Semi-automatic Segmentation Tool Research Articles

Accelerated muscle mass estimation from CT images through transfer learning

BackgroundThe cost of labeling to collect training data sets using deep learning is especially high in medical applications compared to other fields. Furthermore, due to variances in images depending on the computed tomography (CT) devices, a deep learning based segmentation model trained with a certain device often does not work with images from a different device.MethodsIn this study, we propose an efficient learning strategy for deep learning models in medical image segmentation. We aim to overcome the difficulties of segmentation in CT images by training a VNet segmentation model which enables rapid labeling of organs in CT images with the model obtained by transfer learning using a small number of manually labeled images, called SEED images. We established a process for generating SEED images and conducting transfer learning a model. We evaluate the performance of various segmentation models such as vanilla UNet, UNETR, Swin-UNETR and VNet. Furthermore, assuming a scenario that a model is repeatedly trained with CT images collected from multiple devices, in which is catastrophic forgetting often occurs, we examine if the performance of our model degrades.ResultsWe show that transfer learning can train a model that does a good job of segmenting muscles with a small number of images. In addition, it was confirmed that VNet shows better performance when comparing the performance of existing semi-automated segmentation tools and other deep learning networks to muscle and liver segmentation tasks. Additionally, we confirmed that VNet is the most robust model to deal with catastrophic forgetting problems.ConclusionIn the 2D CT image segmentation task, we confirmed that the CNN-based network shows better performance than the existing semi-automatic segmentation tool or latest transformer-based networks.

Distribution of MRI-derived T2 values as a biomarker for in vivo rapid screening of phenotype severity in mdx mice

Background The pathology in Duchenne muscular dystrophy (DMD) is characterized by degenerating muscle fibers, inflammation, fibro-fatty infiltrate, and edema, and these pathological processes replace normal healthy muscle tissue. The mdx mouse model is one of the most commonly used preclinical models to study DMD. Mounting evidence has emerged illustrating that muscle disease progression varies considerably in mdx mice, with inter-animal differences as well as intra-muscular differences in pathology in individual mdx mice. This variation is important to consider when conducting assessments of drug efficacy and in longitudinal studies. We developed a magnetic resonance imaging (MRI) segmentation and analysis pipeline to rapidly and non-invasively measure the severity of muscle disease in mdx mice. Methods Wildtype and mdx mice were imaged with MRI and T2 maps were obtained axially across the hindlimbs. A neural network was trained to rapidly and semi-automatically segment the muscle tissue, and the distribution of resulting T2 values was analyzed. Interdecile range and Pearson Skew were identified as biomarkers to quickly and accurately estimate muscle disease severity in mice. Results The semiautomated segmentation tool reduced image processing time approximately tenfold. Measures of Pearson skew and interdecile range based on that segmentation were repeatable and reflected muscle disease severity in healthy wildtype and diseased mdx mice based on both qualitative observation of images and correlation with Evans blue dye uptake. Conclusion Use of this rapid, non-invasive, semi-automated MR image segmentation and analysis pipeline has the potential to transform preclinical studies, allowing for pre-screening of dystrophic mice prior to study enrollment to ensure more uniform muscle disease pathology across treatment groups, improving study outcomes.

Improving Reproducibility of Volumetric Evaluation Using Computed Tomography in Pediatric Patients with Congenital Heart Disease.

The volumetric data obtained from the cardiac CT scan of congenital heart disease patients is important for defining patient's status and making decision for proper management. The objective of this study is to evaluate the intra-observer, inter-observer, and interstudy reproducibility of left ventricular (LV) and right ventricular (RV) or functional single-ventricle (FSV) volume. And compared those between manual and using semi-automated segmentation tool. Total of 127 patients (56 female, 71 male; mean age 82.1months) underwent pediatric protocol cardiac CT from January 2020 to December 2022. The volumetric data including both end-systolic and -diastolic volume and calculated EF were derived from both conventional semiautomatic region growing algorithms (CM, TeraRecon, TeraRecon, Inc., San Mateo, CA, USA) and deep learning-based annotation program (DLS, Medilabel, Ingradient, Inc., Seoul, Republic of Korea) by three readers, who have different background knowledge or experience of radiology or image extraction before. The reproducibility was compared using intra- and inter-observer agreements. And the usability was measured using time for reconstruction and number of tests that were reconfigured before the reconfiguration time was reduced to less than 5min. Inter- and intra-observer agreements showed better agreements degrees in DLS than CM in all analyzers. The time used for reconstruction showed significantly shorter in DLS compared with CM. And significantly small numbers of tests before the reconfiguration is needed in DLS than CM. Deep learning-based annotation program can be more accurate way for measurement of volumetric data for congenital heart disease patients with better reproducibility than conventional method.

Digital twinning of cardiac electrophysiology for congenital heart disease.

In recent years, blending mechanistic knowledge with machine learning has had a major impact in digital healthcare. In this work, we introduce a computational pipeline to build certified digital replicas of cardiac electrophysiology in paediatric patients with congenital heart disease. We construct the patient-specific geometry by means of semi-automatic segmentation and meshing tools. We generate a dataset of electrophysiology simulations covering cell-to-organ level model parameters and using rigorous mathematical models based on differential equations. We previously proposed Branched Latent Neural Maps (BLNMs) as an accurate and efficient means to recapitulate complex physical processes in a neural network. Here, we employ BLNMs to encode the parametrized temporal dynamics of in silico 12-lead electrocardiograms (ECGs). BLNMs act as a geometry-specific surrogate model of cardiac function for fast and robust parameter estimation to match clinical ECGs in paediatric patients. Identifiability and trustworthiness of calibrated model parameters are assessed by sensitivity analysis and uncertainty quantification.

ViNe-Seg: deep-learning-assisted segmentation of visible neurons and subsequent analysis embedded in a graphical user interface.

Segmentation of neural somata is a crucial and usually the most time-consuming step in the analysis of optical functional imaging of neuronal microcircuits. In recent years, multiple auto-segmentation tools have been developed to improve the speed and consistency of the segmentation process, mostly, using deep learning approaches. Current segmentation tools, while advanced, still encounter challenges in producing accurate segmentation results, especially in datasets with a low signal-to-noise ratio. This has led to a reliance on manual segmentation techniques. However, manual methods, while customized to specific laboratory protocols, can introduce variability due to individual differences in interpretation, potentially affecting dataset consistency across studies. In response to this challenge, we present ViNe-Seg: a deep-learning-based semi-automatic segmentation tool that offers (i) detection of visible neurons, irrespective of their activity status; (ii) the ability to perform segmentation during an ongoing experiment; (iii) a user-friendly graphical interface that facilitates expert supervision, ensuring precise identification of Regions of Interest; (iv) an array of segmentation models with the option of training custom models and sharing them with the community; and (v) seamless integration of subsequent analysis steps. ViNe-Seg code and documentation are publicly available at https://github.com/NiRuff/ViNe-Seg and can be installed from https://pypi.org/project/ViNeSeg/.

Volumetric Parameters Derived from CXCR4-Directed PET/CT Predict Outcome in Patients with Gastrointestinal Neuroendocrine Carcinomas

BackgroundGastro-entero-pancreatic neuroendocrine carcinomas (GEP-NECs) are an aggressive subgroup of neuroendocrine neoplasms (NENs). In patients affected with NEN, there is a growing body of evidence that increased C-X-C motif chemokine receptor (CXCR4) expression is linked to decreasing overall survival (OS) in an ex-vivo setting. Thus, we aimed to determine whether the in-vivo-derived CXCR4-directed whole-body PET signal can also determine GEP-NEC patients with shorter OS.MethodsWe retrospectively included 16 patients with histologically proven GEP-NEC, who underwent CXCR4-directed PET/CT for staging and therapy planning. We assessed maximum, peak, and mean standardized uptake values as well as whole-body tumor volume (TV) and total-lesion uptake (TLU = SUVmean × TV) using a semi-automatic segmentation tool with a 50% threshold. Association of PET-based biomarkers and OS or radiographic progression-free survival (rPFS; according to RECIST 1.1 criteria) was analyzed using univariable and multivariable cox regression.ResultsMedian OS and rPFS was 7.5 and 7 months, respectively. A significant correlation between TV and TLU was found for OS (TV: hazard ratio (HR) 1.007 95% confidence interval (CI) 1.000–1.014, p = 0.0309; TLU: HR 1.002 95% CI 1.000–1.003, p = 0.0350) and rPFS (TV: HR 1.010 95% CI 1.002–1.021; p = 0.0275; TLU: HR 1.002 95% CI 1.000–1.004, p = 0.0329), respectively. No significant correlation with OS or rPFS was found for non-volumetric parameters (p > 0.4). TV remained a significant predictive marker for OS and rPFS in multivariable analysis (OS: HR 1.012 95%, CI 1.003–1.022, p = 0.0084; rPFS: HR 1.009, 95% CI 0.9999–1.019, p = 0.0491), whereas TLU remained only prognostic for OS (HR 1.009, 95% CI 0.9999–1.019, p = 0.0194) but narrowly failed significance for rPFS (p = 0.0559).ConclusionIn-vivo assessment of CXCR4 PET-derived volumetric parameters is predictive for outcome of patients with GEP-NEC and could be used as a risk stratification tool, which detects patients prone to early progression.

P067 Volumetric Assessment of Treatment Response in Newly Diagnosed Crohn Disease: A Pilot Study

Abstract Background Traditionally, mural thickness and disease length have been the radiological standards for assessing Crohn's Disease (CD) severity and treatment response. Our study aimed to evaluate and compare these linear measurements with volumetric ones in newly diagnosed pediatric CD, as well as to assess the concordance between manual and semi-automated volumetric methods. Methods In this single-center, prospective study, 20 children with newly diagnosed ileal CD treated with anti-TNF therapy from Dec 2018 to Oct 2021 (8 females; median age=14.5 years) underwent dedicated research MRI examinations at three timepoints – diagnosis (baseline) as well as 6 weeks and 6 months into therapy. For each exam and using T2-weighted Single Shot Fast Spin Echo (SSFSE) images, a single radiologist measured maximum bowel wall thickness (BWT, mm) and length of disease (cm). The inflamed ileum was manually segmented to measure tissue volume (ml) (Entrolytics; Motilent, London, UK). Semi-automated 3D volumetric measurements of the inflamed ileum (GI Seg tool) were created using the same platform. Estimated maximum BWT also was automatically calculated from manually-edited algorithmically generated masks. Mixed-effects models evaluated measurement changes over time. Intra-class correlation assessed absolute agreement between manual and automated volumetric assessments. Results BWT decreased over time with treatment (0.76 vs. 0.58 vs. 0.45 cm; p<0.0001), with a 24% decrease from baseline to 6 weeks and a 41% decrease from baseline to 6 months. Similarly, there was a significant decrease in length of disease over time (19.2 vs. 12.2 vs. 8.0 cm; p=0.002), with a 36% decrease from baseline to 6 weeks and a 58% decrease from baseline to 6 months. Manual volumetric measurements also showed a significant treatment response (19.8 vs. 11.6 vs. 5.1 ml; p<0.0001), with a 41% decrease from baseline to 6 weeks and a 74% decrease from baseline to 6 months. Using the semi-automated segmentation tool, the volume of inflamed bowel also decreased significantly over time (24.0 vs. 15.1 vs. 9.1 ml; p=0.0007), decreasing 37% between baseline and 6 weeks and 62% between baseline and 6 months. There was good agreement between manual and semi-automated volumetric bowel assessments (ICC=0.78 [95% CI: 0.57-0.88]). ICC between manual and code-estimated Max BWT was 0.60 [95% CI: 0.42, 0.72] with a strong positive correlation (r=0.839; p <0.001) between these two methods as shown in the Figure. Conclusion Volumetric assessments, along with automatic measurements of bowel wall thickness, show promise in accurately assessing treatment response in pediatric CD. These methods might provide more consistent and repeatable measurements, essential for long-term disease monitoring and management.

CT Based Tumor Radiomics with Machine Learning Classifiers for Molecular Subtyping of Diffuse Large B Cell Lymphoma.

Diffuse large B-cell lymphoma (DLBL) has been classified into two distinct molecular subtypes based on their cell of origin- namely germinal center B-cell-like (GCB) and activated B-cell-subtype (ABC) based on Han's algorithm. These two entities are also prognostically distinct. We assessed CT based tumor radiomic features using machine learning classifiers to distinguish the two molecular subtypes. One hundred one patients were accrued in the study after institutional ethics committee approval. GCB subtype was diagnosed in 59 while 42 patients were ABC subtype. Lesion with maximum SUV was delineated using a semi-automated segmentation tool (3D slicer) and radiomic features were extracted using Pyradiomics. A total of 1030 features were extracted including Shape features, first, second, third, higher order and wavelet-based features. Recursive feature elimination was used for feature selection and optimal 10 features were selected. Five machine learning (ML) tools were used to build a model to sub- classify the two molecular subtypes. (Table -1) Model performance was assessed using Accuracy, Precision, Recall, F1 score and Area under curve of receiver operating characteristic curve. The models were constructed by two methods.1) A 5-fold internal cross validation strategy & 2) splitting the data into training (70%) and validation set (30%). Recursive feature elimination using a random forest algorithm method was used to select 7 radiomic features. These 7 features were then used to build the model using the 5 ML classifiers. Of the five classifiers, Random Forest Classifier (RFC) using a 70:30 training: test strategy was the best performer among all the ML models, with the highest accuracy of 80%, AUC of 0.87, recall(sensitivity), f1score(specificity) & precision (Positive Predictive value; PPV) or precision of 82% respectively. When an internal 5-fold cross validation strategy was used, RFC again performed better than other ML classifiers with an accuracy of 75%, AUC of 0.80, recall of 79%, PPV of 77% and specificity of 79%. The machine-learning based Radiomics features extracted from pre-treatment CT images can provide a simple and non-invasive method for the prediction of DLBL molecular subtypes with favorable predictive accuracy. The Random Forest Classifier was the most accurate in distinguishing GCB subtype from the ABC subtype.

Improving radiomics reproducibility using deep learning-based image conversion of CT reconstruction algorithms in hepatocellular carcinoma patients.

CT reconstruction algorithms affect radiomics reproducibility. In this study, we evaluate the effect of deep learning-based image conversion on CT reconstruction algorithms. This study included 78 hepatocellular carcinoma (HCC) patients who underwent four-phase liver CTs comprising non-contrast, late arterial (LAP), portal venous (PVP), and delayed phase (DP), reconstructed using both filtered back projection (FBP) and advanced modeled iterative reconstruction (ADMIRE). PVP images were used to train a convolutional neural network (CNN) model to convert images from FBP to ADMIRE and vice versa. LAP, PVP, and DP images were used for validation and testing. Radiomic features were extracted for each patient with a semi-automatic segmentation tool. We used concordance correlation coefficients (CCCs) to evaluate the radiomics reproducibility for original FBP (oFBP) vs. original ADMIRE (oADMIRE), oFBP vs. converted FBP (cFBP), and oADMIRE vs. converted ADMIRE (cADMIRE). In the test group including 30 patients, the CCC and proportion of reproducible features (CCC ≥ 0.85) for oFBP vs. oADMIRE were 0.65 and 32.9% (524/1595) for LAP, 0.65 and 35.9% (573/1595) for PVP, and 0.69 and 43.8% (699/1595) for DP. For oFBP vs. cFBP, the values increased to 0.92 and 83.9% (1339/1595) for LAP, 0.89 and 71.0% (1133/1595) for PVP, and 0.90 and 79.7% (1271/1595) for DP. Similarly, for oADMIRE vs. cADMIRE, the values increased to 0.87 and 68.1% (1086/1595) for LAP, 0.91 and 82.1% (1309/1595) for PVP, and 0.89 and 76.2% (1216/1595) for DP. CNN-based image conversion between CT reconstruction algorithms improved the radiomics reproducibility of HCCs. This study demonstrates that using a CNN-based image conversion technique significantly improves the reproducibility of radiomic features in HCCs, highlighting its potential for enhancing radiomics research in HCC patients. Radiomics reproducibility of HCC was improved via CNN-based image conversion between two different CT reconstruction algorithms. This is the first clinical study to demonstrate improvements across a range of radiomic features in HCC patients. This study promotes the reproducibility and generalizability of different CT reconstruction algorithms in radiomics research.

A computer-aided diagnosis system for brain tumors based on artificial intelligence algorithms.

The choice of treatment and prognosis evaluation depend on the accurate early diagnosis of brain tumors. Many brain tumors go undiagnosed or are overlooked by clinicians as a result of the challenges associated with manually evaluating magnetic resonance imaging (MRI) images in clinical practice. In this study, we built a computer-aided diagnosis (CAD) system for glioma detection, grading, segmentation, and knowledge discovery based on artificial intelligence algorithms. Neuroimages are specifically represented using a type of visual feature known as the histogram of gradients (HOG). Then, through a two-level classification framework, the HOG features are employed to distinguish between healthy controls and patients, or between different glioma grades. This CAD system also offers tumor visualization using a semi-automatic segmentation tool for better patient management and treatment monitoring. Finally, a knowledge base is created to offer additional advice for the diagnosis of brain tumors. Based on our proposed two-level classification framework, we train models for glioma detection and grading, achieving area under curve (AUC) of 0.921 and 0.806, respectively. Different from other systems, we integrate these diagnostic tools with a web-based interface, which provides the flexibility for system deployment.

Selective ensemble methods for deep learning segmentation of major vessels in invasive coronary angiography.

Invasive coronary angiography (ICA) is a primary imaging modality that visualizes the lumen area of coronary arteries for diagnosis and interventional guidance. In the current practice of quantitative coronary analysis (QCA), semi-automatic segmentation tools require labor-intensive and time-consuming manual correction, limiting their application in the catheterization room. This study aims to propose rank-based selective ensemble methods that improve the segmentation performance and reduce morphological errors that limit fully automated quantification of coronary artery using deep-learning segmentation of ICA. Two selective ensemble methods proposed in this work integrated the weighted ensemble approach with per-image quality estimation. The segmentation outcomes from five base models with different loss functions were ranked either by mask morphology or estimated dice similarity coefficient (DSC). The final output was determined by imposing different weights according to the ranks. The ranking criteria based on mask morphology were formulated from empirical insight to avoid frequent types of segmentation errors (MSEN), while the estimation of DSCs was performed by comparing the pseudo-ground truth generated from a meta-learner (ESEN). Five-fold cross-validation was performed with the internal dataset of 7426 coronary angiograms from 2924 patients, and prediction model was externally validated with 556 images of 226 patients. The selective ensemble methods improved the segmentation performance with DSCs up to 93.07% and provided a better delineation of coronary lesion with local DSCs of up to 93.93%, outperforming all individual models. Proposed methods also minimized the chances of mask disconnection in the most narrowed regions to 2.10%. The robustness of the proposed methods was also evident in the external validation. Inference time for major vessel segmentation was approximately one-sixth of a second. Proposed methods successfully reduced morphological errors in the predicted masks and were able to enhance the robustness of the automatic segmentation. The results suggest better applicability of real-time QCA-based diagnostic methods in routine clinical settings.

The Infrared Thermography Toolbox: An Open-access Semi-automated Segmentation Tool for Extracting Skin Temperatures in the Thoracic Region including Supraclavicular Brown Adipose Tissue

Infrared thermography (IRT) is widely used to assess skin temperature in response to physiological changes. Yet, it remains challenging to standardize skin temperature measurements over repeated datasets. We developed an open-access semi-automated segmentation tool (the IRT-toolbox) for measuring skin temperatures in the thoracic area to estimate supraclavicular brown adipose tissue (scBAT) activity, and compared it to manual segmentations. The IRT-toolbox, designed in Python, consisted of image pre-alignment and non-rigid image registration. The toolbox was tested using datasets of 10 individuals (BMI = 22.1 ± 2.1 kg/m2, age = 22.0 ± 3.7 years) who underwent two cooling procedures, yielding four images per individual. Regions of interest (ROIs) were delineated by two raters in the scBAT and deltoid areas on baseline images. The toolbox enabled direct transfer of baseline ROIs to the registered follow-up images. For comparison, both raters also manually drew ROIs in all follow-up images. Spatial ROI overlap between methods and raters was determined using the Dice coefficient. Mean bias and 95% limits of agreement in mean skin temperature between methods and raters were assessed using Bland–Altman analyses. ROI delineation time was four times faster with the IRT-toolbox (01:04 min) than with manual delineations (04:12 min). In both anatomical areas, there was a large variability in ROI placement between methods. Yet, relatively small skin temperature differences were found between methods (scBAT: 0.10 °C, 95%LoA[-0.13 to 0.33 °C] and deltoid: 0.05 °C, 95%LoA[-0.46 to 0.55 °C]). The variability in skin temperature between raters was comparable between methods. The IRT-toolbox enables faster ROI delineations, while maintaining inter-user reliability compared to manual delineations. (Trial registration number (ClinicalTrials.gov): NCT04406922, [May 29, 2020]).

Pretreatment PET Metrics and Clinical Outcomes of Anal Cancer in Patients Living with and without HIV

<h3>Purpose/Objective(s)</h3> This study investigates the association between pre-treatment positron emission tomography (PET) metrics and clinical outcomes in anal squamous cell carcinoma (ASCC) patients treated with definitive radiation therapy (RT). We hypothesize that pre-treatment PET metrics may differentially predict clinical outcomes in ASCC patients without HIV infection (HIV-) and ASCC patients living with HIV (PLWH) with detectable or undetectable viral load (VL). <h3>Materials/Methods</h3> Patients treated with definitive RT for non-metastatic ASCC between 2005 - 2021 at a single institution were retrospectively identified, and clinical and demographic data, including HIV status and pretreatment VL, were tabulated. Pre-treatment PETs were imported to MIM, and semiautomatic gradient-based segmentation tool algorithms (PET_Edge) were used to calculate PET metrics, including maximum standardized uptake value (SUVmax), metabolic tumor volume (MTV), and total lesion glycolysis (TLG). Cox proportional hazard modeling was used to investigate the association between primary tumor PET metrics and clinical outcomes, including overall survival (OS), progression free survival (PFS), and loco-regional control (LRC) <h3>Results</h3> 141 ASCC patients were identified, 49 of whom were PLWH. 12 had detectable HIV VL. The median follow-up time was 51.8 months (IQR: 0.5-191.2 months). There were 22 loco-regional failures and 24 deaths. 3- and 5- year LRC was 85.6% and 83.3%, 3- and 5- year PFS was 75.9% and 71.6%, and 3- and 5- year OS was 89.3% and 86.6% for the entire cohort. There was no significant difference in LRC, PFS, or OS between PLWH and HIV- patients (p>0.50). PLWH with detectable VL had similar LRC and PFS but trended towards worse 5-year OS than did HIV- patients (66.3% vs 89.7%, p=0.16). MTV and TLG from primary tumors were significantly associated with LRC, PFS, and OS for all patients and in sub-group analysis for HIV-, PLWH, and PLWH with undetectable VL (p<0.05). SUVmax was significantly associated with LRC, PFS, and OS for all patients, and on sub-group analysis only for HIV- patients (p<0.05). On multivariable analysis, after adjusting for stage and Charlson Comorbidity Index, MTV split at median was significantly associated with LRC (HR 3.59 (95% CI:1.21-10.6), p=0.02) and PFS (HR 3.12 (95%CI:1.48-6.56), p=0.003), and TLG split at median was significantly associated with LRC (HR 4.36 (95%CI:1.36-14.0), p=0.01) and PFS (HR 3.13 (95%CI:1.48-6.64), p=0.003). <h3>Conclusion</h3> Patients with ASCC had similar outcomes regardless of HIV status. Advanced pre-treatment PET metrics, especially MTV and TLG, may have prognostic or risk-stratification utility in HIV- patients and undetectable PLWH with ASCC, but were not as strongly associated with clinical outcomes for PLWH with detectable VL, perhaps due to small numbers of patients with detectable VL in this cohort.

A Multi-Layer Auto-Segmentation Quality Assurance and Correction Pipeline for MR-Guided Adaptive Radiotherapy

<h3>Purpose/Objective(s)</h3> Fast and accurate auto-segmentation on daily image is essential for MR-guided adaptive radiotherapy (MRgART). However, the state-of-the-art deep learning auto-segmentation (DLAS) still has limited success particularly for complex structures such as those in the abdomen. Inaccurate contours from DLAS need to be inspected and edited and this manual process can be labor intensive, time consuming and subject to human errors. This work aims to develop a multi-layer pipeline that includes an automatic contour quality assurance (ACQA) test, an automatic contour refinement (ACR) process, and an interactive semi-automatic segmentation (ISS) tool to quickly identify and correct for inaccurate auto-segmented contours, thus replacing the manual process in MRgART. <h3>Materials/Methods</h3> The proposed pipeline includes the ACQA, ACR and ISS modules that are developed to detect and correct for inaccurate auto-segmented contours in the following steps: (1) identifying contours/image slices of acceptable or with minor or major errors from the input auto-segmented contours using ACQA, (2) correcting the identified inaccurate contours/slices using ACR, (3) repeating steps (1) and (2) on the contours obtained from the previous ACR until the contours/slices become acceptable or the performance of ACR becomes saturated, and (4) correcting the un-correctable contours from ACR using ISS tool. The steps 1-3 are designed to be fully automated. Details on ACQA, ACR and ISS are being investigated in separate studies. Briefly, the ACQA module uses either machine learning or DL classification algorithms based on organ and image properties to validate contour quality. The ACR module utilizes multiple correction methods, including an active contour model (ACM), U-Net and Dense UNet DL models to sequentially refine the inaccurate contours with minor and/or major errors. The ISS tool uses the accurate contours on an adjacent slice as prior knowledge informant to guide a 2D UNet DL model to predict contours on a subsequent slice. We have demonstrated the performance of the multi-layer pipeline on complex abdominal organs on various MRIs. <h3>Results</h3> The ACQA models achieved AUC/accuracy of 0.95/0.89 for pancreas and 0.92/0.88 for stomach contours. The ACM model improved 54%/49% of bowels/pancreas contours from major into minor errors and the UNet/Dense UNet ACR model improved 43%/33% of the minor bowel/stomach contours to be acceptable. The ISS tool achieved DSC of 0.92, 0.91, 0.91, 0.87, and 0.87 for colon, small bowel, stomach, pancreas, and duodenum. The expected execution time of the pipeline on an auto-segmented MRI was within 5 minutes. <h3>Conclusion</h3> This proposed multi-layer pipeline can complement deep learning auto-segmentation and can improve the segmentation efficiency and accuracy. As part of our on-going efforts, the pipeline is being implemented as a post DLAS process, substantially speeding up daily recontouring in MRgART.

An Introductory Module in Medical Image Segmentation for BME Students

To support recent trends toward the use of patient-specific anatomical models from medical imaging data, we present a learning module for use in the undergraduate BME curriculum that introduces image segmentation, the process of partitioning digital images to isolate specific anatomical features. Five commercially available software packages were evaluated based on their perceived learning curve, ease of use, tools for segmentation and rendering, special tools, and cost: ITK-SNAP, 3D Slicer, OsiriX, Mimics, and Amira. After selecting the package best suited for a stand-alone course module on medical image segmentation, instructional materials were developed that included a general introduction to imaging, a tutorial guiding students through a step-by-step process to extract a skull from a provided stack of CT images, and a culminating assignment where students extract a different body part from clinical imaging data. This module was implemented in three different engineering courses, impacting more than 150 students, and student achievement of learning goals was assessed. ITK-SNAP was identified as the best software package for this application because it is free, easiest to learn, and includes a powerful, semi-automated segmentation tool. After completing the developed module based on ITK-SNAP, all students attained sufficient mastery of the image segmentation process to independently apply the technique to extract a new body part from clinical imaging data. This stand-alone module provides a low-cost, flexible way to bring the clinical and industry trends combining medical image segmentation, CAD, and 3D printing into the undergraduate BME curriculum.

P18.03.B Long term effects of therapeutic embolization of radiographically suspected meningiomas regarding tumour- and necrosis volume as well as patients' symptoms

Abstract Background Preoperative embolization of radiographically suspected intracranial meningiomas is performed to reduce intraoperative blood loss and to facilitate resection. However, there is paucity of evidence examining the effect of embolization as a sole intervention on the tumor volume and the patient's neurocognitive symptoms. The objective of this study was to investigates the influence of therapeutic embolization on tumor- and necrosis volume and neurocognitive symptoms in absence of surgical intervention. Material and Methods Between January 2019 and June 2021, 4 patients (median age 82, 1 female) with 5 radiographically suspected intracranial meningiomas, underwent endovascular embolization with liquid embolizate and coils, without subsequent surgery. Indication for embolization was history of previous multiple surgery contraindicating further surgical intervention or patients will. Neurological status (on admission, discharge from hospital and 5 months after embolization), medical history, medication, MRI (location, tumor- and necrosis volume, development of edema, complications), DSA (extent of embolization, arterial supply) and procedural complications were recorded. In 3 patients with 4 tumors, an MRI scan was performed. The calculation of tumor- and necrosis volume was done manually with a semiautomatic segmentation tool. One patient with one tumor received CT scans due to a cardiac pacemaker and was therefore excluded from volumetric analysis. Results The average tumor volume was 44,83 ± 13,82 cm3 on admission, 45,33 ± 16,3 cm3 5 days after embolization (+1%) and 39,93 ± 29,39 cm3 (-11%) after 5 months. The average necrosis diameter was 41,76 ± 9,33 mm after 5 days and 6,05 ± 6,43 mm after 5 months. The perifocal oedema decreased in 2/3 Patients. Tumor blood supply was via 2,4 ± 1,34 vessels, 1,2 ± 0,45 vessels were embolised. On admission, 3 patients showed neurocognitive symptoms. At discharge and in the follow-up, in 2 patients the symptoms had improved, one patient experienced a complete symptom regression. In one patient the embolization was complicated by dissection of the middle meningeal artery. The same patients perifocal oedema increased in the further clinical course with accompanying deterioration of the pre-existing hemiparesis from moderate to severe. Supported with dexamethasone medication, the oedema and symptoms declined to a mild residual paresis. None of the patients developed a long-term neurocognitive deficit related to embolization. Conclusion Therapeutic endovascular embolization of radiographically suspected intracranial meningiomas reduced the tumor volume and led to long-term improvement of pre-existing neurocognitive symptoms. In the future, these effects should be verified in larger studies.

Shedding of ctDNA, radiomics assessment of tumor disease volume (TDV), and concordance of mutations (mut) in synchronous liquid and tumor biopsies in metastatic breast cancer (MBC).

1086 Background: Genomic alterations driving MBC progression may be better captured by ctDNA reflecting clonal evolution, but it is currently unknow whether ctDNA analysis can replace tumor sequencing for clinical decision purposes. Aim: to study the concordance between mut in synchronous plasma and tumor samples prospectively collected from patients (pts) with MBC progressing on their last systemic therapy. Methods: MiSeq Amplicon-based NGS (custom panel of 60 cancer-related genes; BRCA1/2 and PALB2 not included) was performed in both tumor biopsy and plasma. The concordance of ESCAT Tier I and II mut ( PIK3CA, AKT1, ERBB2, ESR1, PTEN) was determined and correlated with mutant allele fraction (MAF), TDV, and clinical features. Findings from liquid biopsies were classified as true positive (TP-ctDNA) if a given mut was detected in both tumor and plasma and false negative (FN-ctDNA) if only in the tumor. TDV: all metastasis volume assessed by CT scan (excluding sclerotic bone metastasis), and analyzed by an experienced radiologist using the 3DSlicer semiautomatic segmentation tool (TDV = pixel size x number of pixels). Non-shedding cases were those where any mut was detected in tumor but none in plasma. Results: 88 cases were collected (luminal 64, HER2+ 17, triple negative 7). Median age at diagnosis 49 years (range 28-80). Radiomics assessment could be performed in 78/88 cases. The plasma/tissue concordance at case level was 74%. Discordance came from 23 cases; in 15 cases mut was only found in tissue and in 8 cases it was only detected in plasma. At gene level, PIK3CA had the highest concordance (79%); in ESR1 it was 52%. Higher concordance associated with non-luminal subtype (OR 0.08, 95%CI 0.002 – 0.59) and shorter interval between primary diagnosis and metastatic relapse (20.3 vs 51 months; p =.02), but not with MAF. FN-ctDNA occurred in 15/49 cases (31%) and associated with luminal subtype (p =.02), but not with other clinical variables. Non-shedding cases associated with older age (p =.03), luminal subtype (p =.007), low TDV (p =.0006) and &lt; 3 metastatic sites (p =.05). In patients with visceral metastasis (n = 45), higher TDV associated with lower probability of FN-ctDNA (p =.03). All non-luminal subtypes were shedders and all but one were TP-ctDNA. In multivariate analysis, higher probability of TP-ctDNA in luminal tumors associated with tumor sampling from a progressing lesion (OR 10.8; 95% 1.5 – 122; p =.03) and shorter interval between diagnosis of metastatic disease and biopsy (OR 0.96, 95% CI 0.92 – 0.99; p =.03). Conclusions: Our results suggest that ctDNA can detect a significant proportion of clinically relevant mut in MBC. Patients’ characteristics, tumor subtype, type of gene, and tumor volume should be integrated with ctDNA results to better inform clinical decisions.

Three-dimensional quantification of the relationship between the upper first molar and maxillary sinus.

ObjectivesThe present study aims to describe the relationship between upper first molar roots and maxillary sinus, for the first time with a truly three‐dimensional approach.MethodsFrom a retrospective cone‐beam computed tomography (CBCT) sample of the upper jaw, a total of 105 upper first molars in contact with maxillary sinus from 74 patients (male 24, female 50, mean age 42) were included in the present study. Segmentation of the upper first molar and maxillary sinus in CBCT was performed utilizing a semiautomatic livewire segmentation tool in MeVisLab v.3.1. The segmentations were analyzed in 3‐matic Medical 20.0 for root volume and the contact area between upper first molar roots and maxillary sinus. Analysis of variance test was applied to detect statistically significant differences between the roots.ResultsThe palatal root had the largest contact area with maxillary sinus 27.8 ± 21.4 mm2 (20% of the root area) followed by the mesiobuccal 20.5 ± 17.9 mm2 (17% of the root area) and distobuccal root 13.7 ± 12 mm2 (14% of the root area). A significant difference in the contact area of the different roots of the upper first molar was seen.ConclusionsThis study showed that 70% of the upper first molars were in contact with the maxillary sinus. The palatal root had on average a fifth of its root surface in contact with the sinus, while for mesiobuccal this was a sixth of its root surface and distobuccal roots this was somewhat less. The true 3D relationship could help to better understand maxillary anatomy in relation to occurring pathologies and treatment planning in this area.

Volumetric accuracy of different imaging modalities in acute intracerebral hemorrhage

BackgroundFollow-up imaging in intracerebral hemorrhage is not standardized and radiologists rely on different imaging modalities to determine hematoma growth. This study assesses the volumetric accuracy of different imaging modalities (MRI, CT angiography, postcontrast CT) to measure hematoma size.Methods28 patients with acute spontaneous intracerebral hemorrhage referred to a tertiary stroke center were retrospectively included between 2018 and 2019. Inclusion criteria were (1) spontaneous intracerebral hemorrhage (supra- or infratentorial), (2) noncontrast CT imaging performed on admission, (3) follow-up imaging (CT angiography, postcontrast CT, MRI), and (4) absence of hematoma expansion confirmed by a third cranial image within 6 days. Two independent raters manually measured hematoma volume by drawing a region of interest on axial slices of admission noncontrast CT scans as well as on follow-up imaging (CT angiography, postcontrast CT, MRI) using a semi-automated segmentation tool (Visage image viewer; version 7.1.10). Results were compared using Bland–Altman plots.ResultsMean admission hematoma volume was 18.79 ± 19.86 cc. All interrater and intrarater intraclass correlation coefficients were excellent (1; IQR 0.98–1.00). In comparison to hematoma volume on admission noncontrast CT volumetric measurements were most accurate in patients who received postcontrast CT (bias of − 2.47%, SD 4.67: n = 10), while CT angiography often underestimated hemorrhage volumes (bias of 31.91%, SD 45.54; n = 20). In MRI sequences intracerebral hemorrhage volumes were overestimated in T2* (bias of − 64.37%, SD 21.65; n = 10). FLAIR (bias of 6.05%, SD 35.45; n = 13) and DWI (bias of-14.6%, SD 31.93; n = 12) over- and underestimated hemorrhagic volumes.ConclusionsVolumetric measurements were most accurate in postcontrast CT while CT angiography and MRI sequences often substantially over- or underestimated hemorrhage volumes.

Morphometric and Microstructural Changes During Murine Retinal Development Characterized Using In Vivo Optical Coherence Tomography.

PurposeThe purpose of this study was to develop an in vivo optical coherence tomography (OCT) system capable of imaging the developing mouse retina and its associated morphometric and microstructural changes.MethodsThirty-four wild-type mice (129S1/SvlmJ) were anesthetized and imaged between postnatal (P) day 7 and P21. OCT instrumentation was developed to optimize signal intensity and image quality. Semi-automatic segmentation tools were developed to quantify the retinal thickness of the nerve fiber layer (NFL), inner plexiform layer (IPL), inner nuclear layer (INL), and the outer retinal layers (ORL), in addition to the total retina. The retinal maturation was characterized by comparing layer thicknesses between consecutive time points.ResultsFrom P7 to P10, the IPL increased significantly, consistent with retinal synaptogenesis. From P10 to P12, the IPL and ORL also increased, which is coherent with synaptic connectivity and photoreceptor maturation. In contrast, during these periods, the INL decreased significantly, consistent with cellular densification and selective apoptotic “pruning” of the tissue during nuclear migration. Thereafter from P12 to P21, the INL continued to thin (significantly from P17 to P21) whereas the other layers remained unchanged. No time-dependent changes were observed in the NFL. Overall, changes in the total retina were attributed to those in the IPL, INL, and ORL. Regions of the retina adjacent to the optic nerve head were thinner than distal regions during maturation.ConclusionsChanges in retinal layer thickness are consistent with retinal developmental mechanisms. Accordingly, this report opens new horizons in using our system in the mouse to characterize longitudinally developmental digressions in models of human diseases.