arXiv subjects

Hongru Chen

Publications and source records attributed to Hongru Chen.

6 recordsLinked to original sources

arXiv2026

Patient-Level Diagnosis of Acute Myeloid Leukemia via Deep Learning Analysis of Bone Marrow Smear

Bone marrow smear review remains important for acute myeloid leukemia (AML) assessment, but manual single-cell interpretation is labor-intensive and patient-level diagnosis requires aggregation of many cellular observations. We present a cell-to-patient deep learning pipeline for AML-assisted diagnosis from bone marrow smear images. The study included 258 patients from six anonymized centers, including a main cohort of 169 patients from Centers 1-3 and an external validation cohort of 89 patients from Centers 4-6. A 16-category cell annotation vocabulary was used to describe the global cellular composition, including granulocytic, monocytic, erythroid, lymphoid, eosinophilic, and other cells. Rather than identifying strict AML blasts or leukemic blasts, the model targets an expert-defined composite category termed Composite Blast-like Cells (CBLC), comprising N, N1, M, M1, R, R1, J, and J1 according to the project-wide morphological standard. A fixed YOLO-based segmentation module detected cells, predicted contours were matched to expert polygon annotations by contour IoU, and standardized single-cell crops were generated. An EfficientNet-B0 classifier was trained through a two-stage GT-to-YOLO and YOLO-to-YOLO strategy with class-imbalance correction, center-border regularization, and morphology-assisted supervision. Cell-level predictions were aggregated into patient-level CBLC ratios for AML-oriented diagnostic support. The pipeline achieved stable internal validation and maintained external generalization, with ensemble weighted F1-scores of 0.9076, 0.8696, and 0.9124 on Centers 4, 5, and 6, respectively.

cs.CV↗

arXiv2026

Dense Point-to-Mask Optimization with Reinforced Point Selection for Crowd Instance Segmentation

Crowd instance segmentation is a crucial task with a wide range of applications, including surveillance and transportation. Currently, point labels are common in crowd datasets, while region labels (e.g., boxes) are rare and inaccurate. The masks obtained through segmentation help to improve the accuracy of region labels and resolve the correspondence between individual location coordinates and crowd density maps. However, directly applying currently popular large foundation models such as SAM does not yield ideal results in dense crowds. To this end, we first propose Dense Point-to-Mask Optimization (DPMO), which integrates SAM with the Nearest Neighbor Exclusive Circle (NNEC) constraint to generate dense instance segmentation from point annotations. With DPMO and manual correction, we obtain mask annotations from the existing point annotations for traditional crowd datasets. Then, to predict instance segmentation in dense crowds, we propose a Reinforced Point Selection (RPS) framework trained with Group Relative Policy Optimization (GRPO), which selects the best predicted point from a sampling of the initial point prediction. Through extensive experiments, we achieve state-of-the-art crowd instance segmentation performance on ShanghaiTech, UCF-QNRF, JHU-CROWD++, and NWPU-Crowd datasets. Furthermore, we design new loss functions supervised by masks that boost counting performance across different models, demonstrating the significant role of mask annotations in enhancing counting accuracy.

cs.CV↗

arXiv2026

Exclusivity-Guided Mask Learning for Semi-Supervised Crowd Instance Segmentation and Counting

Semi-supervised crowd analysis is a prominent area of research, as unlabeled data are typically abundant and inexpensive to obtain. However, traditional point-based annotations constrain performance because individual regions are inherently ambiguous, and consequently, learning fine-grained structural semantics from sparse anno tations remains an unresolved challenge. In this paper, we first propose an Exclusion-Constrained Dual-Prompt SAM (EDP-SAM), based on our Nearest Neighbor Exclusion Circle (NNEC) constraint, to generate mask supervision for current datasets. With the aim of segmenting individuals in dense scenes, we then propose Exclusivity-Guided Mask Learning (XMask), which enforces spatial separation through a discriminative mask objective. Gaussian smoothing and a differentiable center sampling strategy are utilized to improve feature continuity and training stability. Building on XMask, we present a semi-supervised crowd counting framework that uses instance mask priors as pseudo-labels, which contain richer shape information than traditional point cues. Extensive experiments on the ShanghaiTech A, UCF-QNRF, and JHU++ datasets (using 5%, 10%, and 40% labeled data) verify that our end-to-end model achieves state-of-the-art semi-supervised segmentation and counting performance, effectively bridging the gap between counting and instance segmentation within a unified framework.

cs.CV↗

arXiv2026

Data-Driven Prediction of Chaotic Transition in Periapsis Poincar\'e Maps

Chaotic trajectories in multi-body dynamical systems play a crucial role in designing low-energy trajectories in astrodynamics. However, predicting these trajectories is inherently difficult, as small errors in initial conditions can grow exponentially, making long-term predictions unreliable. This study introduces a novel methodology using Dynamic Mode Decomposition (DMD) to predict chaotic transitions in the periapsis Poincar\'e map of the circular restricted three-body problem (CRTBP). Unlike standard DMD approaches that model continuous equations of motion, the proposed method approximates deformations in a low-dimensional Poincar\'e map, enabling trajectory prediction and revealing transition structures. Two approaches are developed: the Local Deformation Map-based DMD (LDMD) and the Global Deformation Map-based DMD (GDMD). LDMD constructs discrete maps to track local deformations of periapsis sets, while GDMD captures global deformations using widely distributed data. A key advantage of this framework is that it approximates nonlinear chaotic transport using a linear operator, which enables fast prediction of periapsis evolution via matrix powers and direct access to geometric structures. To validate the proposed method, the deformation map is applied to design ballistic transfer trajectories to the Moon using a targeting strategy, demonstrating its practical relevance in astrodynamics. This work highlights the potential of data-driven modeling to bridge chaotic dynamics with systematic trajectory design.

nlin.CD↗

arXiv2023

Capacity of Sun-driven Lunar Swingby Sequences and Their Application in Asteroid Retrieval

For deep-space mission design, the gravity of the Sun and the Moon can be first considered and utilized. Their gravity can provide the energy change for launching spacecraft and retrieving spacecraft as well as asteroids. Regarding an asteroid retrieval mission, it can lead to the mitigation of asteroid hazards and an easy exploration and exploitation of the asteroid. This paper discusses the application of the Sun-driven lunar swingby sequence for asteroid missions. Characterizing the capacity of this technique is not only interesting in terms of dynamic insights but also non-trivial for trajectory design. The capacity of a Sun-driven lunar swingby sequence is elucidated in this paper with the help of the "Swingby-Jacobi" graph. The capacity can be represented by a range of the Jacobi integral that encloses around 660 asteroids currently cataloged. To facilitate trajectory design, a database of Sun-perturbed Moon-to-Moon transfers, including multi-revolution cases is generated and employed. Massive trajectory options for spacecraft launch and asteroid capture can then be explored and optimized. Finally, a number of asteroid flyby, rendezvous, sample-return, and retrieval mission options enabled by the proposed technique are obtained.

astro-ph.EP↗

arXiv2020

Accuracy of meteor positioning from space-and ground-based observations

Aims. The knowledge of the orbits and origins derived from meteors is important for the study of meteoroids and of the early solar system. With an increase in nano-satellite projects dedicated to Earth observations or directly to meteor observations (e.g., the Meteorix CubeSat), we investigate the stereoscopic measurement of meteor positions using a pair of cameras, one deployed in space and one on the ground, and aim to understand the accuracy and the main driving factors. This study will reveal the requirements for system setups and the geometry favorable for meteor triangulation. Methods. This Letter presents the principle of the stereoscopic measurement from space and the ground, and an error analysis. Specifically, the impacts of the resolutions of the cameras, the attitude and orbit determination accuracy of the satellite, and the geometry formed by the moving target and observers are investigated. Results. To reach a desirable positioning accuracy of 1 km it is necessary to equip the satellite with high-accuracy sensors (e.g., star tracker and GPS receiver) to perform fine attitude and orbit determination. The best accuracy can occur when the target is at an elevation of 30 deg. with respect to the ground station.

astro-ph.EP↗