3D Metrics Research Articles

Dry mass grassland estimation using UAV ultra-wide RGB images

Abstract. Dry mass is an important parameter to optimise grassland management. Traditionally, dry mass values are estimated manually by cutting, drying, and weighing vegetation samples. In large areas of cultivation, this becomes a time-consuming and costly activity. In recent years, many researchers have studied different sensors embedded in Unmanned Aerial Vehicles (UAV) to collect spatial data and estimate biomass using machine learning algorithms for forest and agricultural applications. However, there needs to be more research dealing with estimating production indices for pasture, especially in Brazil, as stated. This study evaluates the feasibility of using the GoPro wide-angle RGB camera on UAVs (Unmanned Aerial Vehicles) to estimate the dry mass of pastures. Different data analysis methods were compared, including the combination of vegetation indices (VIs) values and three-dimensional metrics (3D) extracted from the Canopy Height Model (CHM): all metrics (ALL), three VIs plus four 3D metrics (VI3 + CHM4) and only 3D metrics. Random Forest (RF) machine learning algorithm was used to estimate dry mass. The best results were obtained when merging all the variables from the two flight campaigns, with a coefficient of determination (R2) of 0.80 for the model and a Pearson Correlation Coefficient (PCC) of 0.85 for validation, with a Root Mean Square Error (RMSE%) of 20.5%. In summary, using RGB sensors embedded in UAVs is a promising technique for estimating farm grazing parameters.

Whole Organ Volumetric Sensing Ultrasound Localization Microscopy for Characterization of Kidney Structure.

Glomeruli are the filtration units of the kidney and their function relies heavily on their microcirculation. Despite its obvious diagnostic importance, an accurate estimation of blood flow in the capillary bundle within glomeruli defies the resolution of conventional imaging modalities. Ultrasound Localization Microscopy (ULM) has demonstrated its ability to image in-vivo deep organs in the body. Recently, the concept of sensing ULM or sULM was introduced to classify individual microbubble behavior based on the expected physiological conditions at the micrometric scale. In the kidney of both rats and humans, it revealed glomerular structures in 2D but was severely limited by planar projection. In this work, we aim to extend sULM in 3D to image the whole organ and in order to perform an accurate characterization of the entire kidney structure. The extension of sULM into the 3D domain allows better localization and more robust tracking. The 3D metrics of velocity and pathway angular shift made glomerular mask possible. This approach facilitated the quantification of glomerular physiological parameter such as an interior traveled distance of approximately 7.5 ±0.6 microns within the glomerulus. This study introduces a technique that characterize the kidney physiology which can serve as a method to facilite pathology assessment. Furthermore, its potential for clinical relevance could serve as a bridge between research and practical application, leading to innovative diagnostics and improved patient care.

The effect of handedness and biological sex on fingermark ridge height and volume as examined by 3D imaging.

For decades, detectives and forensic scientists have relied on friction ridge skin comparisons to identify individuals in criminal investigations. This method involves examining (latent) fingermarks, typically captured as two-dimensional (2D) images, and comparing them with known inked impressions or scans. The comparisons focus on general patterns, the location and orientation of minutiae, and additional characteristics like scars or pores. However, this common process overlooks other potentially valuable factors, such as handedness, biological sex, and the time since deposition, which could provide crucial investigative leads, especially in cases with partial or unclear ridge patterns. This study explored the potential of three-dimensional (3D) optical profilometry (OP) to observe changes in fingermarks ridge height and volume over 7 days while considering hand dominance and biological sex. The results indicated that all the assessed 3D metrics were generally consistent in detecting the effects of sex and handedness, mostly noticeable on aged fingermarks. Specifically, females and impressions from their non-dominant hand exhibited higher values. Additionally, the influence of time was evident, with fresh fingermarks showing significantly larger ridges in terms of height and volume. These findings suggest that 3D OP can enhance the probative value of fingermarks by inferring additional characteristics such as sex, handedness, and the age of the impression. These activity-level propositions could provide useful information in cases where only partial or poor-quality fingermarks are recovered, providing new data dimensions that could lead to more comprehensive forensic examinations.

Whole-body PET image denoising for reduced acquisition time.

A reduced acquisition time positively impacts the patient's comfort and the PET scanner's throughput. AI methods may allow for reducing PET acquisition time without sacrificing image quality. The study aims to compare various neural networks to find the best models for PET denoising. Our experiments consider 212 studies (56,908 images) for 7MBq/kg injected activity and evaluate the models using 2D (RMSE, SSIM) and 3D (SUVpeak and SUVmax error for the regions of interest) metrics. We tested 2D and 2.5D ResNet, Unet, SwinIR, 3D MedNeXt, and 3D UX-Net. We have also compared supervised methods with the unsupervised CycleGAN approach. The best model for PET denoising is 3D MedNeXt. It improved SSIM on 38.2% and RMSE on 28.1% in 30-s PET denoising and on 16.9% and 11.4% in 60-s PET denoising when compared to the original 90-s PET reducing at the same time SUVmax discrepancy dispersion.

Three-dimensional assessment of subchondral arc and hip joint coverage angles in the asymptomatic young adult hip.

The primary objective of this study was to develop a custom algorithm to assess three-dimensional (3D) acetabular coverage of the femoral head based on surface models generated from computed tomography (CT) imaging. The secondary objective was to apply this algorithm to asymptomatic young adult hip joints to assess the regional 3D acetabular coverage variability and understand how these novel 3D metrics relate to traditional two-dimensional (2D) radiographic measurements of coverage. The algorithm developed automatically identifies the lateral- and medial-most edges of the acetabular lunate at one-degree intervals around the acetabular rim based on local radius of curvature. The acetabular edges and the center of a best-fit sphere to the femoral head are then used to compute the mean 3D subchondral arc angles and hip joint coverage angles in five acetabular octants. This algorithm was applied to hip models generated from pelvis/hip CT imaging or abdomen/pelvis CT angiograms of 50 patients between 17 and 25 years of age who had no history of congenital or developmental hip pathology, neuromuscular conditions, or bilateral pelvic and/or femoral fractures. Corresponding 2D acetabular coverage measures of lateral center edge angle (LCEA) and acetabular arc angle (AAA) were assessed on the patients' clinical or digitally reconstructed radiographs. The 3D subchondral arc angle in the superior region (58.0 [54.6-64.8] degrees) was significantly higher (p < 0.001) than all other acetabular subregions. The 3D hip joint coverage angle in the superior region (26.2 [20.7-28.5] degrees) was also significantly higher (p < 0.001) than all other acetabular subregions. 3D superior hip joint coverage angle demonstrated the strongest correlation with 2D LCEA (r = 0.649, p < 0.001), while 3D superior-anterior subchondral arc angle demonstrated the strongest correlation with 2D AAA (r = 0.718, p < 0.001). The 3D coverage metrics in the remaining acetabular regions did not strongly correlate with typical 2D radiographic measures. The discrepancy between standard 2D measures of radiographic acetabular coverage and actual 3D coverage identified on advanced imaging indicates potential discord between anatomic coverage and the standard clinical measures of coverage on 2D imaging. As 2D measurement of acetabular coverage is increasingly used to guide surgical decision-making to address acetabular deformities, this work would suggest that 3D measures of acetabular coverage may be important to help discriminate local coverage deficiencies, avoid inconsistencies resulting from differences in radiographic measurement techniques, and provide a better understanding of acetabular coverage in the hip joint, potentially altering surgical planning and guiding surgical technique.

3D Object Quality Metrics and their Differences: How Can We Evaluate Quality of Digitization?

3D Object Quality Metrics and their Differences: How Can We Evaluate Quality of Digitization?

Deciphering the effects of 2D/3D urban morphology on diurnal cooling efficiency of urban green space

Urban green spaces (UGS) have emerged as a main nature-based solution proven to effectively cool urban areas, and cooling efficiency (CE) serves as a prominent indicator for assessing the performance of UGS. Yet few studies have explored the diurnal dynamics of CE and the comprehensive effects of both 2D and 3D urban morphology on CE over a 24-h period. In this study, we employed diurnal land surface temperature (LST) data from Ecosystem Spaceborne Thermal Radiometer Experiment on Space Station (ECOSTRESS) to delve into the diurnal CE of UGS within the urban core area of megacity Paris, France. Meanwhile, using high-resolution land cover data, digital height model, and explainable machine learning models, we elucidated the intricate relationship between various 2D and 3D urban morphology metrics and CE during a 24-h cycle. In summary, city-scale continuous monitoring revealed diurnal CE variations, with values ranging from 0.008 °C to 0.13 °C and a standard deviation of 0.04 °C. On an average of a day, the relative importance of the 2D and 3D urban form indicators was 77.25 % and 22.75 %, respectively. Enhancing the dispersion of trees or improving the integrity of non-tree vegetation patches proves to be an effective strategy for augmenting CE. The results can be further used in urban planning to improve urban thermal resilience and adaptability.

The cooling capacity of urban vegetation and its driving force under extreme hot weather: A comparative study between dry-hot and humid-hot cities

Extreme heat events in cities necessitate closer examination of urban vegetation’ capacity to mitigate extreme hot weather. However, the cooling capacity of vegetation (CCV) and its driving forces of 2D and 3D landscape metrics under extreme hot weather conditions in dry-hot and humid-hot cities, particularly within different functional zones is still not well understood. The results indicate that the CCV is more pronounced in humid-hot cities than in dry-hot cities, particularly during periods of extreme hot weather. This CCV varied significantly across different functional zones, being the lowest in commercial zones and the highest in green space zones. Moreover, residential zones in humid-hot cities have higher CCV when the weather is more extreme. In dry-hot and humid-hot cities, urban vegetation landscape configuration metrics (such as patch and edge densities) contribute significantly to the CCV, and landscape composition (vegetation coverage) also contributes, with a significant increase during periods of extreme hot weather, particularly in humid-hot cities. Additionally, the 3D building morphology plays a crucial role in moderating the CCV; however, its impact diminishes during extreme heat. We also found that the CCV does not increase further after vegetation coverage is greater than 30–40 % in dry-hot cities, compared to approximately 60–80 % in humid-hot cities. The CCV increased nonlinearly with increasing urban vegetation largest patch index and decreasing urban vegetation patch density in dry-hot and humid-hot cities, with different thresholds of vegetation coverage and largest patch index under normal and extreme hot weather. A higher building height increased the CCV in humid-hot cities and reduced the CCV in dry-hot cities, particularly under extreme hot weather, and sparse high-rise buildings were better suited to humid-hot cities for increasing urban CCV. These findings provide valuable insights for urban planners to enhance the urban CCV, contributing to long-term urban resilience and public health protection during extreme heat events.

Study on the contributions of 2D and 3D urban morphologies to the thermal environment under local climate zones

With the changes in global climate and rapid urbanization, identifying the contribution of urban morphology to the thermal environment under different scenarios could be helpful in proposing differentiated optimization responses of different scenarios for improving the thermal environment by means of natural cooling. This study calculated 2D and 3D metrics and retrieved surface temperature based on building vector data, land use, electronic map, the remote sensing images. Then, the study of spatial scale sensitivity was developed to build a nested multi-scale Local Climate Zone (LCZ) (NMSLCZ) map. Finally, the eXtreme Gradient Boost (XGBoost) regression model was performed to explore the relative contributions of 2D and 3D morphologies of different LCZ scenarios to surface temperatures. The results of scale sensitivity analysis showed the size of the optimal spatial scale decreased with the increase in building height. The 3D morphologies show more significant effects on LST variations than 2D morphologies. The surface temperature mainly by built type is significantly higher than that mainly by land cover type. Scenarios representing dense trees, shrubs, low vegetation, and water have significant cooling effects. In the scenarios of built types, the surface temperature of compact and large low-rise buildings is higher than that of the other building forms. Our findings provide a new perspective for LCZ mapping of scale optimization of the thermal environment. Meanwhile, the contributions of 2D and 3D morphologies on LSTs could help urban planners and managers understand the impacts of climate change and propose relevant strategies toward sustainable cities.

OATracker: Object-aware anti-occlusion 3D multiobject tracking for autonomous driving

Efficient and accurate 3D multiobject tracking (MOT) is essential for autonomous driving. However, existing methods still struggle with multiobject occlusion, in which interruptions or even loss of crucial tracked objects is likely. Therefore, a more intelligent object-aware anti-occlusion 3D MOT for autonomous driving, named OATracker, is developed in this paper. First, an updated 3D cascaded detection (3DCD) is adopted as the detection backbone, with incorporation of an updatable component based on the OpenPCDet framework to increase the longevity. Second, a reliable object state-constrained affinity (OSA) metric is designed that adaptively combines 3D geometric and motion metrics to dynamically measure objects’ size and distance in both spatial and temporal dimensions. Third, an object-aware trajectory management (OTM) module is developed for trajectory management and status updating. Specifically, the object identity-aware adaptive tracking life manager (ATLM) allows the tracker to adaptively manage the trajectories of captured objects through a confidence metric for each objects, and the ghost-aware trajectory postoptimizer (GTP) associates the historical and detected motion states, further overcoming id-switch and fragmentation problems. Finally, experiments on the KITTI-Car tracking benchmark and practical autonomous driving car application platform show that the OATracker can achieve competitive and robust tracking performance even in occlusion conditions.

Enhanced mesoscale and macroscale parameters for accurate 3D shape characterization

This paper investigates the limitations of widely used 3D Wadell's roundness and true sphericity metrics for characterizing granular materials. It introduces two improved indices: the mesoscale interlocking index and the radial distance factor. The study involves diverse particles scanned using optical interferometry and X-ray computed tomography. A mean curvature method is employed for surface segmentation, and a computationally efficient 3D mesh simplification procedure is proposed. Python scripts are developed for surface segmentation, mesh simplification, and 3D shape characterization. The new indices are validated by assessing angular and rounded 3D particles and changes in the shape of aggregates through the Los Angeles abrasion test. The results demonstrate the ability of the proposed indices to differentiate visually similar particles with identical Wadell's roundness and true sphericity, showcasing their effectiveness in accurately describing particle shape. The statistical distribution of 3D particle curvatures offers an alternative method for understanding particles' shape and abrasion.

Towards healthy cities: Modeling restorative potential of urban environments by coupling LiDAR-derived 3D metrics with panorama-based online survey

With the growing concern for the psychological well-being of urban dwellers, there has been a recognized need to explore the role of green environments in facilitating mental restorative benefits from the demands of city life. However, a significant gap exists in quantitatively modeling the restorative potential of urban environments, especially in terms of comprehensive 3D characteristics beyond mere greenery. This research gap has hindered our understanding of how to plan and design urban spaces that promote mental health. To bridge this gap, we conducted a study using Singapore as a case study to model the restorative potential for landscape planning and design by integrating LiDAR-derived 3D metrics, outcomes of an online survey, and a machine learning-based regularized regression model. We first collected data from 100 scenes, including both residential neighborhoods and urban parks, by extracting point cloud information from LiDAR scans for calculating a diverse set of 41 3D spatial metrics. Alongside these objective measurements, we conducted an online survey (N = 1500) using the Perceived Restorativeness Scale (PRS) where respondents evaluated the restorative potential of these scenes presented as interactive panoramic images. We subsequently utilized a regularized regression technique based on machine learning by conducting K-fold cross-validation (K = 10) with 1251 tests to identify the feature metrics significantly contributing to restorative potential. Our findings highlighted the importance of the green volumetric ratio as the most significant positive factor for restorative potential. Additionally, the results on the metrics of grey horizontal diversity and grey cluster number indicated a positive impact of unregulated and unaggregated grey components on restorative potential. The results on green horizontal diversity suggested that a uniform and concentrated spatial arrangement of greenery is a positive predictor, while other factors like front-background ratio, colorfulness, and slope variation also positively influenced restorative potential. In essence, our study not only provides novel insights but also introduces innovative methods for quantitatively assessing and characterizing the quality of urban environments that enhance psychological well-being. These findings inform the practical planning and design strategies aimed at fostering healthy and sustainable urban spaces.

3D Back Contour Metrics in Predicting Idiopathic Scoliosis Progression: Retrospective Cohort Analysis, Case Series Report and Proof of Concept.

Adolescent Idiopathic Scoliosis is a 3D spinal deformity commonly characterized by serial radiographs. Patients with AIS may have increased average radiation exposure compared to unaffected patients and thus may be implicated with a modest increase in cancer risk. To minimize lifetime radiation exposure, alternative imaging modalities such as surface topography are being explored. Surface topography (ST) uses a camera to map anatomic landmarks of the spine and contours of the back to create software-generated spine models. ST has previously shown good correlation to radiographic measures. In this study, we sought to use ST in the creation of a risk stratification model. A total of 38 patients met the inclusion criteria for curve progression prediction. Scoliotic curves were classified as progressing, stabilized, or improving, and a predictive model was created using the proportional odds logistic modeling. The results showed that surface topography was able to moderately appraise scoliosis curvatures when compared to radiographs. The predictive model, using demographic and surface topography measurements, was able to account for 86.9% of the variability in the future Cobb angle. Additionally, attempts at classification of curve progression, stabilization, or improvement were accurately predicted 27/38 times, 71%. These results provide a basis for the creation of a clinical tool in the tracking and prediction of scoliosis progression in order to reduce the number of X-rays required.

Efficacy of Cranial Orthosis for Plagiocephaly Based on 2D and 3D Evaluation.

Background With the advent of cranial orthoses as therapeutic medical devices for the treatment of severe positional head deformities in Japan, an increasing number of patients are being treated with them. However, assessing the effectiveness of a treatment is often difficult due to the use of different metrics. This study aimed to evaluate the effectiveness of cranial orthoses for deformational plagiocephaly using two- (2D) and three-dimensional (3D) evaluation metrics. Methods We conducted a retrospective study of infant patients with deformational plagiocephaly who underwent cranial orthosis treatment. We evaluated the severity of deformational plagiocephaly using cranial asymmetry (CA) and the cranial vault asymmetry index (CVAI) as 2D metrics, and anterior and posterior symmetry ratios as 3D metrics. The patients were divided into 24 subgroups based on the initial severity of each outcome and their age at the start of treatment. We analyzed the changes in outcomes and correlations within improvements across the age and severity categories. Results Overall, 1,038 infants were included in this study. The mean CA, CVAI, and anterior and posterior symmetry ratios improved significantly after cranial orthosis treatment. The improvement in each score was greater in patients with more severe initial deformities and in those who underwent treatment at a younger age. Conclusion Cranial orthosis treatment was effective in correcting deformational plagiocephaly in infants, as demonstrated by improvements in both 2D and 3D metrics. Patients with more severe initial deformities and those who underwent treatment at a younger age showed greater improvement.

A Variational Autoencoder Cascade Generative Adversarial Network for Scalable 3D Object Generation and Reconstruction

Generative Adversarial Networks (GANs) for 3D volume generation and reconstruction, such as shape generation, visualization, automated design, real-time simulation, and research applications, are receiving increased amounts of attention in various fields. However, challenges such as limited training data, high computational costs, and mode collapse issues persist. We propose combining a Variational Autoencoder (VAE) and a GAN to uncover enhanced 3D structures and introduce a stable and scalable progressive growth approach for generating and reconstructing intricate voxel-based 3D shapes. The cascade-structured network involves a generator and discriminator, starting with small voxel sizes and incrementally adding layers, while subsequently supervising the discriminator with ground-truth labels in each newly added layer to model a broader voxel space. Our method enhances the convergence speed and improves the quality of the generated 3D models through stable growth, thereby facilitating an accurate representation of intricate voxel-level details. Through comparative experiments with existing methods, we demonstrate the effectiveness of our approach in evaluating voxel quality, variations, and diversity. The generated models exhibit improved accuracy in 3D evaluation metrics and visual quality, making them valuable across various fields, including virtual reality, the metaverse, and gaming.

Applying 3D spatial metrics for landscape planning: Creating and measuring landscape scenarios by a point cloud-based approach

The current planning and design process predominantly relies on experts' experiences or preferences, which is beneficial and straightforward but can be subjective and time-consuming. Although there are some quantitative assessment tools available to aid decision-makers in understanding the perceived quality of landscape scenarios, how to integrate them into the design thinking process to improve landscape design is challenging. This study presents a 3D digital model integrating the quantitative representation of visual quality into the design thinking process to improve urban landscapes' restorative potential. The model consists of generating landscape scenarios in a 3D environment, converting them into point clouds and voxelising them, quantitatively assessing the restorative potential, and interactively modifying landscape designs. To demonstrate the feasibility and effectiveness of the model, we have applied it to improve the restorative potential of a park in Singapore. The results show that the model provides a clear assessment of landscape scenarios and identifies specific areas needing improvement. The significance of this study lies in an innovative approach to generating and assessing the landscape scenario, ultimately enhancing the overall quality of landscapes during the digital planning and design process.

Abstract 16846: Three-Dimensional Assessment of Ascending Aortic Motion by Dynamic CT in Sporadic and Marfan Thoracic Aortic Aneurysm

Background: The ascending aorta exhibits a complex, 3D motion related to cyclic pressure changes and downward traction forces related to left ventricular contraction. However, most metrics of disease severity are static (i.e., diameter) and among prior studies that have investigated ascending aortic motion, 2D imaging is typically employed. Vascular deformation mapping (VDM) is an image analysis technique that accurately quantifies 3D displacement of the aortic wall. Aims We aimed to quantify 3D metrics of cyclic ascending aortic motion and distensibility in patients with aTAA, and evaluate associations between motion metrics and patient characteristics in sporadic TAA (SaTAA) and Marfan syndrome (MFS). Methods: We included adult patients with aTAA, either SaTAA or MFS, and multi-phasic CTA of the thoracic aorta. VDM analysis of CTA data was performed with metrics including: 1) Axial, in-plane and rotational displacements at the root, 2) distensibility assessed by cyclic 2D cross-sectional area change (Dist-2D) at the root [(ΔA/A diastole х PP)], and 3) distensibility assessed by VDM-derived cyclic surface area stretch [(Area ratio-1)*1/PP] at the root and tubular ascending (Dist-3D). Aortic growth rate (GR) was measured from follow-up CTA. Results: We analyzed 51 with SaTAA (33% female, 62 ± 11 years old), and 14 with MFS (50% female, 42 ± 13 years old). In SaTAA, Dist-2D was associated (β=0.59, p=0.047) with axial displacement and AI was strongly associated (β=1.03 p=0.001) with rotation. In MFS, BAV and diameter were associated with in-plane displacement (β=5.50, p=0.004; β=-0.026, p=0.001). Among MFS patients with follow-up CTAs (n=10), ascending distensibility (Dist-3D) was marginally negatively associated with ascending GR (β= -0.27, p=0.058) Conclusion: 3D analysis of aortic motion in aTAA showed differences in root motion, namely related to valve disease and age, as well as differences in aortic distensibility that may predict ascending aortic GR in MFS.

Representing the landscape visual quality of residential green spaces in Singapore with 3D spatial metrics

Visual quality relates to people’s views, feelings about visiting landscapes, and restorative potential, which is crucial in planning urban spaces. This study has investigated the visual quality, particularly in its connection to restorative potential, of Singapore’s residential green spaces (RGSs) that are easily accessible and frequently visited by residents. We deployed a laser scanner to document 152 RGSs in high-density public housing estates and represented their landscape visual quality (LVQ) with previously developed 3D spatial metrics. The spatial metrics correlated to the Perceived Restorative Scale (PRS) of 1500 participants from an online survey were employed for cluster analysis to identify RGSs’ main features. The results concluded five main RGS clusters in Singapore: “High naturalness and low visual scale”, “Low naturalness and coherence”, “High visual scale and low coherence”, “High coherence and low visual scale”, and “Medium-high visual and low complexity”. A set of practical recommendations based on the weaknesses of each cluster were also provided to support their LVQ improvement. The significance of the study lies in the point cloud-based approach to document and characterise RGSs’ LVQ with 3D spatial metrics and paves a typological way to facilitate RGSs planning.

Measuring the Consistency and Diversity of 3D Face Generation

3D-aware GANs have shown their impressive power on 3D controlling for synthesized portraits. While the plausible facial reality is achieved, the inherent 3D properties of the generated results have actually not been well analyzed. One of the reasons is that the wildly-used metrics, such as Inception Score (IS) or Fréchet Inception Distance (FID), focus more on the perceptual features rather than explicit 3D clues. In this paper, we propose two novel 3D metrics, which measure the face consistency and diversity on a 3D level, to compensate for IS or FID on GAN evaluation. With such metrics, the generated faces are systematically analyzed on a wide range of GANs, where we obtain reasonable but different conclusions from the wildly-used approaches. Inspired by the performance on the proposed metrics, we then discuss what is crucial for the GANs to get robust 3D properties, and what may bring uncertainty or inconsistency to the generating procedure. Finally, we propose two simple but efficient solutions to contribute to superior 3D synthesizing accuracy across different architectures, which further demonstrate the significance of the proposed metrics.

Complexity Metrics and Planning Dose-Based Pretreatment Patient-Specific Quality Assurance Prediction: Classification, Gamma Passing Rates, and DVH Deviation

Patient-specific quality assurance (QA) prediction before treatment is beneficial to the clinical resource allocation and the dosimetric monitoring of the patient plans. The aim of this study is to investigate the potential of complexity metrics of radiotherapy plan and patient planning dose to predict QA result, gamma passing rates and dose-volume indices deviation. Planning dose from treatment planning system (TPS), reconstructed dose from a vendor provided QA phantom and complexity metrics of the 499 radiotherapy plans of patients in our institution from March 2022 to September 2022 were used for methodology verification. Gamma passing rate (3%/2mm,10% threshold) 90% was regarded as criterion of QA pass or fail. A deep learning model ResNet-50 was modified to 3D dose processing and a multilayer perceptron (MLP) with three layers were adopted to extract features from 3D dose and 1D metrics in two parallel ways, then, the features were concatenate together to predict QA results. The dataset was split into 349 for train, 50 for validation and 100 for testing. Evaluation of predictions was based on absolute value deviation and area under the curves (AUC) of receiver operator characteristic (ROC) curve. In this dataset, 71% (355/499) plans pass the pretreatment QA test. For QA passing prediction in 100 testing cases, the AUC of ROC could achieve 0.92. For gamma passing rates prediction, a mean absolute error (MAE) of 1.8% could be observed for cases with gamma passing rates bigger than 90%, and a MAE of 4.5% deviation could be observed for cases with gamma passing rates from 80% to 90%. For PTV ΔD95 (%) and PTV ΔHI (%), the MAE of prediction and ground truth is 1%. The model with only complexity metrics and only 3D dose could achieve the AUC of ROC 0.91 and 0.84, respectively. The complexity metrics and 3D planning dose-based model could predict pretreatment patient specific QA results with high accuracy and the complexity metrics play a leading role in the model. Dose-volume metrics deviations of PTV could be predicted and more clinically useful information could be provided.