Patient-specific Segmentation Research Articles

Patient-specific placental vessel segmentation with limited data

A major obstacle in applying machine learning for medical fields is the disparity between the data distribution of the training images and the data encountered in clinics. This phenomenon can be explained by inconsistent acquisition techniques and large variations across the patient spectrum. The result is poor translation of the trained models to the clinic, which limits their implementation in medical practice. Patient-specific trained networks could provide a potential solution. Although patient-specific approaches are usually infeasible because of the expenses associated with on-the-fly labeling, the use of generative adversarial networks enables this approach. This study proposes a patient-specific approach based on generative adversarial networks. In the presented training pipeline, the user trains a patient-specific segmentation network with extremely limited data which is supplemented with artificial samples generated by generative adversarial models. This approach is demonstrated in endoscopic video data captured during fetoscopic laser coagulation, a procedure used for treating twin-to-twin transfusion syndrome by ablating the placental blood vessels. Compared to a standard deep learning segmentation approach, the pipeline was able to achieve an intersection over union score of 0.60 using only 20 annotated images compared to 100 images using a standard approach. Furthermore, training with 20 annotated images without the use of the pipeline achieves an intersection over union score of 0.30, which, therefore, corresponds to a 100% increase in performance when incorporating the pipeline. A pipeline using GANs was used to generate artificial data which supplements the real data, this allows patient-specific training of a segmentation network. We show that artificial images generated using GANs significantly improve performance in vessel segmentation and that training patient-specific models can be a viable solution to bring automated vessel segmentation to the clinic.

Abstract PO1-11-10: Use of a standardized tool for the evaluation and treatment of cancer related cognitive impairment (chemo-brain) among breast cancer patients

Abstract There is a steep linear increase in patients surviving a cancer diagnosis. Clinical solutions that address the high prevalence of cancer-related treatment effects (CRTE) are urgently needed. A stark example of this need is the dearth of screening and treatment programs for cancer related cognitive impairment (CRCI) (colloquially called “chemo-brain”). Up to 75% of cancer patients report neurocognitive changes that leave patients unable to cope with the expectations of employers, colleagues, families, and friends. The prevalence of neurological disorders is 275% higher among adult cancer patients, when compared to an adult cancer-free population. For many adolescent and young adult cancer patients, the onset of CRCI occurs during "cognitive prime time"; a time when individuals are pursuing education, careers, and raising families. To address a gap in CRTE we have implemented a virtual multi-disciplinary intervention to measure and treat CRCI. Our CRCI intervention follows National Comprehensive Cancer Network and American Society of Clinical Oncology (ASCO) guidelines for CRCI. These guidelines indicate a multi-disciplinary approach with emphasis on increasing physical activity levels, cognitive-behavioral therapy, memory adaptive training, mood management, and sleep optimization. Our approach also follows recent ASCO guidelines for nutrition counseling. Our CRCI program utilizes medical oncologists, physician assistants, mental health therapists, registered dietitians, and occupational therapists. A nationally validated patient reported outcome (PRO) measurement, analysis, and reporting platform informs clinical treatment pathways, monitors patient progress, guides clinical decision making, and improves patient/provider communication. Using health status measures for CRCI, and following American Cancer Society and ASCO guidelines, patients are risk-stratified into three personalized cancer survivor care pathways: 1) low-complexity patients; 2) medium-complexity patients; and 3) high-complexity patients. Low-complexity patients baseline PROs score in the normal range, medium-complexity patients are those who score below the norm (moderate to large effect size difference from normal), and high-complexity patients are those who score well below the norm (greater than larger effect size difference from normal). Low-complexity patients are enrolled into a self-management and continuation of care program. Medium-complexity patients are enrolled into a shared care program. High-complexity patients are enrolled into an acute care program. All three clinical pathways include continued risk-factor vital-sign surveillance. Nearly 100% (99%) of patients who enroll in our primary cancer survivor care program complete our PRO baseline (N=80). From a convenient sample of breast cancer patients (N=42), provisional results of patient-specific segmentation analyses reveal a nearly perfect bell-shaped curve, with 51% of the population classified within the definition of medium-complexity and even split of outliers (high-complexity 24% and low-complexity 24%). Additionally, of all patients in our primary cancer survivor care program who completed at least three measurement time points for health status measures of cognitive impartment (N=38), 92% had a positive experience, 71% had a clinically significant improvement in cognitive health, 21% had a no change in cognitive status, and only 8% had a decline in cognitive health. If these data, and PRO implementation approach hold true, this program could be scaled to national and international cancer survivor population suffering from CRCI. Citation Format: John Librett, Mark Kosinski, Gregory Litton. Use of a standardized tool for the evaluation and treatment of cancer related cognitive impairment (chemo-brain) among breast cancer patients [abstract]. In: Proceedings of the 2023 San Antonio Breast Cancer Symposium; 2023 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2024;84(9 Suppl):Abstract nr PO1-11-10.

Dosimetric comparison of autocontouring techniques for online adaptive proton therapy

Objective. Anatomical and daily set-up uncertainties impede high precision delivery of proton therapy. With online adaptation, the daily plan is reoptimized on an image taken shortly before the treatment, reducing these uncertainties and, hence, allowing a more accurate delivery. This reoptimization requires target and organs-at-risk (OAR) contours on the daily image, which need to be delineated automatically since manual contouring is too slow. Whereas multiple methods for autocontouring exist, none of them are fully accurate, which affects the daily dose. This work aims to quantify the magnitude of this dosimetric effect for four contouring techniques. Approach. Plans reoptimized on automatic contours are compared with plans reoptimized on manual contours. The methods include rigid and deformable registration (DIR), deep-learning based segmentation and patient-specific segmentation. Main results. It was found that independently of the contouring method, the dosimetric influence of using automatic OAR contours is small (<5% prescribed dose in most cases), with DIR yielding the best results. Contrarily, the dosimetric effect of using the automatic target contour was larger (>5% prescribed dose in most cases), indicating that manual verification of that contour remains necessary. However, when compared to non-adaptive therapy, the dose differences caused by automatically contouring the target were small and target coverage was improved, especially for DIR. Significance. The results show that manual adjustment of OARs is rarely necessary and that several autocontouring techniques are directly usable. Contrarily, manual adjustment of the target is important. This allows prioritizing tasks during time-critical online adaptive proton therapy and therefore supports its further clinical implementation.

Deep Learning Segmentation of the Nucleus Basalis of Meynert on 3T MRI.

The nucleus basalis of Meynert is a key subcortical structure that is important in arousal and cognition and has been explored as a deep brain stimulation target but is difficult to study due to its small size, variability among patients, and lack of contrast on 3T MR imaging. Thus, our goal was to establish and evaluate a deep learning network for automatic, accurate, and patient-specific segmentations with 3T MR imaging. Patient-specific segmentations can be produced manually; however, the nucleus basalis of Meynert is difficult to accurately segment on 3T MR imaging, with 7T being preferred. Thus, paired 3T and 7T MR imaging data sets of 21 healthy subjects were obtained. A test data set of 6 subjects was completely withheld. The nucleus was expertly segmented on 7T, providing accurate labels for the paired 3T MR imaging. An external data set of 14 patients with temporal lobe epilepsy was used to test the model on brains with neurologic disorders. A 3D-Unet convolutional neural network was constructed, and a 5-fold cross-validation was performed. The novel segmentation model demonstrated significantly improved Dice coefficients over the standard probabilistic atlas for both healthy subjects (mean, 0.68 [SD, 0.10] versus 0.45 [SD, 0.11], P = .002, t test) and patients (0.64 [SD, 0.10] versus 0.37 [SD, 0.22], P < .001). Additionally, the model demonstrated significantly decreased centroid distance in patients (1.18 [SD, 0.43] mm, 3.09 [SD, 2.56] mm, P = .007). We developed the first model, to our knowledge, for automatic and accurate patient-specific segmentation of the nucleus basalis of Meynert. This model may enable further study into the nucleus, impacting new treatments such as deep brain stimulation.

Feasibility study of adaptive radiotherapy for esophageal cancer using artificial intelligence autosegmentation based on MR-Linac.

We proposed a scheme for automatic patient-specific segmentation in Magnetic Resonance (MR)-guided online adaptive radiotherapy based on daily updated, small-sample deep learning models to address the time-consuming delineation of the region of interest (ROI) in the adapt-to-shape (ATS) workflow. Additionally, we verified its feasibility in adaptive radiation therapy for esophageal cancer (EC). Nine patients with EC who were treated with an MR-Linac were prospectively enrolled. The actual adapt-to-position (ATP) workflow and simulated ATS workflow were performed, the latter of which was embedded with a deep learning autosegmentation (AS) model. The first three treatment fractions of the manual delineations were used as input data to predict the next fraction segmentation, which was modified and then used as training data to update the model daily, forming a cyclic training process. Then, the system was validated in terms of delineation accuracy, time, and dosimetric benefit. Additionally, the air cavity in the esophagus and sternum were added to the ATS workflow (producing ATS+), and the dosimetric variations were assessed. The mean AS time was 1.40 [1.10-1.78 min]. The Dice similarity coefficient (DSC) of the AS model gradually approached 1; after four training sessions, the DSCs of all ROIs reached a mean value of 0.9 or more. Furthermore, the planning target volume (PTV) of the ATS plan showed a smaller heterogeneity index than that of the ATP plan. Additionally, V5 and V10 in the lungs and heart were greater in the ATS+ group than in the ATS group. The accuracy and speed of artificial intelligence-based AS in the ATS workflow met the clinical radiation therapy needs of EC. This allowed the ATS workflow to achieve a similar speed to the ATP workflow while maintaining its dosimetric advantage. Fast and precise online ATS treatment ensured an adequate dose to the PTV while reducing the dose to the heart and lungs.

508 Deep Learning Segmentation of the Nucleus Basalis of Meynert for Deep Brain Stimulation Surgical Planning

INTRODUCTION: The nucleus basalis of Meynert (NBM) is a subcortical structure involved in arousal and cognition being explored as a deep brain stimulation (DBS) target for Alzheimer’s disease and other disorders. Given the small size of the NBM and variability between patients, accurate patient-specific targeting for DBS is needed. Accurate patient-specific manual segmentation is possible but requires high resolution 7T MRI. Therefore, the current standard is a non-patient-specific probabilistic atlas (Zaborszky 2008). METHODS: Paired 3T and 7T MRI datasets of 21 healthy subjects were obtained and the NBM was expertly segmented on 7T MRI. The 7T NBM segmentation was then used on the 3T MRI. To increase generalizability, we augmented the dataset to a total of 210 images. An external dataset of 14 patients with temporal lobe epilepsy (TLE) was used to validate the network’s performance on brains with pathological changes (Alvim 2016). A 3D-Unet convolutional neural network was constructed, and a 5-fold cross-validation was performed for model selection. The model was evaluated on healthy subjects using the held-out test dataset and on the external dataset of 14 TLE patients. RESULTS: When tested on the held-out dataset, the deep learning network demonstrated significantly improved dice coefficient compared against the probabilistic atlas for both healthy subjects (0.68 ± 0.08, 0.47 ± 0.06) and patients with TLE (0.63 ± 0.08, 0.38 ± 0.19). Additionally, the centroid distance was significantly decreased when using the network in patients with TLE (1.22 ± 0.33 mm, 3.25 ± 2.57 mm). CONCLUSIONS: We developed the first network, to our knowledge, for automatic and accurate patient-specific segmentation of the NBM using deep learning. The proposed network has less error than accepted targeting margins. This segmentation strategy allows accurate patient-specific targeting of the NBM for DBS.

Effect of transcatheter edge-to-edge repair device position on diastolic hemodynamic parameters: An echocardiography-based simulation study.

BackgroundTranscatheter edge-to-edge repair (TEER) has developed from innovative technology to an established treatment strategy of mitral regurgitation (MR). The risk of iatrogenic mitral stenosis after TEER is, however, a critical factor in the conflict of interest between maximal reduction of MR and minimal impairment of left ventricular filling. We aim to investigate systematically the impact of device position on the post treatment hemodynamic outcome by involving the patient-specific segmentation of the diseased mitral valve.Materials and methodsTransesophageal echocardiographic image data of ten patients with severe MR (age: 57 ± 8 years, 20% female) were segmented and virtually treated with TEER at three positions by using a position based dynamics approach. Pre- and post-interventional patient geometries were preprocessed for computational fluid dynamics (CFD) and simulated at peak-diastole with patient-specific blood flow boundary conditions. Simulations were performed with boundary conditions mimicking rest and stress. The simulation results were compared with clinical data acquired for a cohort of 21 symptomatic MR patients (age: 79 ± 6 years, 43% female) treated with TEER.ResultsVirtual TEER reduces the mitral valve area (MVA) from 7.5 ± 1.6 to 2.6 ± 0.6 cm2. Central device positioning resulted in a 14% smaller MVA than eccentric device positions. Furthermore, residual MVA is better predictable for central than for eccentric device positions (R2 = 0.81 vs. R2 = 0.49). The MVA reduction led to significantly higher maximal diastolic velocities (pre: 0.9 ± 0.2 m/s, post: 2.0 ± 0.5 m/s) and pressure gradients (pre: 1.5 ± 0.6 mmHg, post: 16.3 ± 9 mmHg) in spite of a mean flow rate reduction by 23% due to reduced MR after the treatment. On average, velocities were 12% and pressure gradients were 25% higher with devices in central compared to lateral or medial positions.ConclusionVirtual TEER treatment combined with CFD is a promising tool for predicting individual morphometric and hemodynamic outcomes. Such a tool can potentially be used to support clinical decision making, procedure planning, and risk estimation to prevent post-procedural iatrogenic mitral stenosis.

Patient-Specific Analysis of Ascending Thoracic Aortic Aneurysm with the Living Heart Human Model

In ascending thoracic aortic aneurysms (ATAAs), aneurysm kinematics are driven by ventricular traction occurring every heartbeat, increasing the stress level of dilated aortic wall. Aortic elongation due to heart motion and aortic length are emerging as potential indicators of adverse events in ATAAs; however, simulation of ATAA that takes into account the cardiac mechanics is technically challenging. The objective of this study was to adapt the realistic Living Heart Human Model (LHHM) to the anatomy and physiology of a patient with ATAA to assess the role of cardiac motion on aortic wall stress distribution. Patient-specific segmentation and material parameter estimation were done using preoperative computed tomography angiography (CTA) and ex vivo biaxial testing of the harvested tissue collected during surgery. The lumped-parameter model of systemic circulation implemented in the LHHM was refined using clinical and echocardiographic data. The results showed that the longitudinal stress was highest in the major curvature of the aneurysm, with specific aortic quadrants having stress levels change from tensile to compressive in a transmural direction. This study revealed the key role of heart motion that stretches the aortic root and increases ATAA wall tension. The ATAA LHHM is a realistic cardiovascular platform where patient-specific information can be easily integrated to assess the aneurysm biomechanics and potentially support the clinical management of patients with ATAAs.

Deep Cerebellar Nuclei Segmentation via Semi-Supervised Deep Context-Aware Learning from 7T Diffusion MRI.

Deep cerebellar nuclei are a key structure of the cerebellum that are involved in processing motor and sensory information. It is thus a crucial step to accurately segment deep cerebellar nuclei for the understanding of the cerebellum system and its utility in deep brain stimulation treatment. However, it is challenging to clearly visualize such small nuclei under standard clinical magnetic resonance imaging (MRI) protocols and therefore precise segmentation is not feasible. Recent advances in 7 Tesla (T) MRI technology and great potential of deep neural networks facilitate automatic patient-specific segmentation. In this paper, we propose a novel deep learning framework (referred to as DCN-Net) for fast, accurate, and robust patient-specific segmentation of deep cerebellar dentate and interposed nuclei on 7T diffusion MRI. DCN-Net effectively encodes contextual information on the patch images without consecutive pooling operations and adding complexity via proposed dilated dense blocks. During the end-to-end training, label probabilities of dentate and interposed nuclei are independently learned with a hybrid loss, handling highly imbalanced data. Finally, we utilize self-training strategies to cope with the problem of limited labeled data. To this end, auxiliary dentate and interposed nuclei labels are created on unlabeled data by using DCN-Net trained on manual labels. We validate the proposed framework using 7T B0 MRIs from 60 subjects. Experimental results demonstrate that DCN-Net provides better segmentation than atlas-based deep cerebellar nuclei segmentation tools and other state-of-the-art deep neural networks in terms of accuracy and consistency. We further prove the effectiveness of the proposed components within DCN-Net in dentate and interposed nuclei segmentation.

Generic thrombus segmentation from pre- and post-operative CTA.

Abdominal aortic aneurysm (AAA) is a localized, permanent and irreversible enlargement of the artery, with the formation of thrombus into the inner wall of the aneurysm. A precise patient-specific segmentation of the thrombus is useful for both the pre-operative planning to estimate the rupture risk, and for post-operative assessment to monitor the disease evolution. This paper presents a generic approach for 3D segmentation of thrombus from patients suffering from AAA using computed tomography angiography (CTA) scans. A fast and versatile thrombus segmentation approach has been developed. It is composed of initial centerline detection and aorta lumen segmentation, an optimized pre-processing stage and the use of a 3D deformable model. The approach has been designed to be very generic and requires minimal user interaction. The proposed method was tested on different datasets with 145 patients overall, including pre- and post-operative CTAs, abdominal aorta and iliac artery sections, different calcification degrees, aneurysm sizes and contrast enhancement qualities. The thrombus segmentation approach showed very accurate results with respect to manual delineations for all datasets ([Formula: see text] and [Formula: see text] for abdominal aorta sections on pre-operative CTA, iliac artery sections on pre-operative CTAs and aorta sections on post-operative CTA, respectively). Experiments on the different patient and image conditions showed that the method was highly versatile, with no significant differences in term of precision. Comparison with the level-set algorithm also demonstrated the superiority of the 3D deformable model. Average processing time was [Formula: see text]. We presented a near-automatic and generic thrombus segmentation algorithm applicable to a large variability of patient and imaging conditions. When integrated in an endovascular planning system, our segmentation algorithm shows its compatibility with clinical routine and could be used for pre-operative planning and post-operative assessment of endovascular procedures.

OP-269 Computational Patient-Specific Segmentation and Modelling of Degenerative Mitral Valve Disease from 3D Echocardiography

OP-269 Computational Patient-Specific Segmentation and Modelling of Degenerative Mitral Valve Disease from 3D Echocardiography

Patient-Specific Modeling of Cerebral Blood Flow: Geometrical Variations in a 1D Model

Cerebral aneurysm growth is influenced by hemodynamical factors, which are commonly estimated by means of a computational fluid dynamics (CFD) model of a small 3D patient-specific segment of the arterial tree. Since direct in vivo assessment of the boundary conditions for the 3D segment is often not possible, the flow pulses are generally computed with a patient-generic 1D model. The aim of this study is to evaluate the necessity of, and requirements for, full patient-specific modeling, i.e., both geometry and boundary conditions based on patient data. The flow waveforms obtained with several models from literature and a patient-specific geometry of the arterial tree are compared and analysed. By comparing patient-specific and patient-generic models, it is investigated whether different but all reasonable modeling assumptions result in substantial differences in flow distribution. The flow pulses obtained at the boundaries of the arterial segment of interest are significantly affected by the 1D geometry and boundary conditions associated with the different modeling assumptions. The geometry as well as the in- and outflow conditions of the 1D model should be patient-specific, as they significantly influence the flow pulses to be prescribed at the boundaries of the 3D arterial segment.

A comparative study on in vitro and in vivo effects of topical vasodilators in human internal mammary, radial artery and great saphenous vein

As an important prognostic factor of coronary artery bypass grafting (CABG), graft vasospasm can be observed in all currently used graft conduits. Radial artery (RA) vasospasm is more prone to occur in comparisons with internal mammary artery (IMA) and great saphenous vein (GSV). There is still controversy about which antispasmodic agent is superior to different grafts, especially to RA conduits. The aim of this pilot study was to investigate the relaxation response of four topical vasodilators to different in vitro grafts and how these vasodilators affect the blood flow of the vessel in situ during RA harvesting. Vasodilatory properties of diltiazem, nitroglycerin, urapidil and nicorandil were compared in matched patient-specific segments of RA, IMA and GSV harvested from 12 patients. The vasodilatory response of the RA to intraradial administration of nitroglycerin, diltiazem and urapidil was compared in vivo (n=10 per group) by assessing the free blood flow of RA. (1) The maximal relaxations occurring with urapidil, nitroglycerin and nicorandil in IMA, RA and SGV were significantly greater than that with diltiazem. The reactivity of all three graft conduits showed similar relaxation with nitroglycerin or with diltiazem, but the relaxation with urapidil in RA showed greater than that of IMA and GSV, and RA and GSV showed greater relaxation with nicorandil than IMA. (2) A dose of 10(-5)mol/l of nitroglycerin, urapidil and nicorandil but not diltiazem significantly inhibited the RA response to PE. (3) In vivo, urapidil and nitroglycerin significantly increased the RA blood flow, the potency of which was greater than that caused by diltiazem. (1) Comparing with nicorandil, urapidil and diltiazem, nitroglycerin caused a significant relaxation in all three graft vessels tested. (2) Nitroglycerin, nicorandil and urapidil were more effective in preventing RA spasm than diltiazem.

New approaches to catheter navigation for interventional radiology simulation

For over 20 years, interventional methods have improved the outcomes of patients with cardiovascular disease. However, these procedures require an intricate combination of visual and tactile feedback and extensive training. In this paper, we describe a series of novel approaches that have led to the development of a high-fidelity simulation system for interventional neuroradiology. In particular, we focus on a new approach for real-time deformation of devices such as catheters and guidewires during navigation inside complex vascular networks. This approach combines a real-time incremental Finite Element Model (FEM), an optimization strategy based on substructure decomposition, and a new method for handling collision response in situations where the number of contact points is very large. We also briefly describe other aspects of the simulation system, from patient-specific segmentation to the simulation of contrast agent propagation and fast volume-rendering techniques for generating synthetic X-ray images in real time. Although currently targeted at stroke therapy, our results are applicable to the simulation of any interventional radiology procedure.