OBJECTIVES: One of the significant factors influencing the cost of building computer vision algorithms applied to pathology images is the need for per-pixel labeling (PPL) by a trained physician. Multiple instance learning (MIL) technique allows to train algorithms without having PPL, only with the global diagnosis of the patient. Therefore, we evaluate the work done with this technique and its influence in reducing healthcare resources (HRU).

METHODS: We conducted a targeted literature review until January 2023, searching for: "multiple instance learning" AND "attention” AND “whole slide imaging pathology” OR "WSI pathology”. A data extraction grid was created to analyze the following variables: proportion of applications per disease or medical specialty, number of WSI with global label, type of dataset, HRU component addressed, and registered metrics for performance. Categorical data is presented as percentage and continuous data as means.

RESULTS: 62 articles underwent full text screening and data extraction. Therapeutic area/ Medical specialties included: Oncology: 57 (91%), Gastroenterology: 3 (5%), Hematology: 1(2%), Infectious Diseases: 1(2%). Subsets for oncology included: Breast (28), Lung (14) and Gastrointestinal (12) cancer. The WSI samples employed ranged from 24 to 20,229 (mean: 226). 57 (91%) articles applied labels extracted from diagnostic registries, 27 (43%) articles used data from the Cancer Genome Atlas (TCGA) and 39 (58%) mentioned any element involving HRU optimization (reduction in number of trained physicians: 26, reduction in hours to diagnosis: 8; both: 5). The most used metric was accuracy (37, 60%), with a maximum (0.99), minimum (0.68), and average (0.89).

CONCLUSIONS: MIL research applied to healthcare is emerging. Oncology is a therapy area with high number of WSI samples employed, due to their potential benefit in reducing the need of trained physicians to generate labels. Further research is required to understand a value-based perspective of its application at the healthcare system level.