September 2026
M	T	W	T	F	S	S
	1	2	3	4	5	6
7	8	9	10	11	12	13
14	15	16	17	18	19	20
21	22	23	24	25	26	27
28	29	30

Cressida Bradley: Navigating AI Challenges in Endometriosis Diagnosis

Cressida Bradley/LinkedIn

Mar 1, 2026, 16:35

Cressida Bradley: Navigating AI Challenges in Endometriosis Diagnosis

Cressida Bradley, Program Manager and Research Advisor at APNA, shared a post by Imagendo Aus on LinkedIn about a paper by Alison Deslandes et al. published in Human Reproduction:

“The recent disturbing allegations about endometriosis treatment in Australia have shone a light on the challenges of diagnosis and management and reignited the debate about gold standard approaches. While the path forward is complex, AI is likely to play a crucial role in future endometriosis diagnosis and management approaches. This excellent new Australian research article illustrates the inherent difficulties in finding suitable ‘ground truth’ for training and evaluating ML models when the real-world training data is multimodal, complex and noisy.”

Quoting Imagendo Aus’s post:

“Endometriosis diagnosis has never been simple – and building AI in this space has made that reality impossible to ignore.

Over the past five years, our team has been developing and refining AI models for endometriosis imaging. In doing so, we repeatedly encountered the same challenge: diagnostic ‘ground truth’ is not absolute.

Symptoms vary.

Imaging quality varies.

Surgical and histopathological confirmation have limitations.

When AI is trained on imperfect reference standards, it risks encoding those imperfections into its predictions. Our newly published paper in Human Reproduction is the culmination of five years of technical development, iteration, and critical reflection.

The goal is not to eliminate complexity. It is to harness it.

All diagnostic tests have limitations.

Understanding them is how we build better intelligence.

As AI becomes embedded in clinical imaging workflows, these conversations are no longer theoretical – they are foundational.”

Title: The problem with the ‘truth’: rethinking ground truth for artificial intelligence in endometriosis diagnosis

Authors: Alison Deslandes, Yuan Zhang, Mathew Leonardi, Hsiang-Ting Chen, Gustavo Carneiro, Jodie Avery, George Condous, Steven Knox, M Louise Hull, the IMAGENDO Team

Read the full article.

Cressida Bradley