Abstract PR-06: Utilizing biological domain knowledge and machine learning methods to improve cellular segmentation on multiplex fluorescence and imaging mass cytometry datasets improves the quality of single-cell data obtained
Bibliographic record
Abstract
Abstract Currently, many existing multi-modality image analysis methods rely heavily on manual annotations for the extraction of quantitative readouts. This limits both the analytical throughput and the complexity to which one can analyze the data – for every hour of imaging time, several additional hours of tedious manual labor are require to extract even rudimentary metrics. Recent advances in computer vision-based techniques, including the use of machine and deep learning methods, have shown tremendous promise for solving previously intractable image segmentation and classification problems. This presents an opportunity for the development of image analysis pipelines that utilize these technical advances to solve known biomedical imaging challenges. We have developed a series of image analysis pipelines for multiplex image segmentation that draws from several interdisciplinary collaborations between biological scientists, pathologists, and biomedical image analysis specialists. For example, we have recently utilized biomedical domain knowledge to develop a customized cellular segmentation methodology for Imaging Mass Cytometry, to identify invading immune cells of various biomarker-identified types, and measure their density and proximity to blood vessels (a marker of active invasion) in multiple sclerosis lesions in the brain. A robust image segmentation and classification pipeline permits us to move these complex datasets from the format of “pixels embedded in spatial coordinates” towards the format of a “single-cell proteomic” dataset, that also permits spatial relationships between markers or tissue regions to be queried. Likewise, we have illustrated a number of methods for single-cell and per-vessel analysis of hypoxia and proliferation gradients within solid tumor tissue sections. Development of single-cell “tissue cytometric” methods permits the in-depth study of spatial relationships that would be difficult or impossible to quantify with more rudimentary whole-tissue analysis approaches. The spatial relationships between tissue components are tightly integrated with tissue metabolism and biomedical transport phenomena, so the study of these relationships permits us to better model tissue physiology in silico, through more accurate measurement of relevant physiological parameters. Clinical validation and deployment of analytical methodologies developed in the laboratory, along with robust methods for quality control and validation of the analytical outputs, allows us to move these promising research tools towards the ultimate goal of clinically approved diagnostic algorithms and medical devices. Such efforts hold the promise to deliver a profound positive impact on the healthcare system, reducing tedious manual steps like counting cells or reviewing scans, through an optimal combination of the advantages of validated automated methods with clinical wisdom and experience. Citation Format: Trevor D. McKee, Mark Zaidi, Veronica Cojocari. Utilizing biological domain knowledge and machine learning methods to improve cellular segmentation on multiplex fluorescence and imaging mass cytometry datasets improves the quality of single-cell data obtained [abstract]. In: Proceedings of the AACR Virtual Special Conference on Artificial Intelligence, Diagnosis, and Imaging; 2021 Jan 13-14. Philadelphia (PA): AACR; Clin Cancer Res 2021;27(5_Suppl):Abstract nr PR-06.
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 | 0.003 | 0.004 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.002 | 0.002 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.002 | 0.002 |
| Insufficient payload (model declined to judge) | 0.005 | 0.004 |
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".