arXiv subjects

D. Ma

Publications and source records attributed to D. Ma.

5 recordsLinked to original sources

arXiv2025

Comprehensive Evaluation of OCT-based Automated Segmentation of Retinal Layer, Fluid and Hyper-Reflective Foci: Impact on Clinical Assessment of Diabetic Retinopathy Severity

Diabetic retinopathy (DR) is a leading cause of vision loss, requiring early and accurate assessment to prevent irreversible damage. Spectral Domain Optical Coherence Tomography (SD-OCT) enables high-resolution retinal imaging, but automated segmentation performance varies, especially in cases with complex fluid and hyperreflective foci (HRF) patterns. This study proposes an active-learning-based deep learning pipeline for automated segmentation of retinal layers, fluid, and HRF, using four state-of-the-art models: U-Net, SegFormer, SwinUNETR, and VM-UNet, trained on expert-annotated SD-OCT volumes. Segmentation accuracy was evaluated with five-fold cross-validation, and retinal thickness was quantified using a K-nearest neighbors algorithm and visualized with Early Treatment Diabetic Retinopathy Study (ETDRS) maps. SwinUNETR achieved the highest overall accuracy (DSC = 0.7719; NSD = 0.8149), while VM-UNet excelled in specific layers. Structural differences were observed between non-proliferative and proliferative DR, with layer-specific thickening correlating with visual acuity impairment. The proposed framework enables robust, clinically relevant DR assessment while reducing the need for manual annotation, supporting improved disease monitoring and treatment planning.

eess.IV↗

arXiv2024

Real-Time Analysis of Nanoscale Dynamics in Membrane Protein Insertion via Single-Molecule Imaging

Membrane proteins often need to be inserted into or attached on the cell membrane to perform their functions. Understanding their transmembrane topology and conformational dynamics during insertion is crucial for elucidating their roles. However, it remains challenging to monitor nanoscale changes in insertion depth of individual proteins in membranes. Here, we introduce two single molecule imaging methods, SIFA and LipoFRET, designed for in vitro observation of the nanoscale architecture of membrane proteins within membranes. These methods have demonstrated their efficacy in studying biomolecules interacting with bio-membranes with sub-nanometer precision.

physics.bio-ph↗

arXiv2024

With Greater Text Comes Greater Necessity: Inference-Time Training Helps Long Text Generation

Long text generation, such as novel writing and discourse-level translation with extremely long contexts, presents significant challenges to current language models. Existing methods mainly focus on extending the model's context window through strategies like length extrapolation. However, these approaches demand substantial hardware resources during the training and/or inference phases. Our proposed method, Temp-Lora, introduces an alternative concept. Instead of relying on the KV cache to store all context information, we embeds this information directly into a temporary Lora module. In the process of long text generation, this module is progressively trained with text generated previously. This approach not only efficiently preserves contextual knowledge but also prevents any permanent alteration to the model's parameters given that the module is discarded post-generation. Extensive experiments on the PG19 language modeling benchmark and the GuoFeng discourse-level translation benchmark validate the effectiveness of Temp-Lora. Our results show that: 1) Temp-Lora substantially enhances generation quality for long text, as indicated by a 13.2% decrease in perplexity (PPL) on a subset of PG19, and a 29.3% decrease in PPL along with a 113.2% increase in BLEU score on a subset of GuoFeng, 2) Temp-Lora is compatible with and enhances most existing long text generation methods, and 3) Temp-Lora can greatly reduce computational costs by shortening the context window. For example, we can ensure a moderate improvement in generation quality (a decrease of 3.8% in PPL) while enabling a 51.5% memory usage reduction and a 60.0% decrease in latency for inference.

cs.CL↗

arXiv2016

Clinically applicable Monte Carlo-based biological dose optimization for the treatment of head and neck cancers with spot-scanning proton therapy

Purpose: To demonstrate the feasibility of fast Monte Carlo (MC) based inverse biological planning for the treatment of head and neck tumors in spot-scanning proton therapy. Methods: Recently, a fast and accurate Graphics Processor Unit (GPU)-based MC simulation of proton transport was developed and used as the dose calculation engine in a GPU-accelerated IMPT optimizer. Besides dose, the dose-averaged linear energy transfer (LETd) can be simultaneously scored, which makes biological dose (BD) optimization possible. To convert from LETd to BD, a linear relation was assumed. Using this novel optimizer, inverse biological planning was applied to 4 patients: 2 small and 1 large thyroid tumor targets, and 1 glioma case. To create these plans, constraints were placed to maintain the physical dose (PD) within 1.25 times the prescription while maximizing target BD. For comparison, conventional IMRT and IMPT plans were created for each case in Eclipse (Varian, Inc). The same critical structure PD constraints were used for the IMRT, IMPT and bio-optimized plans. The BD for the IMPT plans were obtained through MC re-calculations. Results: Compared to standard IMPT, the bio-optimal plans for patients with small tumor targets displayed a BD escalation that was around twice the PD increase. Dose sparing to critical structures was improved compared to both IMRT and IMPT. No significant BD increase could be achieved for the large thyroid case, and when the presence of critical structures mitigated the contribution of additional fields. The calculation of the bio-optimized plans can be completed in a clinically viable time (<30 minutes) on a 24-GPU system. Conclusion: By exploiting GPU acceleration, MC-based, biologically optimized plans were created for small-target tumor patients. This optimizer will be used in an upcoming feasibility trial on LETd painting for radio-resistant tumors.

physics.med-ph↗

arXiv2008

Ab-initio simulation and experimental validation of beta-titanium alloys

In this progress report we present a new approach to the ab-initio guided bottom up design of beta-Ti alloys for biomedical applications using a quantum mechanical simulation method in conjunction with experiments. Parameter-free density functional theory calculations are used to provide theoretical guidance in selecting and optimizing Ti-based alloys with respect to three constraints: (i) the use of non-toxic alloy elements; (ii) the stabilization of the body centered cubic beta phase at room temperature; (iii) the reduction of the elastic stiffness compared to existing Ti-based alloys. Following the theoretical predictions, the alloys of interest are cast and characterized with respect to their crystallographic structure, microstructure, texture, and elastic stiffness. Due to the complexity of the ab initio calculations, the simulations have been focused on a set of binary systems of Ti with two different high melting bcc metals, namely, Nb and Mo. Various levels of model approximations to describe mechanical and thermodynamic properties are tested and critically evaluated. The experiments are conducted both, on some of the binary alloys and on two more complex engineering alloy variants, namely, Ti-35wt.%Nb-7wt.%Zr-5wt.%Ta and a Ti-20wt.%Mo-7wt.%Zr-5wt.%Ta.

cond-mat.mtrl-sci↗