Predictive models of response to neoadjuvant chemotherapy in muscle-invasive bladder cancer using nuclear morphology and tissue architecture

• Published in: Cell reports. Medicine 2021-09, Vol.2 (9), p.100382, Article 100382
• Main Authors: Mi, Haoyang, Bivalacqua, Trinity J., Kates, Max, Seiler, Roland, Black, Peter C., Popel, Aleksander S., Baras, Alexander S.
• Format: Article
• Language: English
• Subjects: Aged; Aged, 80 and over; bladder cancer; Cell Nucleus - pathology; chemotherapy; Cohort Studies; digital pathology; Female; Humans; image processing; Image Processing, Computer-Assisted; Machine Learning; Male; Middle Aged; Models, Biological; Muscles - pathology; neoadjuvant; Neoadjuvant Therapy; Neoplasm Invasiveness; nucleus morphology; predictive biomarkers; Survival Analysis; tissue architecture; Tumor Microenvironment; Urinary Bladder Neoplasms - drug therapy; Urinary Bladder Neoplasms - pathology
• ISSN: 2666-3791; 2666-3791
• DOI: 10.1016/j.xcrm.2021.100382

Characterizing likelihood of response to neoadjuvant chemotherapy (NAC) in muscle-invasive bladder cancer (MIBC) is an important yet unmet challenge. In this study, a machine-learning framework is developed using imaging of biopsy pathology specimens to generate models of likelihood of NAC response. Developed using cross-validation (evaluable N = 66) and an independent validation cohort (evaluable N = 56), our models achieve promising results (65%–73% accuracy). Interestingly, one model—using features derived from hematoxylin and eosin (H&E)-stained tissues in conjunction with clinico-demographic features—is able to stratify the cohort into likely responders in cross-validation and the validation cohort (response rate of 65% for predicted responder compared with the 41% baseline response rate in the validation cohort). The results suggest that computational approaches applied to routine pathology specimens of MIBC can capture differences between responders and non-responders to NAC and should therefore be considered in the future design of precision oncology for MIBC. [Display omitted] Using imaging of pathology samples to predict chemotherapy response in bladder cancerMulti-modal integration of cell nuclear and tissue architectural featuresModels using H&E images and basic clinical features able to enrich for respondersPredictive features suggest response-modulating factors in tumor microenvironment Using multi-modal machine-learning leveraging features from digital pathology, Mi et al. develop models to predict response to chemotherapy in muscle-invasive bladder cancer. Models using handcrafted features derived from conventional H&E TMAs in conjunction with basic clinico-demographic features significantly stratify likelihood of response in both discovery and independent validation cohorts.