mostly planned in 2D. Thus, current environment models cannot be generated directly on the basis of existing data and a holistic approach on how to build such a factory model in a highly automated fashion is mostly non-existent. Major steps in generating an environment model of a production plant include data collection, data pre-processing and object identification as well as pose estimation. In this work, we elaborate on a methodical modelling approach, which starts with the digitalization of large-scale indoor environments and ends with the generation of a static environment or simulation model. The object identification step is realized using a Bayesian neural network capable of point cloud segmentation. We elaborate on the impact of the uncertainty information estimated by a Bayesian segmentation framework on the accuracy of the generated environment model. The steps of data collection and point cloud segmentation as well as the resulting model accuracy are evaluated on a real-world data set collected at the assembly line of a large-scale automotive production plant. The Bayesian segmentation network clearly surpasses the performance of the frequentist baseline and allows us to considerably increase the accuracy of the model placement in a simulation scene.

Abstract PO-084: Automated detection of pancreatic ductal adenocarcinoma (PDAC) on CT scans using artificial intelligence (AI): Impact of inclusion of automated pancreas segmentation on the accuracy of 3D-convolutional neural network (CNN)

Abstract Purpose: Around 30% of PDAC less than 2-cm tend to go undetected on CT due to their subtle imaging signatures. Automated detection of PDAC using AI represents an opportunity to augment physician expertise and to improve outcomes through early detection of PDAC. Our purpose was to develop a 3D-CNN for fully automated detection of PDAC and to further evaluate the impact of inclusion of pancreas segmentation on the accuracy of this 3D-CNN. Methods: A Medical Imaging Data Readiness Scale (MIDaR) level A dataset (portal venous phase CTs, slice thickness ≤ 3.75 mm) of 466 treatment-naïve biopsy-proven PDAC and 1994 subjects with normal pancreas was created after exclusion of CTs with suboptimal image quality or biliary stents. Volumetric pancreas and tumor segmentations on CTs were done by two radiologists using 3D Slicer. A total of 370 CTs with PDAC and 370 CTs with normal pancreas were randomly selected for separate training and validation sets, and 396 CTs (96 CTs with PDAC and 300 CTs with normal pancreas) were utilized for testing. Two separate 3D-CNNs were trained. A three-stage bounding-box-only model (A): stage 1 was based on a UNET-like architecture and localized the pancreas on CT with a bounding box; stage 2 utilized an Inception ResNet architecture and classified each slice through the pancreas into PDAC vs. normal; and stage 3 utilized the output of stage 2 to generate final classification for a given CT. Conversely, a four-stage pancreas segmentation-based model (B) included stage 1 of model A followed by an additional stage of automated pancreas and tumor segmentation (stage 2), classification of each slice through the pancreas into PDAC vs. normal (stage 3) and, finally, generation of final classification score (stage 4) for a given CT. Area under the receiver operating characteristic curve (AUROC) of the two models were compared on the test set. Results: Mean (SD) PDAC diameter in the test set was 1.1 (0.43) cm. Model A (three-stage bounding-box-only) correctly classified 305 (77%) out of 396 CTs from the test set into PDAC vs. normal. It incorrectly classified 12/96 (12.5%) CTs with PDAC as normal and 79/300 (26%) normal CTs as PDAC. AUROC for model A was 0.85. Model B (four-stage pancreas segmentation-based) correctly classified 351 (88%) out of 396 CTs. It incorrectly classified 13/96 (13.5%) CTs with PDAC as normal and 32/300 (10.7%) normal CTs as PDAC. AUROC for model B was 0.94. AUROC for model B was significantly higher than model A (p&amp;lt;0.005). Conclusion: A 3D-CNN can detect small PDAC with high accuracy using automated localization of pancreas with a bounding box without relying on separate pancreas segmentation. Inclusion of an additional automated pancreas segmentation step reduced false positives with consequent incremental gain in the model’s accuracy. Prospective validation and subsequent integration of such models into clinical workflows has the potential to reduce inadvertent errors in detection of subtle or small PDAC on standard-of-care CT scans. Citation Format: Anurima Patra, Korfiatis Panagiotis, Garima Suman, Ananya Panda, Sushil Kumar Garg, Ajit Goenka. Automated detection of pancreatic ductal adenocarcinoma (PDAC) on CT scans using artificial intelligence (AI): Impact of inclusion of automated pancreas segmentation on the accuracy of 3D-convolutional neural network (CNN) [abstract]. In: Proceedings of the AACR Virtual Special Conference on Artificial Intelligence, Diagnosis, and Imaging; 2021 Jan 13-14. Philadelphia (PA): AACR; Clin Cancer Res 2021;27(5_Suppl):Abstract nr PO-084.

The Aneurysm Occlusion Assistant, an AI platform for real time surgical guidance of intracranial aneurysms.

In recent years, endovascular treatment has become the dominant approach to treat intracranial aneurysms (IAs). Despite tremendous improvement in surgical devices and techniques, 10-30% of these surgeries require retreatment. Previously, we developed a method which combines quantitative angiography with data-driven modeling to predict aneurysm occlusion within a fraction of a second. This is the first report on a semi-autonomous system, which can predict the surgical outcome of an IA immediately following device placement, allowing for therapy adjustment. Additionally, we previously reported various algorithms which can segment IAs, extract hemodynamic parameters via angiographic parametric imaging, and perform occlusion predictions. We integrated these features into an Aneurysm Occlusion Assistant (AnOA) utilizing the Kivy library's graphical instructions and unique language properties for interface development, while the machine learning algorithms were entirely developed within Keras, Tensorflow and skLearn. The interface requires pre- and post-device placement angiographic data. The next steps for aneurysm segmentation, angiographic analysis and prediction have been integrated allowing either autonomous or interactive use. The interface allows for segmentation of IAs and cranial vasculature with a dice index of ~0.78 and prediction of aneurysm occlusion at six months with an accuracy 0.84, in 6.88 seconds. This is the first report on the AnOA to guide endovascular treatment of IAs. While this initial report is on a stand-alone platform, the software can be integrated in the angiographic suite allowing direct communication with the angiographic system for a completely autonomous surgical guidance solution.