← Back to Research Papers

Advancing fair and explainable machine learning for neuroimaging dementia pattern classification in multi-racial and multi-ethnic populations.

Authors: Ho NH, Charisis S, Honnorat N, Brandigampala SR, Wang D, Heckbert SR, Fox PT, Martinez D, Wang DH, Hughes TM, Archer DB, Hohman TJ, Seshadri S, Davatzikos C, Habes M
Journal: Nature communications
mental health psychology open access

Abstract

Histopathologic assessment of the presence of tumor budding (TB) yields a promising biomarker for colorectal cancer (CRC). TB is defined as isolated single cells or small clusters of up to four tumor cells located at the invasive tumor front. These biomarkers possess independent prognostic information, in addition to TNM staging of CRC. TB is of particular importance in early CRC, as it is frequently encountered in population screening programs. The assessment is necessary for improved decision making on therapeutic strategies and for prevention of possible undertreatment and overtreatment. Although TB assessment is recognized in the WHO classification and international clinical guidelines, clinical adoption is hampered as human visual scoring remains labor intensive and subjective, resulting in a moderate interobserver agreement in hematoxylin and eosin slides.