张春成

V2

2023/02/12阅读:25主题:默认主题

RDM与图像关系分析

RDM与图像关系分析

本文记录近期开发的一套前后端工具,它利用 RDM 矩阵进行图像关系分析,而 RDM 矩阵的来源为 fMRI 和 MEG 采集的神经影像数据。文中附录部分为 AI 补写的 RDM 矩阵的细节,虽然有点啰嗦但十分靠谱。


  • RDM与图像关系分析[1]
    • RDM 数据[2]
    • RDM 与图像关系[3]
  • 附录:Understanding Relative Dissimilar Matrix[4]
    • Introduction[5]
    • What is Relative Dissimilar Matrix[6]
    • How to Calculate Relative Dissimilar Matrix[7]
    • Uses of Relative Dissimilar Matrix[8]
    • Conclusion[9]

RDM 数据

由于 fMRI 和 MEG 的神经影像数据十分昂贵和有限,因此目前只有两组数据,它们的细节如下。

Untitled
Untitled

其中, MEG 数据的 RDM 矩阵大小为 代表 15 名被试和 20 个时间点, 代表图像数量,它包括 78images 和 92images 两组图像;而fMRI 数据与之类似,但只有 1 个时间点。另外,78images 图像为自然场景,其直方图特征较为复杂;92images 图像为分割后的纯粹物体,其直方图特征较为简单。

Untitled
Untitled
Untitled
Untitled

RDM 与图像关系

接下来,我们用 RDM 矩阵对图像之间的相似性进行度量。典型情况下,RDM 矩阵是对称方阵,它的每个元素 代表对应索引的内容 之间的距离,距离越大代表两个内容越不相似,反之距离越小代表两个内容越相似

如下图所示,它的左侧是在被试观看 78images 图像时,对较高级视觉皮层(用 late 代表)进行 MEG 扫描,通过得到的神经活动数据进行计算,而得到的 RDM 矩阵,其中颜色越偏蓝代表距离越小,从而表示两图越相似。可以看到在右下角存在较明显的蓝色区域,与周围明显不同,这代表这些图像具有“排他”的相似性,它们是人的面部图像。

Untitled
Untitled

使用另外 92images 图像集发现了同样的情况。而更有意思的是,由于这个图像集中有一张猴子的面部图像,因此猴子图像所在的列呈现出一条明显的蓝线,这进一步验证之前的假设,即神经活动相似性能够反映在其计算的 RDM 矩阵中。

Untitled
Untitled
Untitled
Untitled

附录:Understanding Relative Dissimilar Matrix

Introduction

Relative dissimilarity is a statistical measure used to compare the similarity or dissimilarity of two different sets of data. This measure is often used to compare the distance between two points in a data set, or to compare two sets of data. In this blog post, we will discuss how to understand relative dissimilar matrix, what it means and how it can be used.

What is Relative Dissimilar Matrix

A relative dissimilar matrix (RDM) is a matrix that describes the relative dissimilarity between points in a data set. It is based on the distance between each point in the set, and is used to compare the similarity or dissimilarity between sets of data. As an example, a RDM can be used to compare the distance between two cities, or the distance between two points on a graph.

How to Calculate Relative Dissimilar Matrix

The relative dissimilar matrix is calculated by taking the absolute differences between each point in the data set, and then normalizing the differences. This means that the values are scaled between 0 and 1, where 0 is the most similar point and 1 is the least similar. The RDM is then calculated by taking the square root of the differences and adding them together.

Uses of Relative Dissimilar Matrix

The relative dissimilar matrix can be used to compare the similarity or dissimilarity of sets of data. This can be useful for determining the degree of similarity between two points in a data set, or for comparing two sets of data. It can also be used to identify clusters of data points that are more or less similar.

Conclusion

In summary, relative dissimilar matrix is a statistical measure used to compare the similarity or dissimilarity of two different sets of data. It is based on the distance between each point in the set, and is used to compare the similarity or dissimilarity between sets of data. It can be used to compare the similarity or dissimilarity of sets of data, to identify clusters of data points, and to compare two points in a data set.

参考资料

[1]

RDM与图像关系分析: #rdm与图像关系分析

[2]

RDM 数据: #rdm-数据

[3]

RDM 与图像关系: #rdm-与图像关系

[4]

附录:Understanding Relative Dissimilar Matrix: #附录understanding-relative-dissimilar-matrix

[5]

Introduction: #introduction

[6]

What is Relative Dissimilar Matrix: #what-is-relative-dissimilar-matrix

[7]

How to Calculate Relative Dissimilar Matrix: #how-to-calculate-relative-dissimilar-matrix

[8]

Uses of Relative Dissimilar Matrix: #uses-of-relative-dissimilar-matrix

[9]

Conclusion: #conclusion

分类:

后端

标签:

后端

作者介绍

张春成
V2