Journal article
A Novel Artificial Intelligence-based Parameterization Approach of the Stromal Landscape in Merkel Cell Carcinoma: a Multi-Institutional Study
Laboratory investigation, Vol.104(9), 102123
09/2024
DOI: 10.1016/j.labinv.2024.102123
PMID: 39147033
Abstract
Tumor-Stroma Ratio (TSR) has been recognized as a valuable prognostic indicator in various solid tumors. This study aimed to examine the clinicopathological relevance of TSR in Merkel cell carcinoma (MCC) using artificial intelligence (AI)-based parameterization of the stromal landscape and validate TSR scores generated by our AI model versus human-assessed.
112 MCC cases with Whole Slide Images (WSIs) were collected from four different institutions. WSIs were first partitioned into 128x128-pixel “mini-patches” then classified by a novel framework, termed Pre-TumOr And STroma (Pre-TOAST) and TOAST, whose output equaled the probability of the mini-patch representing tumor cells rather than stroma. Hierarchical random samplings of 50 mini-patches per region were performed throughout 50 regions per slide. TSR and Tumor-Stromal Landscape (TSL) parameters were estimated by the maximum-likelihood algorithm.
Receiver Operating Characteristic (ROC) curves showed the areas under the curve (AUCs) of Pre-TOAST in discriminating classed of interest (COI) including tumor cells, collagenous stroma, and lymphocytes from non-classes of interest (non-COI) including hemorrhage, space, and necrosis were 1.00. AUCs of TOAST in differentiating tumor cells from related stroma were 0.93. MCC stroma was categorized into TSR-high (TSR≥50%) and TSR-low (TSR<50%) using both AI- and human pathology-based methods. AI-based TSR-high subgroup exhibited notably shorter Metastasis-Free Survival (MFS) with a statistical significance of p=0.029. Interestingly, pathologist-determined TSR subgroups lacked statistical significance in Recurrence-Free Survival (RFS), MFS, and Overall Survival (OS) (p>0.05). Density-based spatial clustering of applications with noise (DBSCAN) analysis identified two distinct Tumor-Stroma Landscape (TSL) clusters: TSL1 and TSL2. TSL2 showed significantly shorter RFS (p=0.045) and markedly reduced MFS (p<0.001) compared to TSL1.
TSL classification appears to offer better prognostic discrimination than traditional TSR evaluation in MCC. TSL can be reliably calculated using an AI-based classification framework and predict various prognostic features of MCC.
Details
- Title: Subtitle
- A Novel Artificial Intelligence-based Parameterization Approach of the Stromal Landscape in Merkel Cell Carcinoma: a Multi-Institutional Study
- Creators
- Chau M. Bui - University of RochesterMinh-Khang Le - University of YamanashiMasataka Kawai - University of Yamanashi HospitalHuy Gia Vuong - University of IowaKristin J. Rybski - University of RochesterKathleen Mannava - University of RochesterTetsuo Kondo - University of Yamanashi HospitalTakashi Okamoto - University of Yamanashi HospitalLeah Laageide - University of RochesterBrian Swick - University of IowaBonnie Balzer - Cedars-Sinai Medical CenterBruce R. Smoller - University of Rochester
- Resource Type
- Journal article
- Publication Details
- Laboratory investigation, Vol.104(9), 102123
- DOI
- 10.1016/j.labinv.2024.102123
- PMID
- 39147033
- NLM abbreviation
- Lab Invest
- ISSN
- 0023-6837
- eISSN
- 1530-0307
- Publisher
- Elsevier Inc
- Language
- English
- Electronic publication date
- 08/13/2024
- Date published
- 09/2024
- Academic Unit
- Dermatology; Pathology
- Record Identifier
- 9984696780802771
Metrics
41 Record Views