OSCR

Probing the content of semantic representations in body-selective regions.

Code ↔ Paper

25 matches between paragraphs of the paper and lines of its authors' code, computed by the harvester (lexical-v1). Click a colored paragraph or line to see its counterpart.

The 25 matches
  1. [1] § Materials and Methods › Co-occurrence analysis › Encoding models based on text descriptions of natural scene images ↔ scripts/train_encmodel.py, lines 1–47 · score 0.94 · fractional ridge regression, evaluate prediction accuracy, Pearson correlation, cross validation, caption embeddings, encoding model
  2. [2] § Materials and Methods › Co-occurrence analysis › Defining a co-occurrence matrix from captions ↔ src/interpret_semrep/co_occurrence/make_matrix.py, lines 157–281 · score 0.93 · co occurrence matrix, MS COCO, superordinate categories, baby, baseball, bat
  3. [3] § Materials and Methods › Co-occurrence analysis › Encoding models based on text descriptions of natural scene images ↔ src/interpret_semrep/encoding/train_model.py, lines 15–64 · score 0.91 · fractional ridge regression, weights shrunk, ridge regression models, 0.1–1, cross validation, folds
  4. [4] § Materials and Methods › Correlation analysis using body-related image features ↔ src/interpret_semrep/correlation_analysis/compute_correlation.py, lines 378–489 · score 0.88 · Benjamini Hochberg, discovery rate, negative correlations, correlation coefficient, brain responses, permuted
  5. [5] § Materials and Methods › Co-occurrence analysis › Defining a co-occurrence matrix from captions ↔ src/interpret_semrep/co_occurrence/perform_nmf.py, lines 38–173 · score 0.84 · Bayesian information criterion, co occurrence matrices, coefficient matrices, matrix factorization, reconstructing, triangular
  6. [6] § Materials and Methods › Large-scale fMRI data ↔ src/interpret_semrep/nsd/get_roibeta.py, lines 47–184 · score 0.83 · noise ceiling signal, fithrf_GLMdenoise_RR, noise ratio, space, scored, NCSNR
  7. [7] § Materials and Methods › Large-scale fMRI data ↔ src/interpret_semrep/nsd/load_nsddata.py, lines 64–119 · score 0.80 · noise ceiling signal, fithrf_GLMdenoise_RR, noise ratio, scored, NCSNR, beta
  8. [8] § Results › Semantic representations in the body-selective regions are composed of multiple body-related features ↔ src/interpret_semrep/utils/stats.py, lines 59–140 · score 0.75 · Wilcoxon signed rank, pairwise comparisons, FDR correction, body selective, RMS, MST
  9. [9] § Materials and Methods › Co-occurrence analysis › Defining a co-occurrence matrix from captions ↔ src/interpret_semrep/co_occurrence/perform_nmf.py, lines 38–173 · score 0.74 · log likelihood, Poisson distribution, co occurrence matrices, components, captions
  10. [10] § Materials and Methods › Co-occurrence analysis › Defining a co-occurrence matrix from captions ↔ src/interpret_semrep/co_occurrence/perform_nmf.py, lines 177–234 · score 0.73 · co occurrence matrix, appliance, electronic, food, furniture, indoor
  11. [11] § Materials and Methods › Co-occurrence analysis › Defining a co-occurrence matrix from captions ↔ src/interpret_semrep/co_occurrence/make_matrix.py, lines 157–281 · score 0.72 · multiple co occurrence, co occurrence matrices, superordinate categories, nouns, NSD image, trained
  12. [12] § Materials and Methods › Correlation analysis using body-related image features ↔ src/interpret_semrep/variance_partition/compute_variance.py, lines 396–487 · score 0.70 · Benjamini Hochberg, discovery rate, permuted, NSD subject, FDR, permutation
  13. [13] § Materials and Methods › Variance partitioning ↔ src/interpret_semrep/variance_partition/compute_variance.py, lines 155–193 · score 0.66 · linear regression, variance partitioning, implied motion, fold, fitted, R2
  14. [14] § Results › Data-driven analysis of object co-occurrence revealed object pairs associated with the EBA response ↔ src/interpret_semrep/co_occurrence/perform_nmf.py, lines 177–234 · score 0.66 · co occurrence matrix, appliance, electronic, food, furniture, indoor
  15. [15] § Results › Semantic representations in the body-selective regions are composed of multiple body-related features ↔ src/interpret_semrep/utils/stats.py, lines 59–140 · score 0.64 · Wilcoxon signed rank, pairwise comparisons, FDR corrected, RMS, MST, MT
  16. [16] § Materials and Methods › Co-occurrence analysis › Defining a co-occurrence matrix from captions ↔ src/interpret_semrep/co_occurrence/make_matrix.py, lines 1–34 · score 0.63 · co occurrence matrices, matrix factorization, matrix represents, magnitude, captions
  17. [17] § Materials and Methods › Variance partitioning ↔ src/interpret_semrep/encoding/train_model.py, lines 15–64 · score 0.63 · fold cross validation, regression model, fitted, trained, coefficient, prediction
  18. [18] § Results › Variance partitioning identified the largest functional cluster uniquely explained by implied motion ↔ src/interpret_semrep/variance_partition/compute_variance.py, lines 155–193 · score 0.59 · linear regression, variance partitioning, implied motion, fold, fitted, maps
  19. [19] § Results › Variance partitioning identified the largest functional cluster uniquely explained by implied motion ↔ scripts/make_figures.py, lines 1–51 · score 0.58 · Cortical surface maps, Venn diagrams, variance partition, vertex, implied, motion
  20. [20] § Results › Variance partitioning identified the largest functional cluster uniquely explained by implied motion ↔ src/interpret_semrep/surfmap/make_surfmaps.py, lines 1–18 · score 0.57 · Venn diagrams, cortical maps, variance partition
  21. [21] § Materials and Methods › Variance partitioning ↔ scripts/train_encmodel.py, lines 1–47 · score 0.57 · prediction accuracy, cross validated, regression, threshold, models, vertex
  22. [22] § Materials and Methods › Implied motion rating task › Procedure ↔ src/interpret_semrep/rating/rating.py, lines 21–64 · score 0.56 · passive viewing session, blocks, twice, implied motion, animal, vehicle
  23. [23] § Materials and Methods › Variance partitioning ↔ src/interpret_semrep/variance_partition/compute_variance.py, lines 1–21 · score 0.55 · cross validated, FDR corrected, R2, partitioning, Variance, permutation
  24. [24] § Results › Data-driven analysis of object co-occurrence revealed object pairs associated with the EBA response ↔ src/interpret_semrep/co_occurrence/make_matrix.py, lines 38–81 · score 0.51 · trained encoding models, predicted responses, co occurrence, matrix, embeddings, captions
  25. [25] § Results › Variance partitioning identified the largest functional cluster uniquely explained by implied motion ↔ scripts/make_figures.py, lines 1–51 · score 0.50 · Venn diagrams, variance partition, implied motion, cortical, maps, vertex

Paper

Loaded from Europe PMC by your browser, not stored by OSCR: doi.org · Europe PMC

The paper is loaded when this pane is shown.

The authors' code

Python · 281 lines · 13 KB · MIT · 4 matches

  1. """
  2. Build co-occurrence matrices for NSD captions using MS COCO categories.
  4. This module creates co-occurrence matrices from NSD image captions, where each matrix represents
  5. word category co-occurrences in a subset of images sorted by neural response strength.
  6. Matrices are organized by subject, ROI, and grouped by image response magnitude, suitable for downstream
  7. non-negative matrix factorization (NMF) analysis.
  9. Main functions:
  10. - load_nsd_pred_betas: Load or compute predicted neural responses from caption embeddings.
  11. - cooccur_matrix: Build a co-occurrence matrix from a set of captions.
  12. - multi_coocur_matrix_nsd: Create and save multiple co-occurrence matrices grouped by response strength.
  14. When you use the script the first time, you may need to download the following NLTK resources for tokenization and lemmatization:
  15. nltk.download('punkt')
  16. nltk.download('wordnet')
  17. nltk.download('wordnet_ic')
  18. nltk.download('stopwords')
  19. nltk.download('averaged_perceptron_tagger')
  20. """
  22. import nltk
  23. from nltk.stem import WordNetLemmatizer
  24. import numpy as np
  25. import os
  26. import pandas as pd
  27. import os
  28. import itertools
  30. from interpret_semrep.nsd.load_nsddata import nsd_trial_stimidx, load_nsd_betas, load_saved_captions, load_saved_mpnet_embeddings
  32. from interpret_semrep.paths import NSD_ROOT, RESULTS_ROOT
  33. ENC_WEIGHT_DIR = RESULTS_ROOT / 'encoding' / 'weights'
  34. RESULT_DIR = RESULTS_ROOT / 'co_occurrence'
  38. def load_nsd_pred_betas(subject, roi='EBA', which_embed=0):
  39. """
  40. Compute and load predicted beta values for a given image caption using trained encoding models based on language embeddings
  42. Parameters
  43. ----------
  44. subject : int
  45. NSD subject number (1-8), or 99 to use predicted betas averaged across all subjects
  46. roi : str
  47. ROI name
  48. which_embed : int or str
  49. Caption index to use for each image (0-4), or 'all' to use all five captions. The default value is 0, which corresponds to the first caption for each image.
  50. """
  52. if subject == 99: # if subject is 99, we compute predicted responses averaged across all subjects
  54. # load embeddings of captions for all 73000 NSD images
  55. embeds = load_saved_mpnet_embeddings(which_embed=which_embed)
  57. for sub in range(1, 9):
  58. # load the coefficients of fractional ridge regression (output from train_encoding_model.py)
  59. coef = np.load(ENC_WEIGHT_DIR / f'subj{sub:02d}_{roi}_coef.npy')
  61. # compute predicted beta values for all subjects and ROIs
  62. pred_betas_sub = np.dot(embeds, coef)
  64. if sub == 1:
  65. pred_betas = pred_betas_sub
  66. else:
  67. pred_betas += pred_betas_sub
  69. pred_betas /= 8 # average across all subjects
  71. else:
  72. # load the coefficients of fractional ridge regression (output from train_encoding_model.py)
  73. coef = np.load(ENC_WEIGHT_DIR / f'subj{subject:02d}_{roi}_coef.npy')
  75. # load embeddings of captions for all 73000 NSD images
  76. embeds = load_saved_mpnet_embeddings(which_embed=which_embed)
  78. # compute predicted beta values for the given NSD subject and ROI
  79. pred_betas = np.dot(embeds, coef)
  81. return pred_betas
  84. def cooccur_matrix(wordlist, labels, captions, pos_type, diag_occ=False):
  85. """
  86. Make a co-occurrence matrix based on hierarchical clustering results.
  88. Parameters
  89. ----------
  90. wordlist : np.ndarray
  91. A list of frequently-used words in the NSD captions, including 80 MS COCO categories and additional person categories.
  92. labels : np.ndarray
  93. MS COCO superordinate category labels for each word in wordlist.
  94. captions : list of str
  95. A list of captions to use for making the co-occurrence matrix.
  96. pos_type : str
  97. Part of speech type to use for lemmatization, either 'noun' or 'verb'.
  98. diag_occ : bool, optional
  99. Whether to record the number of occurrence for each label in the diagonal elements of the co-occurrence matrix.
  100. The default value is False, which means that we will not record the number of occurrence in the diagonal elements.
  101. """
  103. if pos_type == 'noun':
  104. pos_type2 = 'n'
  105. elif pos_type == 'verb':
  106. pos_type2 = 'v'
  108. n_labels = len(np.unique(labels))
  110. # load frequently-used words and clustering labels for each word
  112. # make a co-occurrence matrix
  113. cooccur_mat = np.zeros((n_labels, n_labels))
  114. # for each caption, find the frequently-used words and their clustering labels
  115. for caption in captions:
  117. # tokenize the caption
  118. tokens = nltk.word_tokenize(caption)
  120. # lemmatize the tokens
  121. lemmatizer = WordNetLemmatizer()
  122. lemmatized_words = [lemmatizer.lemmatize(word, pos=pos_type2) for word in tokens]
  124. # if a lemmatized token is included in the wordlist, add its label to a list
  125. cooccur_label = []
  126. for lemword in lemmatized_words:
  127. if lemword in wordlist:
  128. idx = np.where(wordlist == lemword)[0]
  129. if not int(labels[idx]) in cooccur_label:
  130. cooccur_label.append(int(labels[idx]))
  132. # Some COCO categories are a phrase (e.g., traffic light, street sign), not a single word
  133. # we need to consider the original caption to see if the phrase is included
  134. phrase_or_not = [i for i, word in enumerate(wordlist) if ' ' in word]
  135. phrases = wordlist[phrase_or_not]
  136. for phrase in phrases:
  137. if phrase in caption and not int(labels[np.where(wordlist == phrase)[0]]) in cooccur_label:
  138. cooccur_label.append(int(labels[np.where(wordlist == phrase)[0]]))
  140. # for all label pairs, increment the corresponding element in the co-occurrence matrix
  141. if len(cooccur_label) > 1:
  142. allpairs = list(itertools.combinations(cooccur_label, 2))
  143. for pair in allpairs:
  144. label1 = pair[0]
  145. label2 = pair[1]
  146. cooccur_mat[label1-1, label2-1] += 1
  147. cooccur_mat[label2-1, label1-1] += 1
  149. if diag_occ:
  150. # we also record the number of occurrence for each label in the diagonal elements
  151. for l in cooccur_label:
  152. cooccur_mat[l-1, l-1] += 1
  154. return cooccur_mat
  157. def multi_cooccur_matrix_nsd(subject, roi='EBA', group_size=1000, img_type='all', which_embed=0):
  158. """
  159. Make multiple co-occurrence matrices for NSD captions and save them as .npy files for subsequent non-negative matrix factorization (NMF).
  160. Each co-occurrence matrix is based on a subset of NSD images, sorted by their predicted or actual beta values in the given subject and ROI.
  161. Note that even though each NSD image has five captions from five different annotators,
  162. we use one of these captions (first one by default)
  164. Parameters
  165. ----------
  166. subject : int
  167. NSD subject number (1-8), or 99 to use predicted betas averaged across all subjects
  168. roi : str
  169. ROI name
  170. group_size : int
  171. Number of images to include in each co-occurrence matrix. The default value is 1000, which means that we will make 73 co-occurrence matrices for the 73000 NSD images.
  172. img_type : str, default='all'
  173. Which images to use for making co-occurrence matrices: 'all' uses all 73000 NSD images, 'shared' uses the 1000 special images that were viewed by all subjects, and 'unique' uses the images that were viewed uniquely by one of the subjects.
  174. When img_type is 'shared' or 'unique', the actual beta values are used to sort the images, because some of the 1000 shared images were not presented to all subjects twice or more, so we cannot compute predicted betas for these images for all subjects.
  175. which_embed : int or str
  176. Caption index to use for each image (0-4), or 'all' to use all five captions. The default value is 0, which corresponds to the first caption for each image.
  177. """
  180. # a list of categories (MS COCO + additional person categories)
  181. wordlist = np.array([
  182. "person", "man", "woman", "boy", "girl", "child", "baby",
  183. "bicycle", "car", "motorcycle", "airplane", "bus", "train", "truck", "boat",
  184. "traffic light", "fire hydrant", "street sign", "stop sign", "parking meter", "bench",
  185. "bird", "cat", "dog", "horse", "sheep", "cow", "elephant", "bear", "zebra", "giraffe",
  186. "hat", "backpack", "umbrella", "shoe", "eye glasses", "handbag", "tie", "suitcase", "frisbee", "skis",
  187. "snowboard", "sports", "ball", "kite", "baseball", "bat", "glove", "skateboard", "surfboard",
  188. "tennis", "racket", "bottle", "plate", "wine glass", "cup", "fork", "knife", "spoon", "bowl",
  189. "banana", "apple", "sandwich", "orange", "broccoli", "carrot", "hot dog", "pizza",
  190. "donut", "cake", "chair", "couch", "potted plant", "bed", "mirror", "dining table", "window", "desk", "toilet",
  191. "door", "tv", "laptop", "mouse", "remote", "keyboard", "cell phone", "microwave", "oven",
  192. "toaster", "sink", "refrigerator", "blender", "book", "clock", "vase", "scissors", "teddy bear",
  193. "hair drier", "toothbrush", "hair brush"
  194. ])
  196. labels = np.array([1] * 7 + [2] * 8 + [3] * 6 + [4] * 10 + [5] * 8 + [6] * 13 + [7] * 8 + [8] * 10 + [9] * 10 + [10] * 6 + [
  197. 11] * 6 + [12] * 8)
  198. # each number denotes MS COCO superordinate category
  200. # number of class labels
  201. n_labels = len(np.unique(labels))
  203. captions = load_saved_captions(which_embed=which_embed)
  205. results_dir = RESULT_DIR / f'subj{subject:02d}'
  206. os.makedirs(results_dir, exist_ok=True)
  208. if subject == 99 and img_type not in ['all', 'shared']:
  209. raise ValueError("subject=99 is only supported with img_type='all' or img_type='shared'.")
  211. def _mean_img_response_for_subject(sub, stim_indices):
  212. # Compute mean ROI response for each image index across repeated trials.
  213. beta = load_nsd_betas(subject=sub, roi=roi, thres_ncsnr=0.2, thres_tval=1)
  214. trial_stim_index = np.array(nsd_trial_stimidx(sub))
  215. beta_mean = np.mean(beta, axis=1)
  217. beta_img = np.full(len(stim_indices), np.nan)
  218. for i, idx in enumerate(stim_indices):
  219. img_trials = beta_mean[trial_stim_index == idx]
  220. if img_trials.size > 0:
  221. beta_img[i] = np.mean(img_trials)
  223. return beta_img
  225. if img_type == 'all':
  226. predbetas = load_nsd_pred_betas(subject=subject, roi=roi, which_embed=which_embed)
  227. unique_stim_index = np.arange(0, 73000)
  229. sort_indices = np.argsort(predbetas)[::-1]
  231. elif img_type == 'shared':
  232. tsv_path = NSD_ROOT / 'nsddata' / 'stimuli' / 'nsd' / 'shared1000.tsv' # the file containing the numbers of the 1000 shared images that are used as test set for all subjects
  233. spimg_ind = pd.read_csv(tsv_path, sep='\t', header=None).to_numpy().flatten() # indices of special images
  234. spimg_ind = spimg_ind - 1 # indices of special images in the range of 0~72999
  235. unique_stim_index = np.array(spimg_ind)
  237. if subject == 99:
  238. # For virtual subject 99, average shared-image responses across all 8 subjects.
  239. beta_img_by_sub = []
  240. for sub in range(1, 9):
  241. beta_img_by_sub.append(_mean_img_response_for_subject(sub, unique_stim_index))
  242. beta_img = np.nanmean(np.vstack(beta_img_by_sub), axis=0) # for some shared images, some subjects do not have valid beta values because these images were not presented to these subjects, so we take the average across valid subjects ignoring nan values.
  243. else:
  244. beta_img = _mean_img_response_for_subject(subject, unique_stim_index)
  246. sort_indices = np.argsort(np.nan_to_num(beta_img, nan=-np.inf))[::-1]
  248. elif img_type == 'unique':
  249. # load betas in roi for each trial
  250. beta = load_nsd_betas(subject=subject, roi=roi, thres_ncsnr=0.2, thres_tval=1)
  251. trial_stim_index = nsd_trial_stimidx(subject) # which image was presented in each trial
  252. trial_stim_index = np.array(trial_stim_index)
  254. unique_stim_index = np.unique(trial_stim_index)
  255. unique_stim_index = sorted(unique_stim_index)
  256. unique_stim_index = np.array(unique_stim_index)
  258. beta_mean = np.mean(beta, axis=1)
  260. # compute mean response across repetitions for each image
  261. beta_img = np.zeros(len(unique_stim_index))
  262. for i, idx in enumerate(unique_stim_index):
  263. beta_img[i] = np.mean(beta_mean[trial_stim_index == idx])
  265. sort_indices = np.argsort(beta_img)[::-1]
  267. n_imgs = len(unique_stim_index)
  268. cooccur_mat_all = np.zeros((n_labels, n_labels, int(n_imgs/group_size)))
  270. for w in range(int(n_imgs/group_size)):
  271. sub_indices = unique_stim_index[sort_indices[w * group_size:+ (w + 1) * group_size]]
  272. sub_captions = captions[sub_indices]
  273. cmat = cooccur_matrix(wordlist, labels, sub_captions, pos_type='noun')
  274. cooccur_mat_all[:, :, w] = cmat
  276. if img_type == 'all':
  277. np.save(results_dir / f'coocmat_nsdcaps_subj{subject:02d}_{roi}_groupsize{group_size}_nlabel{n_labels}.npy', cooccur_mat_all)
  278. elif img_type in 'shared':
  279. np.save(results_dir / f'coocmat_nsdcaps_shared_subj{subject:02d}_{roi}_groupsize{group_size}_nlabel{n_labels}.npy', cooccur_mat_all)
  280. elif img_type in 'unique':
  281. np.save(results_dir / f'coocmat_nsdcaps_unique_subj{subject:02d}_{roi}_groupsize{group_size}_nlabel{n_labels}.npy', cooccur_mat_all)

make_matrix.py at commit 4bbd7b4, under MIT · at the source

Overview

Authors: Ryuto Yashiro1,2,3, Masataka Sawayama1,4,5, Ayumu Yamashita1, Kaoru Amano1
ORCID iDs: Kaoru Amano
  1. Graduate School of Information Science and Technology, The University of Tokyo, Tokyo, Japan
  2. Department of Education and Psychology, Freie Universität Berlin, Berlin, Germany
  3. Institute of Cognitive Science, Universität Osnabrück, Osnabrück, Germany
  4. Graduate School of Information Science and Technology, Hokkaido University, Sapporo, Hokkaido, Japan
  5. Prometech CG Research, Tokyo, Japan
Institutions: Osnabrück University (Germany); The University of Tokyo (Japan); Freie Universität Berlin (Germany); Hokkaido University (Japan)
Journal: Imaging neuroscience (Cambridge, Mass.), volume 4, article IMAG.a.1309
Dates: received 16 February 2026; accepted 27 June 2026; published online 27 July 2026
Type: Research article · Language: English
License: CC BY
Identifiers: DOI 10.1162/imag.a.1309 · PMID 42524262 · PMCID PMC13409281 · OpenAlex W4413472814
Open access: diamond, a free copy (OpenAlex)
Preprint: doi:10.1101/2025.08.19.669677
Status: code verified
Categories: human (organism)
Methods: Statistics, Machine learning, Preprocessing, Connectivity, fMRI & imaging
Keywords: body-selective regions, encoding models, semantic representation, co-occurrence, implied motion
Topic: Visual perception and processing mechanisms (Cognitive Neuroscience, Neuroscience), according to OpenAlex
Funding: Japan Society for the Promotion of Science (23KJ0477)
Citations: not cited yet (Europe PMC); 88 references in the paper

Abstract

Recent advances in neural networks trained on natural language have revealed that category-selective regions encode complex semantics and contextual information of natural scenes in addition to object categories. However, the limited interpretability of embeddings derived from these models complicates the characterization of the aspects of natural scenes that contribute to such semantic representations. Here we addressed this question by developing an analysis of the relationship between object co-occurrence in large-scale natural scene captions and corresponding fMRI responses predicted by caption-based encoding models, inspired by the fact that the joint presence of multiple objects generally shapes the overall content of a scene. We performed this analysis on the extrastriate body area (EBA), which responds strongly to human body parts. We found that human bodies co-occurring with sports-related objects drive the strongest predicted responses in the EBA among all image categories, whereas those co-occurring with vehicles or accessories elicit strong but weaker predicted responses. The findings from the co-occurrence analysis helped identify three key body-related features that contribute to the semantic representation in the EBA and fusiform body area: human body motion speed implied in static images of natural scenes is a primary contributor, and the number of people and body size are secondary contributors. Our framework, integrating object co-occurrence with caption-based encoding models, offers an interpretable approach for understanding high-level visual representations underlying natural scene perception.

Reproduced under the paper's license (CC BY), from the paper cited above.

Repositories

Its files are read in the Code ↔ Paper reader above, with 25 matches between paragraphs and lines of code.

figshare 32348043

License: CC-BY-4.0
State: the link answers, verified on 27 September 2026
Evidence: files inventoried
Size: 2 files
Software Heritage: not checked
Found in: “Data and Code Availability”
Not found: README, license file, CITATION.cff, environment file, tests, continuous integration, documentation
Availability: 1 check, the latest on 27 September 2026: the link answers (HTTP 200)
  • 27 September 2026: the link answers (HTTP 200)
At the source: doi.org/10.6084/m9.figshare.32348043

amano-k-lab/interpret_semrep

License: MIT
State: the link answers, verified on 27 September 2026
Evidence: files inventoried
Commit: 4bbd7b43a8b7b29c285402d17c502f08b1bdc5fb, 12 August 2026
Languages: Python (26)
Size: 34 files, 26 scripts
Software Heritage: not archived
Found in: “Data and Code Availability”
Holds: README, license file, environment (pyproject.toml)
Not found: CITATION.cff, tests, continuous integration, documentation
Tools: NumPy (17 files), Matplotlib (7 files), NiBabel (6 files), SciPy (6 files), pandas (4 files), h5py (3 files), seaborn (3 files), statsmodels (3 files), OpenCV (2 files), pycortex (2 files), Pillow (1 file), PyTorch (1 file), scikit-learn (1 file)
Availability: 1 check, the latest on 27 September 2026: the link answers
  • 27 September 2026: the link answers
28 files
At the source: github.com/amano-k-lab/interpret_semrep

The paper's code and data availability statement is in the Data section.

Tracing map

Proposed by the machine: these links were found in the paper and verified at the source, without human review. The map will receive a Zenodo DOI once one of the paper's authors has validated it with their ORCID.

What the map holds:

  • 2 repositories of the authors' code, each at its verified commit, with its license and how the link was found in the paper;
  • 26 scripts, each with its path and the digest of its content;
  • 25 matches between paragraphs of the paper and lines of the code (method lexical-v1);
  • neither the text of the paper nor the code itself.

Its JSON (tracing-map.json) is deposited on Zenodo with its DOI once the map is validated.

Data

No dataset and no data link were found in the paper.

Data and Code Availability

The Natural Scenes Dataset is publicly available at http://naturalscenesdataset.org. The behavioral data and code used for the analyses reported here are available on Figshare at https://doi.org/10.6084/m9.figshare.32348043 and on GitHub at https://github.com/amano-k-lab/interpret_semrep.

Reproduced under the paper's license (CC BY), from the paper cited above.

Versions

The history of this record: each version stored by the harvester or made by a correction of its authors or of the maintainers of its code, and what changed in its facts. The texts of the paper (its abstract, its availability statements) are not part of it; versions that changed only those are not listed.

Version 2, 28 September 2026

  • Funding: added Japan Society for the Promotion of Science: 23KJ0477

Version 1, 27 September 2026: the first record

Recorded: type, language, journal, volume, pages, dates, 4 authors, 5 keywords, 80 references.

Cite

This paper

Yashiro, R., Sawayama, M., Yamashita, A., & Amano, K. (2026). Probing the content of semantic representations in body-selective regions. Imaging neuroscience (Cambridge, Mass.), 4, IMAG.a.1309. https://doi.org/10.1162/imag.a.1309

BibTeX

@article{yashiro2026probing,
author = {Yashiro, Ryuto and Sawayama, Masataka and Yamashita, Ayumu and Amano, Kaoru},
title = {{Probing the content of semantic representations in body-selective regions}},
journal = {Imaging neuroscience (Cambridge, Mass.)},
year = {2026},
month = jul,
volume = {4},
pages = {IMAG.a.1309},
publisher = {MIT Press},
issn = {2837-6056},
doi = {10.1162/imag.a.1309},
url = {https://doi.org/10.1162/imag.a.1309},
pmid = {42524262},
pmcid = {PMC13409281}
}

RIS

TY - JOUR
AU - Yashiro, Ryuto
AU - Sawayama, Masataka
AU - Yamashita, Ayumu
AU - Amano, Kaoru
TI - Probing the content of semantic representations in body-selective regions
T2 - Imaging neuroscience (Cambridge, Mass.)
J2 - Imaging Neurosci (Camb)
PY - 2026
DA - 2026/07/27
VL - 4
SP - IMAG.a.1309
SN - 2837-6056
PB - MIT Press
DO - 10.1162/imag.a.1309
UR - https://doi.org/10.1162/imag.a.1309
LA - en
ER -

CSL-JSON

{
"id": "10.1162/imag.a.1309",
"type": "article-journal",
"title": "Probing the content of semantic representations in body-selective regions",
"container-title": "Imaging neuroscience (Cambridge, Mass.)",
"author": [
{
"family": "Yashiro",
"given": "Ryuto"
},
{
"family": "Sawayama",
"given": "Masataka"
},
{
"family": "Yamashita",
"given": "Ayumu"
},
{
"family": "Amano",
"given": "Kaoru"
}
],
"container-title-short": "Imaging Neurosci (Camb)",
"volume": "4",
"page": "IMAG.a.1309",
"DOI": "10.1162/imag.a.1309",
"PMID": "42524262",
"PMCID": "PMC13409281",
"ISSN": "2837-6056",
"publisher": "MIT Press",
"URL": "https://doi.org/10.1162/imag.a.1309",
"language": "en",
"issued": {
"date-parts": [
[
2026,
7,
27
]
]
}
}

The tracing map gets a citation of its own once an author has validated it and it has a DOI.

Similar papers

The papers with a page that share the most with this one: the tools found in their code, their categories, datasets, cited references and authors, the rarest counting most.

[1] doi:10.1523/jneurosci.0038-26.2026 [code]
Multidimensional Feature Tuning in Category Selective Areas of Human Visual Cortex.
Journal: The Journal of neuroscience : the official journal of the Society for Neuroscience
In common: pycortex, h5py, Pillow, 9 other tools, 11 references
[2] doi:10.1038/s42003-026-10169-0 [code]
Shared representations in brains and models reveal a two-route cortical organization during scene perception.
Journal: Communications biology
In common: h5py, Pillow, statsmodels, 8 other tools, 10 references
[3] doi:10.1038/s41467-026-76098-y [code]
A single computational objective can produce specialization of streams in visual cortex.
Journal: Nature communications
In common: h5py, Pillow, statsmodels, 8 other tools, 8 references
[4] doi:10.1167/jov.26.5.7 [code]
Representations in vision and language converge in a shared, multidimensional space of perceived similarities.
Journal: Journal of vision
In common: pycortex, OpenCV, h5py, 10 other tools, 4 references
[5] doi:10.1038/s41467-026-73938-9 [code]
Higher visual areas act like domain-general filters with strong selectivity and functional specialization.
Journal: Nature communications
In common: Pillow, NiBabel, PyTorch, 4 other tools, 8 references
[6] doi:10.1038/s41597-026-07248-6 [code]
A large-scale fMRI dataset for vision-language semantic association.
Journal: Scientific data
In common: OpenCV, h5py, Pillow, 7 other tools, 6 references
[7] doi:10.7554/elife.107933 [code]
Modality-agnostic decoding of vision and language from fMRI.
Journal: eLife
In common: OpenCV, h5py, Pillow, 9 other tools, 4 references
[8] doi:10.1038/s41597-025-05174-7 [code]
A large-scale MEG and EEG dataset for object recognition in naturalistic scenes
Journal: n/a
In common: OpenCV, h5py, Pillow, 8 other tools, 4 references
[9] doi:10.1162/nol.a.271 [code]
Compositional Complexity in Text and Images.
Journal: Neurobiology of language (Cambridge, Mass.)
In common: OpenCV, Pillow, NiBabel, 7 other tools, 5 references
[10] doi:10.1162/imag.a.1256 [code]
Gamer in the scanner: Event-related analysis of fMRI activity during retro videogame play guided by automated annotations of game content.
Journal: Imaging neuroscience (Cambridge, Mass.)
In common: Pillow, statsmodels, NiBabel, 7 other tools, 4 references

Contribute

The authors of this paper can claim it, correct its record and validate its tracing map, and the maintainers of its code (its owner, or a public member of its organization) correct what it says of their repository; anyone signed in can ask for its removal. Every request goes to OSCR's own machine, which answers it; your account page follows them.

Sign in with ORCID to claim this paper as one of its authors, correct its record or validate its tracing map: when the paper's metadata lists your ORCID iD, you are recognized at once. Maintainers of its code: sign in with GitHub, then claim the repository on your account page.

Request its removal

To ask OSCR to remove this record, the copies of its authors' scripts or its tracing map, use the removal request page: signed in, you say who you are, what to remove and why, then review and confirm the request. Published rules decide every request (how).

Discussion, reproductions, activity

Discussion: questions and error reports about this paper and its code, from signed-in readers and its authors. It opens with sign-in.

Reproductions: reports from readers who ran the authors' code: what they reproduced, with which environment, commit and data. It opens with sign-in.

Activity: what happens around this paper: new versions of its record, its map's validation, discussions and reproductions. It opens with sign-in.