arXiv subjects

Junyong Zhao

Publications and source records attributed to Junyong Zhao.

7 recordsLinked to original sources

arXiv2025

SpinalSAM-R1: A Vision-Language Multimodal Interactive System for Spine CT Segmentation

The anatomical structure segmentation of the spine and adjacent structures from computed tomography (CT) images is a key step for spinal disease diagnosis and treatment. However, the segmentation of CT images is impeded by low contrast and complex vertebral boundaries. Although advanced models such as the Segment Anything Model (SAM) have shown promise in various segmentation tasks, their performance in spinal CT imaging is limited by high annotation requirements and poor domain adaptability. To address these limitations, we propose SpinalSAM-R1, a multimodal vision-language interactive system that integrates a fine-tuned SAM with DeepSeek-R1, for spine CT image segmentation. Specifically, our SpinalSAM-R1 introduces an anatomy-guided attention mechanism to improve spine segmentation performance, and a semantics-driven interaction protocol powered by DeepSeek-R1, enabling natural language-guided refinement. The SpinalSAM-R1 is fine-tuned using Low-Rank Adaptation (LoRA) for efficient adaptation. We validate our SpinalSAM-R1 on the spine anatomical structure with CT images. Experimental results suggest that our method achieves superior segmentation performance. Meanwhile, we develop a PyQt5-based interactive software, which supports point, box, and text-based prompts. The system supports 11 clinical operations with 94.3\% parsing accuracy and sub-800 ms response times. The software is released on https://github.com/6jm233333/spinalsam-r1.

cs.CV↗

arXiv2025

MorphSAM: Learning the Morphological Prompts from Atlases for Spine Image Segmentation

Spine image segmentation is crucial for clinical diagnosis and treatment of spine diseases. The complex structure of the spine and the high morphological similarity between individual vertebrae and adjacent intervertebral discs make accurate spine segmentation a challenging task. Although the Segment Anything Model (SAM) has been proposed, it still struggles to effectively capture and utilize morphological information, limiting its ability to enhance spine image segmentation performance. To address these challenges, in this paper, we propose a MorphSAM that explicitly learns morphological information from atlases, thereby strengthening the spine image segmentation performance of SAM. Specifically, the MorphSAM includes two fully automatic prompt learning networks, 1) an anatomical prompt learning network that directly learns morphological information from anatomical atlases, and 2) a semantic prompt learning network that derives morphological information from text descriptions converted from the atlases. Then, the two learned morphological prompts are fed into the SAM model to boost the segmentation performance. We validate our MorphSAM on two spine image segmentation tasks, including a spine anatomical structure segmentation task with CT images and a lumbosacral plexus segmentation task with MR images. Experimental results demonstrate that our MorphSAM achieves superior segmentation performance when compared to the state-of-the-art methods.

eess.IV↗

arXiv2022

Low-Dose CT Denoising via Sinogram Inner-Structure Transformer

Low-Dose Computed Tomography (LDCT) technique, which reduces the radiation harm to human bodies, is now attracting increasing interest in the medical imaging field. As the image quality is degraded by low dose radiation, LDCT exams require specialized reconstruction methods or denoising algorithms. However, most of the recent effective methods overlook the inner-structure of the original projection data (sinogram) which limits their denoising ability. The inner-structure of the sinogram represents special characteristics of the data in the sinogram domain. By maintaining this structure while denoising, the noise can be obviously restrained. Therefore, we propose an LDCT denoising network namely Sinogram Inner-Structure Transformer (SIST) to reduce the noise by utilizing the inner-structure in the sinogram domain. Specifically, we study the CT imaging mechanism and statistical characteristics of sinogram to design the sinogram inner-structure loss including the global and local inner-structure for restoring high-quality CT images. Besides, we propose a sinogram transformer module to better extract sinogram features. The transformer architecture using a self-attention mechanism can exploit interrelations between projections of different view angles, which achieves an outstanding performance in sinogram denoising. Furthermore, in order to improve the performance in the image domain, we propose the image reconstruction module to complementarily denoise both in the sinogram and image domain.

eess.IV↗

arXiv2021

On the number of zeros to the equation $f(x_1)+...+f(x_n)=a$ over finite fields

Let $p$ be a prime, $k$ a positive integer and let $\mathbb{F}_q$ be the finite field of $q=p^k$ elements. Let $f(x)$ be a polynomial over $\mathbb F_q$ and $a\in\mathbb F_q$. We denote by $N_{s}(f,a)$ the number of zeros of $f(x_1)+\cdots+f(x_s)=a$. In this paper, we show that $$\sum_{s=1}^{\infty}N_{s}(f,0)x^s=\frac{x}{1-qx} -\frac{x { M_f^{\prime}}(x)}{qM_f(x)},$$ where $$M_f(x):=\prod_{m\in\mathbb F_q^{\ast}\atop{S_{f, m}\ne 0}}\Big(x-\frac{1}{S_{f,m}}\Big)$$ with $S_{f, m}:=\sum_{x\in \mathbb F_q}ζ_p^{{\rm Tr}(mf(x))}$, $ζ_p$ being the $p$-th primitive unit root and ${\rm Tr}$ being the trace map from $\mathbb F_q$ to $\mathbb F_p$. This extends Richman's theorem which treats the case of $f(x)$ being a monomial. Moreover, we show that the generating series $\sum_{s=1}^{\infty}N_{s}(f,a)x^s$ is a rational function in $x$ and also present its explicit expression in terms of the first $2d+1$ initial values $N_{1}(f,a), ..., N_{2d+1}(f,a)$, where $d$ is a positive integer no more than $q-1$. From this result, the theorems of Chowla-Cowles-Cowles and of Myerson can be derived.

math.NT↗

arXiv2021

On the number of zeros of diagonal quartic forms over finite fields

Let $\mathbb{F}_q$ be the finite field of $q=p^m\equiv 1\pmod 4$ elements with $p$ being an odd prime and $m$ being a positive integer. For $c, y \in\mathbb{F}_q$ with $y\in\mathbb{F}_q^*$ non-quartic, let $N_n(c)$ and $M_n(y)$ be the numbers of zeros of $x_1^4+...+x_n^4=c$ and $x_1^4+...+x_{n-1}^4+yx_n^4=0$, respectively. In 1979, Myerson used Gauss sum and exponential sum to show that the generating function $\sum_{n=1}^{\infty}N_n(0)x^n$ is a rational function in $x$ and presented its explicit expression. In this paper, we make use of the cyclotomic theory and exponential sums to show that the generating functions $\sum_{n=1}^{\infty}N_n(c)x^n$ and $\sum_{n=1}^{\infty}M_{n+1}(y)x^n$ are rational functions in $x$. We also obtain the explicit expressions of these generating functions. Our result extends Myerson's theorem gotten in 1979.

math.NT↗

arXiv2021

Sums of polynomial-type exceptional units modulo $n$

Let $f(x)\in\mathbb{Z}[x]$ be a nonconstant polynomial. Let $n, k$ and $c$ be integers such that $n\ge 1$ and $k\ge 2$. An integer $a$ is called an $f$-exunit in the ring $\mathbb{Z}_n$ of residue classes modulo $n$ if $\gcd(f(a),n)=1$. In this paper, we use the principle of cross-classification to derive an explicit formula for the number ${\mathcal N}_{k,f,c}(n)$ of solutions $(x_1,...,x_k)$ of the congruence $x_1+...+x_k\equiv c\pmod n$ with all $x_i$ being $f$-exunits in the ring $\mathbb{Z}_n$. This extends a recent result of Anand {\it et al.} [On a question of $f$-exunits in $\mathbb{Z}/{n\mathbb{Z}}$, {\it Arch. Math. (Basel)} {\bf 116} (2021), 403-409]. We derive a more explicit formula for ${\mathcal N}_{k,f,c}(n)$ when $f(x)$ is linear or quadratic.

math.NT↗

arXiv2018

A certain reciprocal power sum is never an integer

By $(\mathbb{Z}^+)^{\infty}$ we denote the set of all the infinite sequences $\mathcal{S}=\{s_i\}_{i=1}^{\infty}$ of positive integers (note that all the $s_i$ are not necessarily distinct and not necessarily monotonic). Let $f(x)$ be a polynomial of nonnegative integer coefficients. Let $\mathcal{S}_n:=\{s_1, ..., s_n\}$ and $H_f(\mathcal{S}_n):=\sum_{k=1}^{n}\frac{1}{f(k)^{s_{k}}}$. When $f(x)$ is linear, Feng, Hong, Jiang and Yin proved in [A generalization of a theorem of Nagell, Acta Math. Hungari, in press] that for any infinite sequence $\mathcal{S}$ of positive integers, $H_f(\mathcal{S}_n)$ is never an integer if $n\ge 2$. Now let deg$f(x)\ge 2$. Clearly, $0 0$, there are positive integers $n_1$ and $n_2$ and infinite sequences $\mathcal{S}^{(1)}$ and $\mathcal{S}^{(2)}$ of positive integers such that $1-\varepsilon<H_f(\mathcal{S}^{(1)}_{n_1})<1$ and $1<H_f(\mathcal{S}^{(2)}_{n_2})<1+\varepsilon$.

math.NT↗