Original title: MRI-based radiomics analysis and machine learning in the differentiation of endometriomas from hemorrhagic ovarian cysts
Armin Doostparast, Zahra Alami, Sajjad Sadeghpour, Maryam Ghandhari +2 · September 2026
Doctors often struggle to tell two types of non-cancerous ovarian cysts apart: endometriomas and hemorrhagic ovarian cysts. Because doctors treat them very differently, an incorrect guess causes problems. Standard MRI scans can look confusingly similar for both conditions. To solve this, researchers turned to machine learning, which uses computer programs that learn patterns from data. They pulled hundreds of invisible quantitative details, called radiomic features, directly from MRI scans of these cysts. These measurements capture subtle details about the cyst shape, patterns, and tissue texture. The computer analyzed these measurements to spot hidden differences human eyes miss. The system worked remarkably well. Out of several tested models, the top program accurately separated the two cysts ninety-six percent of the time. This approach could help doctors make confident, correct diagnoses without surgery, ensuring patients receive the right medical care.
Imagine you have two kinds of fruit that look almost identical from the outside, like a red apple and a red pear. A regular check just looks at color and shape, and often gets confused. Now suppose you have a device that measures the fruit's surface texture, internal density, and how it reflects light in tiny ways. It finds five specific measurements that always differ between the two fruits. In this paper, the MRI images are the fruit, the radiomics features are those fine measurements, and the machine learning model is the device that combines them to tell the cysts apart with high accuracy.
Endometriomas and hemorrhagic ovarian cysts look almost identical on standard MRI, but one often requires surgery while the other can be left alone. Misdiagnosis leads to unnecessary operations or delayed treatment. This method could help doctors avoid those mistakes without invasive procedures.
The abstract notes this is a single-center study with a relatively small sample (160 patients). It explicitly states that external validation is pending, meaning the results may not generalize to other hospitals or MRI machines. The manual segmentation of lesions is also a potential source of variability not addressed here.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
MRI-based radiomics analysis and machine learning in the differentiation of endometriomas from hemorrhagic ovarian cysts — Armin Doostparast, Zahra Alami, Sajjad Sadeghpour et al. — European journal of radiology open — doi:10.1016/j.ejro.2026.100818
Abstract licence: CC BY
Explanation generated by Synthesis from the abstract — it is not the original paper.