Back

Machine Learning-driven Histotype Diagnosis of Ovarian Carcinoma: Insights from the OCEAN AI Challenge

Asadi-Aghbolaghi, M.; Farahani, H.; Zhang, A.; Akbari, A.; Kim, S.; Chow, A.; Dane, S.; OCEAN Challenge Consortium, ; OTTA Consortium, ; G Huntsman, D.; Gilks, C. B.; Ramus, S.; Köbel, M.; N Karnezis, A.; Bashashati, A.

2024-04-23 oncology
10.1101/2024.04.19.24306099 medRxiv
Show abstract

Ovarian cancer poses a significant health burden as one of the deadliest malignancies affecting women globally. Histotype assignment of epithelial ovarian cancers can be challenging due to morphologic overlap, inter-observer variability, and the lack of ancillary diagnostic techniques in some areas of the world. Moreover, rare cancers can pose particular diagnostic difficulties because of a relative lack of familiarity with them, underscoring the necessity for robust diagnostic methodologies. The emergence of Artificial Intelligence (AI) has brought promising prospects to the realm of ovarian cancer diagnosis. While various studies have underscored AIs promise, its validation across multiple healthcare centers and hospitals has been limited. Inspired by innovations in medical imaging driven by public competitions, we initiated the Ovarian Cancer subtypE clAssification and outlier detectioN (OCEAN) challenge -- the most extensive histopathology competition to date.

Matching journals

The top 10 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.