Abstract PO2-02-14: Development of a multi-institutional, photograph-rich clinical dataset to test and validate a novel inflammatory breast cancer (IBC) scoring system
Bibliographic record
Abstract
Abstract Purpose: Susan G. Komen, The Inflammatory Breast Cancer (IBC) Research Foundation, and the Milburn Foundation convened a collaborative of patient advocates, clinicians, and researchers and proposed a novel quantitative scoring system for IBC diagnosis. The score includes timing of symptoms, extent of skin edema and erythema/other discoloration, breast and nipple asymmetry, lympho-vascular space invasion, nodal stage, and diffuse extent disease in the breast on breast imaging. Total scores were categorized (< 14 Not IBC, 14-24 Weak Possibility of IBC, 25-41 Strong Possibility of IBC, > 41 Definitely IBC). Herein, we developed a multi-institutional clinical dataset to retrospectively test and validate the proposed scoring system. Methods: IBC (N= 988) and non-IBC (N=322) cases were identified at two institutions with dedicated multi-disciplinary IBC programs, UT MD Anderson Cancer Center (MDA) and Dana-Farber Cancer Institute (DFCI). The DFCI cohort was diagnosed from 1986-2021 and the MDA cohort was diagnosed from 2010-2020. The non-IBC cohorts were identified as consecutive T4b and T4c cases in existing retrospective databases. Photographs were reviewed for available cases. All scoring variables were reviewed with data extractors and standard operating procedures (SOPs) developed for all ambiguous findings and language. A subset of scores were reviewed by IBC clinicians (FL, WW) and discrepancies discussed to inform the SOPs. Missing data imputation using three methods was examined. Area under the curve (AUC) - Receiver operator characteristic (ROC) curves and values, and sensitivity and specificity for the proposed score cut offs were examined. Results: Of 1320 patients currently in the analysis, 421 (31.9%) have missing data on > 1 characteristics and 47 (3.6%) are missing data on > 4. The rubric characteristic most commonly missing was “nipple abnormalities” accounting for 21.0% and 19.3% for IBC and non-IBC cohorts respectively. Unavailable bilateral breast photograph was the primary source of missing data. The highest scoring finding was present in the following percentage of IBC and non-IBC cases, respectively: timing of symptom onset < 3 months (74.1% vs 38.6%), any peau d’orange (61.2% vs 30.1%), clinically apparent enlargement or new asymmetry of breast size (87.8% vs 90.1%), complete or near complete involvement of the breast skin by erythema or discoloration (55.6% vs 25.6%), new nipple inversion (24.6% vs 9.0%), dermal lymphovascular emboli present (17.9% vs 1.8%), diffuse involvement of the breast parenchyma by imaging (72% vs 77.7%). True IBC cases were categorized as Definitely IBC, Strong Possibility, Weak Possibility, not IBC and unknown 19.7%, 49.1%, 0.4%, 0.1%, and 30.7%. True non-IBC were 0.6%, 51.8%, 9.9%, 2.1%, and 35.5%. Random imputation, lowest value imputation, and median value imputation were compared to impute missing data. AUC-ROC values for each method were similar, 0.80 - 0.84. Conclusion: We describe the development of the largest multi-institutional IBC clinical database to date. SOPs for the rubric characteristics were developed. The importance of bilateral photography to assess the score retrospectively is highlighted. The score is associated with a good AUC-ROC but still leaves much overlap among cases between 25 and 42. Further optimization of the current scoring system and association with outcomes are in progress. Citation Format: Filipa Lynce, Samuel Niman, Megumi Kai, Sean Ryan, Elizabeth Troll, Li Li, Kathy Miller, Reshma Jagsi, Ginny Mason, Beth Overmoyer, H. T. Carisa Le-Petross, Faina Nakhlis, Savitri Krishnamurthy, Beth Harrison, Susie X. Sun, Eren Yeh, Jennifer Bellon, Laura Warren, Michael Stauder, Meredith Regan, Wendy Woodward. Development of a multi-institutional, photograph-rich clinical dataset to test and validate a novel inflammatory breast cancer (IBC) scoring system [abstract]. In: Proceedings of the 2023 San Antonio Breast Cancer Symposium; 2023 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2024;84(9 Suppl):Abstract nr PO2-02-14.
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 | 0.005 | 0.016 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.002 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.002 | 0.002 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.003 | 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".