Demand-driven volume rendering of terascale EM data
Bibliographic record
Abstract
In neuroscience, a very promising bottom-up approach to understanding how the brain works is built on acquiring and analyzing electron microscopy (EM) scans of brain tissue, an area known as Connectomics. This results in volume data of extremely high resolution of 3--5nm per pixel and 25--50nm slice thickness, overall leading to data sizes of many terabytes [Jeong et al. 2010]. To support the work of neurobiologists, interactive exploration and analysis of such volumes requires novel visual computing systems because the requirements differ from those of current systems in several key aspects. In this talk, we describe the system that we are working on to enable neuroscientists to interactively roam terascale EM volumes and support their analysis. A major design principle was to avoid the standard approach of pre-computing a 3D multi-resolution hierarchy such as an octree. Data acquisition proceeds from 2D image tile to 2D image tile, where not only the slices along the z axis are scanned independently, but each slice is itself acquired as many smaller image tiles. These images tiles need to be aligned and stitched, and neurobiologists also want to be able to combine different resolutions used for scanning different regions, without re-sampling everything to a single global resolution. Therefore, we focus on working directly with a stream of individual 2D image tiles, instead of a 3D volume that usually is assumed to exist in its entirety for visualization. We perform interactive volume rendering of a "virtual" volume, where the corresponding physical storage is only represented and populated in a sparse manner with 2D instead of 3D image data on the fly during rendering. Furthermore, these 2D image tiles can be of different resolution, scale, and orientation.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.003 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.003 | 0.001 |
| Open science | 0.002 | 0.002 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.008 | 0.002 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".