Guiding the classification of hepatocellular carcinoma on 3D CT-scans using deep and handcrafted radiological features

E. Sarfati, A. Bône, M-M. Rohé, C. Aubé, M. Ronot, P. Gori, I. Bloch · 2025

Read Guiding the classification of hepatocellular carcinoma on 3D CT-scans using deep and handcrafted radiological features by E. Sarfati, A. Bône, M-M. Rohé, C. Aubé, M. Ronot, P. Gori, I. Bloch (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2501-08097.txt.

Description

Sarfati and co-authors guide a computer diagnosis of liver cancer from three-dimensional CT scans, in a 2025 paper. This cancer is the most common primary liver tumour worldwide. Biopsy is the gold standard, but in practice radiologists read scans against a standard protocol that weighs five features in a decision tree. The authors aim to cut the variation between readers with an automatic method.

Plain deep learning, they show, reads these scans poorly on their hard data set. Their two-step approach instead mimics the radiologists protocol: it joins learned features with hand-crafted ones that match the criteria doctors use. Accuracy rose by six to eighteen points over deep learning baselines of several designs. In clinical validation, the method beat non-expert radiologists and matched experts. The authors present it as decision support grounded in how diagnosis is already done.

Catalog record

CREATORE. Sarfati, A. Bône, M-M. Rohé, C. Aubé, M. Ronot, P. Gori, I. Bloch
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE22 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2501-08097

Browse the full library · About the Tower of Babel