arXiv subjects

Haotian Feng

Publications and source records attributed to Haotian Feng.

11 recordsLinked to original sources

arXiv2026

4$\pi$ Planning for the Reduction of Predicted Hematologic Toxicity Risk in Cervical Cancer Radiotherapy

Purpose: In conventional coplanar radiotherapy for cervical cancer, nearby pelvic bones receive high radiation doses, increasing the risk of acute hematologic toxicity (HT). This study aims to estimate the HT risk reduction achievable with non-coplanar (4$\pi$) radiotherapy. Methods: We retrospectively analyzed 114 cervical cancer patients treated with coplanar volumetric modulated arc therapy (VMAT) between 2021 and 2023. Radiomic features from planning CTs were extracted and combined with clinical and dosimetric data to train predictive machine learning models. Patients were then replanned using integrated non-coplanar beam orientation and fluence map optimization (4$\pi$ planning). The top performing model was applied to these new plans to evaluate potential HT reduction. Results: Models combined radiomics, clinical, and dosimetric features achieved the highest predictive performance (AUC = 0.79). Feature importance analysis highlighted CT radiomic and dosimetric variables as the strongest predictors. Compared to VMAT, 4$\pi$ non-coplanar planning significantly reduced doses to the bone marrow (V10Gy, V20Gy, V30Gy, and V40Gy by 28%, 52%, 47%, and 33%, respectively) while significantly reducing dose to other pelvic organs at risk (OAR). When evaluated by the model, this improved dosimetry translated into a 23% reduction in predicted HT risk (risk ratio: 0.77; 95% CI: 0.75-0.79 and odds ratio: 0.68; 95% CI:0.65-0.71). Conclusions: Non-coplanar 4$\pi$ radiotherapy significantly lowers radiation doses to major pelvic bones during cervical cancer treatment without compromising target coverage or sparing of other OARs. Based on our predictive model, this superior dosimetry should translate to a marked reduction in acute hematologic toxicity.

physics.med-ph↗

arXiv2026

Prediction of Radiotherapy-Induced Hematologic Toxicity in Cervical Cancer with Cohort-Aware Framework

Hematologic toxicity (HT) is a major dose-limiting complication of pelvic radiotherapy for cervical cancer. Although radiomic and dosiomic features improve HT prediction beyond dosimetric metrics, their performance is highly sensitive to contour variability, limiting generalizability. We developed a cohort-aware representation-learning framework to address this challenge. We retrospectively analyzed 152 cervical cancer patients treated with pelvic radiotherapy without concurrent chemotherapy. Patients were divided into two cohorts based on the operators performing pelvic bone segmentation. HT prediction models were developed using cohort-specific training, pooled training, statistical harmonization, and a cohort-aware neural network (CANN) that learns shared and cohort-specific representations with contrastive regularization. Performance was evaluated using cross-validation and an independent test set. Cohort-specific models achieved test AUCs of 0.77 and 0.71, outperforming a dosimetry-only model (AUC=0.58). Directly pooling cohorts reduced performance (test AUC=0.64). Statistical harmonization provided limited benefit, while adversarial and correlation-based alignment further degraded performance. CANN achieved the best balance between robustness and generalizability (test AUC=0.72), with ablation studies confirming the importance of cohort-specific representations and contrastive alignment. These results demonstrate that cohort-aware representation learning effectively mitigates contour variability and improves the generalizability of radiomic and dosiomic models for HT prediction.

physics.med-ph↗

arXiv2025

Spherical Radiomics -- A Novel Approach to Glioblastoma Radiogenomic Analysis of Heterogeneity

We develop and validate a novel spherical radiomics framework for predicting key molecular biomarkers using multiparametric MRI. Conventional Cartesian radiomics extract tumor features on orthogonal grids, which do not fully capture the tumor's radial growth patterns and can be insensitive to evolving molecular signatures. In this study, we analyzed GBM radiomic features on concentric 2D shells, which were then mapped onto 2D planes for radiomics analysis. Radiomic features were extracted using PyRadiomics from four different regions in GBM. Feature selection was performed using ANOVA F-statistics. Classification was conducted with multiple machine-learning models. Model interpretability was evaluated through SHAP analysis, clustering analysis, feature significance profiling, and comparison between radiomic patterns and underlying biological processes. Spherical radiomics consistently outperformed conventional 2D and 3D Cartesian radiomics across all prediction tasks. The best framework reached an AUC of 0.85 for MGMT, 0.80 for EGFR, 0.80 for PTEN, and 0.83 for survival prediction. GLCM-derived features were identified as the most informative predictors. Radial transition analysis using the Mann-Whitney U-test demonstrates that transition slopes between T1-weighted contrast-enhancing and T2/FLAIR hyperintense lesion regions, as well as between T2 intense lesion and a 2 cm peritumoral expansion region, are significantly associated with biomarker status. Furthermore, the observed radiomic changes along the radial direction closely reflected known biological characteristics. Radiomic features extracted on the spherical surfaces at varying radial distances to the GBM tumor centroid are better correlated with important tumor molecular markers and patient survival than the conventional Cartesian analysis.

physics.med-ph↗

arXiv2025

Neural Network approximation power on homogeneous and heterogeneous reaction-diffusion equations

Reaction-diffusion systems represent one of the most fundamental formulations used to describe a wide range of physical, chemical, and biological processes. With the increasing adoption of neural networks, recent research has focused on solving differential equations using machine learning techniques. However, the theoretical foundation explaining why neural networks can effectively approximate such solutions remains insufficiently explored. This paper provides a theoretical analysis of the approximation power of neural networks for one- and two-dimensional reaction-diffusion equations in both homogeneous and heterogeneous media. Building upon the universal approximation theorem, we demonstrate that a two-layer neural network can approximate the one-dimensional reaction-diffusion equation, while a three-layer neural network can approximate its two-dimensional counterpart. The theoretical framework presented here can be further extended to elliptic and parabolic equations. Overall, this work highlights the expressive power of neural networks in approximating solutions to reaction-diffusion equations and related PDEs, providing a theoretical foundation for neural network-based differential equation solvers.

cs.LG↗

arXiv2025

Shedding Light on the Polymer's Identity: Microplastic Detection and Identification Through Nile Red Staining and Multispectral Imaging (FIMAP)

The widespread distribution of microplastics (MPs) in the environment presents significant challenges for their detection and identification. Fluorescence imaging has emerged as a promising technique for enhancing plastic particle detectability and enabling accurate classification based on fluorescence behavior. However, conventional segmentation techniques face limitations, including poor signal-to-noise ratio, inconsistent illumination, thresholding difficulties, and false positives from natural organic matter (NOM). To address these challenges, this study introduces the Fluorescence Imaging Microplastic Analysis Platform (FIMAP), a retrofitted multispectral camera with four optical filters and five excitation wavelengths. FIMAP enables comprehensive characterization of the fluorescence behavior of ten Nile Red-stained MPs: HDPE, LDPE, PP, PS, EPS, ABS, PVC, PC, PET, and PA, while effectively excluding NOM. Using K-means clustering for robust segmentation (Intersection over Union = 0.877) and a 20-dimensional color coordinate multivariate nearest neighbor approach for MP classification (>3.14 mm), FIMAP achieves 90% precision, 90% accuracy, 100% recall, and an F1 score of 94.7%. Only PS was occasionally misclassified as EPS. For smaller MPs (35-104 microns), classification accuracy declined, likely due to reduced stain sorption, fewer detectable pixels, and camera instability. Integrating FIMAP with higher-magnification instruments, such as a microscope, may enhance MP identification. This study presents FIMAP as an automated, high-throughput framework for detecting and classifying MPs across large environmental sample volumes.

cs.CV↗

arXiv2024

Multi-Modality and Temporal Analysis of Cervical Cancer Treatment Response

Cervical cancer presents a significant global health challenge, necessitating advanced diagnostic and prognostic approaches for effective treatment. This paper investigates the potential of employing multi-modal medical imaging at various treatment stages to enhance cervical cancer treatment outcomes prediction. We show that among Gray Level Co-occurrence Matrix (GLCM) features, contrast emerges as the most effective texture feature regarding prediction accuracy. Integration of multi-modal imaging and texture analysis offers a promising avenue for personalized and targeted interventions, as well as more effective management of cervical cancer. Moreover, there is potential to reduce the number of time measurements and modalities in future cervical cancer treatment. This research contributes to advancing the field of precision diagnostics by leveraging the information embedded in noninvasive medical images, contributing to improving prognostication and optimizing therapeutic strategies for individuals diagnosed with cervical cancer.

physics.med-ph↗

arXiv2023

Physics-Constrained Neural Network for Design and Feature-Based Optimization of Weave Architectures

Woven fabrics play an essential role in everyday textiles for clothing/sportswear, water filtration, and retaining walls, to reinforcements in stiff composites for lightweight structures like aerospace, sporting, automotive, and marine industries. Several possible combinations of weave patterns and material choices, which comprise weave architecture, present a challenging question about how they could influence the physical and mechanical properties of woven fabrics and reinforced structures. In this paper, we present a novel Physics-Constrained Neural Network (PCNN) to predict the mechanical properties like the modulus of weave architectures and the inverse problem of predicting pattern/material sequence for a design/target modulus value. The inverse problem is particularly challenging as it usually requires many iterations to find the appropriate architecture using traditional optimization approaches. We show that the proposed PCNN can effectively predict weave architecture for the desired modulus with higher accuracy than several baseline models considered. We present a feature-based optimization strategy to improve the predictions using features in the Grey Level Co-occurrence Matrix (GLCM) space. We combine PCNN with this feature-based optimization to discover near-optimal weave architectures to facilitate the initial design of weave architecture. The proposed frameworks will primarily enable the woven composite analysis and optimization process, and be a starting point to introduce Knowledge-guided Neural Networks into the complex structural analysis.

physics.app-ph↗

arXiv2023

Parameterization-based Neural Network: Predicting Non-linear Stress-Strain Response of Composites

Composite materials like syntactic foams have complex internal microstructures that manifest high-stress concentrations due to material discontinuities occurring from hollow regions and thin walls of hollow particles or microballoons embedded in a continuous medium. Predicting the mechanical response as non-linear stress-strain curves of such heterogeneous materials from their microstructure is a challenging problem. This is true since various parameters, including the distribution and geometric properties of microballoons, dictate their response to mechanical loading. To that end, this paper presents a novel Neural Network (NN) framework called Parameterization-based Neural Network (PBNN), where we relate the composite microstructure to the non-linear response through this trained NN model. PBNN represents the stress-strain curve as a parameterized function to reduce the prediction size and predicts the function parameters for different syntactic foam microstructures. We show that compared to several common baseline models considered in this paper, the PBNN can accurately predict non-linear stress-strain responses and the corresponding parameterized functions using smaller datasets. This is enabled by extracting high-level features from the geometry data and tuning the predicted response through an auxiliary term prediction. Although built in the context of the compressive response prediction of syntactic foam composites, our NN framework applies to predict generic non-linear responses for heterogeneous materials with internal microstructures. Hence, our novel PBNN is anticipated to inspire more parameterization-related studies in different Machine Learning methods.

physics.app-ph↗

arXiv2022

Role of Material Directionality on the Mechanical Response of Miura-Ori Composite Structures

This paper aims to understand the role of directional material properties on the mechanical responses of origami structures. We consider the Miura-Ori structures our target model due to their collapsibility and negative Poisson's ratio (NPR) effects, which are widely used in shock absorbers, disaster shelters, aerospace applications, etc. Traditional Miura-Ori structures are made of isotropic materials (Aluminum, Acrylic), whose mechanical properties like stiffness and NPR are well understood. However, how these responses are affected by directional materials, like Carbon Fiber Reinforced Polymer (CFRP) composites, lacks in-depth understanding. To that end, we study how fiber directions and arrangements in CFRP composites and Miura-Ori's geometric parameters control the stiffness and NPR of such structures. Through finite element analysis, we show that Miura-Ori structures made of CFRP composites can achieve higher stiffness and Poisson's ratio values than those made of an isotropic material like Aluminum. Then through regression analysis, we establish the relationship between different geometric parameters and the corresponding mechanical responses, which is further utilized to discover the Miura-Ori structure's optimal shape. We also show that the shear modulus is a dominant parameter that controls the mechanical responses mentioned above among the individual composite material properties within the Miura-Ori structure. We demonstrate that we can optimize the Miura-Ori structure by finding geometric and material parameters that result in combined stiffest and most compressible structures. We anticipate our research to be a starting point for designing and optimizing more sophisticated origami structures with composite materials incorporated.

physics.app-ph↗

arXiv2021

Densification Mechanics of Polymeric Syntactic Foams

In this paper, a fundamental understanding of the densification mechanics of polymeric syntactic foams under compressive loading is established. These syntactic foams are closed cell composite foams with thin-walled microballoons dispersed in a matrix (resin) whose closed cell structure provides excellent mechanical properties, like high strength and low density. There are several parameters that can contribute towards their mechanical properties, including, microballoon volume fraction, microballoon wall thickness, bonding between the microballoons and the matrix, and the crushing strength of microballoons. Conducting purely experimental testing by varying these parameters can be very time sensitive and expensive. Also, identification of densification mechanics is challenging using experiments only. Higher densification stress and energy are favorable properties under foam compression or crushing. Hence, the influence of key structural and material parameters associated with syntactic foams that dictate the mechanics of densification is studied here by implementing micromechanics based computational models and multiple linear regression analysis. Specifically, specific densification stresses and energy, which are densification stresses and energy normalized by weight, are evaluated which are more relevant for a wide variety of weight saving applications. Microballoon crushing strength and volume fraction are identified as the parameters that have the higher influence on densification stress and energy, and their specific counterparts, whereas the interfacial bonding has the least impact. In addition, designing aspects of syntactic foams with specified overall density are discussed by mapping microballoon volume fraction and wall thickness.

physics.app-ph↗

arXiv2021

Difference-Based Deep Learning Framework for Stress Predictions in Heterogeneous Media

Stress analysis of heterogeneous media, like composite materials, using Finite Element Analysis (FEA) has become commonplace in design and analysis. However, determining stress distributions in heterogeneous media using FEA can be computationally expensive in situations like optimization and multi-scaling. To address this, we utilize Deep Learning for developing a set of novel Difference-based Neural Network (DiNN) frameworks based on engineering and statistics knowledge to determine stress distribution in heterogeneous media, for the first time, with special focus on discontinuous domains that manifest high stress concentrations. The novelty of our approach is that instead of directly using several FEA model geometries and stresses as inputs for training a Neural Network, as typically done previously, we focus on highlighting the differences in stress distribution between different input samples for improving the accuracy of prediction in heterogeneous media. We evaluate the performance of DiNN frameworks by considering different types of geometric models that are commonly used in the analysis of composite materials, including volume fraction and spatial randomness. Results show that the DiNN structures significantly enhance the accuracy of stress prediction compared to existing structures, especially for composite models with random volume fraction when localized high stress concentrations are present.

physics.app-ph↗