Learning to count biological structures with raters’ uncertainty

Ciampi, Luca; Carrara, Fabio; Totaro, Valentino; Mazziotti, Raffaele; Lupori, Leonardo; Santiago, Carlos; Amato, Giuseppe; Pizzorusso, Tommaso; Gennaro, Claudio

doi:10.1016/j.media.2022.102500

Cited by 9 publications

(12 citation statements)

References 50 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Ourscoring models produced detection scores that strongly correlated with the number of raters that detected each object ( Figure S1 B, C). Overall, when tested on objects located by at least three raters, our models proved to be reliable in the detection of PNNs and PV cells (see Ciampi et al, 2022 and section Deep learning models for cell counting in Methods & Materials). We release the pre-trained four models used in this study (link) to allow performing predictions on new images or to fine-tune them based on different experimental setups.…”

Section: Resultsmentioning

confidence: 85%

“…The aim was to produce scores for each putative object that maximally correlate with the raters’ agreement. A detailed description of this method is available in Ciampi et al, 2022.…”

Section: Resultsmentioning

confidence: 99%

“…The aim was to produce scores for each putative object that maximally correlate with the raters' agreement. A detailed description of this method is available in Ciampi et al, 2022. In our multi-rater dataset, the average agreement (Jaccard index) between pairs of expert raters was 64% for PNNs and 72% for PV cells, demonstrating relevant individual differences in counting strategies (Figure S1A). Our scoring models produced detection scores that strongly correlated with the number of raters that detected each object (Figure S1B, C).…”

Section: Pnn and Pv Cells Quantification In The Mouse Brainmentioning

confidence: 94%

“…The deep learning models used in this work are based on a novel counting strategy described in Ciampi et al, 2022 specifically designed to account for the variability between experimenters when counting non-trivial, overlapping, or low-contrast objects like PNNs in histological preparations. Briefly, cell counting for both PNNs and PV cells was done through a two-step pipeline.…”

Section: Methodsmentioning

confidence: 99%

See 3 more Smart Citations

A Comprehensive Atlas of Perineuronal Net Distribution and Colocalization with Parvalbumin in the Adult Mouse Brain

Lupori

Totaro

Cornuti

et al. 2023

Preprint

Self Cite

View full text Add to dashboard Cite

Perineuronal nets (PNNs) surround specific neurons in the brain and are involved in various forms of plasticity and clinical conditions. However, our understanding of the PNN role in these phenomena is limited by the lack of highly quantitative maps of PNN distribution and association with specific cell types. Here, we present the first comprehensive atlas of PNN distribution (in Allen Brain Atlas coordinates) and colocalization with parvalbumin (PV) cells for over 600 regions of the adult mouse brain. Data analysis showed that PV expression is a good predictor of PNN aggregation. In the cortex, PNNs are dramatically enriched in layer 4 of all primary sensory areas in correlation with thalamocortical input density, and their distribution mirrors intracortical connectivity patterns. Gene expression analysis identified many PNN correlated genes. Strikingly, PNN anticorrelated transcripts were enriched in synaptic plasticity genes, generalizing PNN role as circuit stability factors. Overall, this atlas offers novel resources for understanding the organizational principles of the brain extracellular matrix.

show abstract

Section: Resultsmentioning

confidence: 85%

“…The aim was to produce scores for each putative object that maximally correlate with the raters’ agreement. A detailed description of this method is available in Ciampi et al, 2022.…”

Section: Resultsmentioning

confidence: 99%

Section: Pnn and Pv Cells Quantification In The Mouse Brainmentioning

confidence: 94%

Section: Methodsmentioning

confidence: 99%

See 2 more Smart Citations

A Comprehensive Atlas of Perineuronal Net Distribution and Colocalization with Parvalbumin in the Adult Mouse Brain

Lupori

Totaro

Cornuti

et al. 2023

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…In recent years, Computer Vision swerved toward Deep Learning (DL)-based models that learn from vast amounts of annotated data during the supervised learning phase. These models achieved astonishing results in several tasks that nowadays are considered basic, such as image classification, causing interest in addressing more complex domains such as object detection [1], image segmentation [2], visual object counting [3] [4] [5], people tracking [6], or even facial reconstruction [7] and video violence detection [8]. However, these more cumbersome tasks often also require more structured datasets that come with challenges concerning bias, privacy, and cost in terms of human effort for the annotation procedure.…”

Section: Introductionmentioning

confidence: 99%

CrowdSim2: An Open Synthetic Benchmark for Object Detectors

Foszner

Szczęsna

Ciampi

et al. 2023

Proceedings of the 18th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Application

View full text Add to dashboard Cite

Data scarcity has become one of the main obstacles to developing supervised models based on Artificial Intelligence in Computer Vision. Indeed, Deep Learning-based models systematically struggle when applied in new scenarios never seen during training and may not be adequately tested in non-ordinary yet crucial real-world situations. This paper presents and publicly releases CrowdSim2, a new synthetic collection of images suitable for people and vehicle detection gathered from a simulator based on the Unity graphical engine. It consists of thousands of images gathered from various synthetic scenarios resembling the real world, where we varied some factors of interest, such as the weather conditions and the number of objects in the scenes. The labels are automatically collected and consist of bounding boxes that precisely localize objects belonging to the two object classes, leaving out humans from the annotation pipeline. We exploited this new benchmark as a testing ground for some state-of-the-art detectors, showing that our simulated scenarios can be a valuable tool for measuring their performances in a controlled environment.

show abstract