Table 4.
Radiomic prediction model study design characteristics with their radiomics quality score.
| Radiomic Models for Biological Signature | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Study, Year | Location Inclusion Center |
Train (N) |
Test (N) | Age (mean) | Male (%) | Tumor Subside | Tumor Stage | Modality | #Features | Total RQS |
Domains: IM/ FR/VA/PI/LE/OS |
| Gao, 2021 [76] | Hunan, CHN | 237 | 79 | 47.9 | 69.9 | NA | All | T1+c | 530 | 16 | 8/5/6/3/6/0 |
| Zhang, 2020 [77] | Zuhai, CHN | 220 | 44 + 44 * | 47.4 † | 72.7 | NA | All | T1(c), T2 | 2364 | 19 | 8/6/6/5/7/0 |
The table lists study design characteristics, patient numbers, and total and individual domain quality scores of radiomics-based prediction model studies that were designed to align with biological features. Abbreviations: No. = number of selected patients; Train = number of patients analyzed in the training cohort; Test = number of patients analyzed in the test cohort; #Features = number of features collected; NA = nasopharynx; RQS = radiomics quality score; RSQ domains = IM: image protocol and feature reproducibility, FR: feature reduction and validation, VA: biologic/clinical validation and utility, PI: performance index, LE: high level of evidence, OS: open science. Note: † Median value; * 44 patients of an internal validation cohort and 44 in a separate external validation cohort.