Unpaired multi-modal training and single-modal testing for detecting signs of endometriosis
Bibliographic record
Abstract
Endometriosis is a serious multifocal condition that can involve various pelvic structures, with Pouch of Douglas (POD) obliteration being a significant clinical indicator for diagnosis. To circumvent the need for invasive diagnostic procedures like laparoscopy, research has increasingly focused on imaging-based methods such as transvaginal ultrasound (TVUS) and magnetic resonance imaging (MRI). The limited diagnostic accuracy achieved through manual interpretation of these imaging techniques has driven the development of automated classifiers that can effectively utilize both modalities. However, patients often undergo only one of these two examinations, resulting in unpaired data for training and testing POD obliteration classifiers, where TVUS models tend to be more accurate than MRI models, but TVUS scanning are more operator dependent. This prompts a crucial question: Can a model be trained with unpaired TVUS and MRI data to enhance the performance of a model exclusively trained with MRI, while maintaining the high accuracy of the model individually trained with TVUS? In this paper we aim to answer this question by proposing a novel multi-modal POD obliteration classifier that is trained with unpaired TVUS and MRI data and tested using either MRI or TVUS data. Our method is the first POD obliteration classifier that can flexibly take either the TVUS or MRI data, where the model automatically focuses on the uterus region within MRI data, eliminating the need for any manual intervention. Experiments conducted on our endometriosis dataset show that our method significantly improves POD obliteration classification accuracy using MRI from AUC=0.4755 (single-modal training and testing, without automatically focusing on the uterus region) to 0.8023 (unpaired multi-modal training and single modality MRI testing, with automatic uterus region detection), while maintaining the accuracy using TVUS with AUC=0.8921 (single modality TVUS testing using either an unpaired multi-modal training or a single-modality training).
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 | 0.009 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.000 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.004 | 0.001 |
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".