Abstract A13: Quantitative pathology toolbox: Improvement in prediction of progression risk for oral premalignant lesions using both interactive and automated image analysis
Bibliographic record
Abstract
Abstract For the promise of Quantitative Pathology (QP) to be clinically accepted it must provide new information that is clinically usable (and hence actionable). This must be accomplished in a fashion that is not onerous (expensive/time-consuming/require specialized training) to the pathologist relative to the value of the information provided. As an example a major barrier to oral cancer prevention is the inability to predict progression risk for oral premalignant lesions by conventional pathology alone that can be addressed by QP. We present two approaches for the quantitative analysis of FFPE sectioned oral tissue. In one approach the user selects the area of epithelium to be analyzed and visually filters the cells to be analyzed, the other is an automated approach in which the user only circles the epithelium to be analyzed. In both approaches the cell nuclei are automatically segmented, 110 features per nuclei calculated and used to determine how normal or cancer like the nucleus is (and in the fully automated approach if the objects are single intact nucleus or not), then the distribution of nuclei values within the area of interest is used to generate a Quantitative Pathology Scores (QPS) for the tissue. These tissue measures were used alone or in combination with other markers to perform risk assessment in patients from a very large oral cancer prediction longitudinal study. Also the scores can be combined with other risk markers such as Loss of Heterozygosity (LOH) analysis to improve risk stratification. A combination of LOH based predictors and QPS thresholds were trained to refine three previously validated LOH defined- risk groups. The combined model defined a low, a medium and a high risk of progression to cancer categories. For the 104 low risk cases so classified, 98.1% do not progress to cancer (used to define a relative risk [RR] of 1). In contrast, 15% of the 106 classified medium risk cases (RR= 7.85) and 65% of the 26 high-risk cases (RR = 34) progress. This is a substantial improvement over just the LOH based classification and significantly better than dysplasia grade for risk prediction. In a validation set of 43 mild to moderate dysplasia cases with long term follow-up, 100% of the 23 cases classified as low risk by the combined algorithm did not progress, 43% of the 7 cases classified as medium risk by the combined algorithm progressed and 92.3% (12 out 13) of the cases classified as high risk by the combined algorithm progressed. These validation results strongly support the combination of these approaches for facilitating risk prediction and improving patient management. This combined risk model is also a suitable intermediate endpoint biomarker of transformation risk for oral tissue and is being used in multicenter (8) Canadian Optically guides Oral Cancer Surgical Trial as part of the quantitative evaluation of surgical margin tissue. This work was funded by the NIDCR, NIH and by the TFRI. Note: This abstract was withdrawn after the Proceedings were printed and, therefore, was not presented at the conference. Citation Format: Calum MacAulay, Miriam Rosin, Lewei Zhang, Catherine Poh, Michele Williams, Martial Guillaud. Quantitative pathology toolbox: Improvement in prediction of progression risk for oral premalignant lesions using both interactive and automated image analysis. [abstract]. In: Proceedings of the Thirteenth Annual AACR International Conference on Frontiers in Cancer Prevention Research; 2014 Sep 27-Oct 1; New Orleans, LA. Philadelphia (PA): AACR; Can Prev Res 2015;8(10 Suppl): Abstract nr A13.
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 distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.003 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
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 teacher head, 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".