Guide for collecting and reporting metadata on protocol variables and parameters from slide-based histotechnology assays to enhance reproducibility

Building living systematic reviews and reporting standards for comparative microscopic analysis of white diseases in hard corals

“…Recent reviews of biomedical and public health, including slide‐based research, found that articles generally lacked rigour and transparency in one or more fundamental areas needed for reproducing studies (Freedman & Inglese, 2014; Naudet et al., 2018; Wallach et al., 2018). Because practices have been found to be inadequate, various groups have aimed to address transparency through development of reporting guidelines (Chiriboga et al., 2022; Kenall et al., 2015; Marcus, 2016). As highlighted in the recent study by Hawthorn et al.…”

“…The foundational concepts for assessment of diseased tissues are similar across scientific areas (Chiriboga et al., 2022; Gibson‐Corley et al., 2013; Meyerholz & Beck, 2018b; Wolf et al., 2015) and include principles associated with bias control in tissue evaluation (Chiriboga et al., 2022; Meyerholz & Beck, 2018a). For example, bias can be reduced through steps at sample processing stage that involve following a set protocol for slide assessment (Gibson‐Corley et al., 2013; Holland & Holland, 2011).…”

“…The foundational concepts for assessment of diseased tissues are similar across scientific areas (Chiriboga et al, 2022 et al (2013) in addition to (Cross, Germolec et al, 2004;Landis Koch, 1977). At the slide-reading stage authors should also be sure to provide information on the number of sections examined per sample, and the number of sections examined per stain, per sample.…”

Performance of externally validated machine learning models based on histopathology images for the diagnosis, classification, prognosis, or treatment outcome prediction in female breast cancer: A systematic review

“…11 During tissue processing, slide preparation, slide digitization, image compression, and image storage, hidden variables (i.e., image data unrelated to the actual prediction/classification task that can affect the performance of ML models) may be introduced to WSIs. [18][19][20] As protocols, equipment, and consumables utilized during these processes vary among different institutions, 21 the independent datasets used during EV should ideally be extracted from a different data source, such as another clinic or hospital system. 11,[15][16][17] Only EV is considered important evidence of generalizability as patterns learned from hidden variables of training datasets (instead or in addition to those that can be learned from the intended target variables) are not expected to improve ML models' performance when they are tested with independent datasets.…”

“…During tissue processing, slide preparation, slide digitization, image compression, and image storage, hidden variables (i.e., image data unrelated to the actual prediction/classification task that can affect the performance of ML models) may be introduced to WSIs. 18 , 19 , 20 As protocols, equipment, and consumables utilized during these processes vary among different institutions, 21 the independent datasets used during EV should ideally be extracted from a different data source, such as another clinic or hospital system. 11 , 15 , 16 , 17 …”