Sharing Biomedical Data: Strengthening AI Development in Healthcare

Abstract

Artificial intelligence (AI)-based solutions have revolutionized our world, using extensive datasets and computational resources to create automatic tools for complex tasks that, until now, have been performed by humans. Massive data is a fundamental aspect of the most powerful AI-based algorithms. However, for AI-based healthcare solutions, there are several socioeconomic, technical/infrastructural, and most importantly, legal restrictions, which limit the large collection and access of biomedical data, especially medical imaging. To overcome this important limitation, several alternative solutions have been suggested, including transfer learning approaches, generation of artificial data, adoption of blockchain technology, and creation of an infrastructure composed of anonymous and abstract data. However, none of these strategies is currently able to completely solve this challenge. The need to build large datasets that can be used to develop healthcare solutions deserves special attention from the scientific community, clinicians, all the healthcare players, engineers, ethicists, legislators, and society in general. This paper offers an overview of the data limitation in medical predictive models; its impact on the development of healthcare solutions; benefits and barriers of sharing data; and finally, suggests future directions to overcome data limitations in the medical field and enable AI to enhance healthcare. This perspective is dedicated to the technical requirements of the learning models, and it explains the limitation that comes from poor and small datasets in the medical domain and the technical options that try or can solve the problem related to the lack of massive healthcare data.

Exploring AI in Healthcare: How the Acceleration of Data Processing can Impact Life Saving Diagnoses

“…Model representing patient and data relationship for improving models, while balancing impact on patient security [17] With a continuation of these trends as found in this research, AI systems that make improvements to data standardization, would increase the accuracy of diagnoses drastically enough to make ML an extremely valuable tool in healthcare. Within Yoo's study, the application of AI was limited because the specific cohort of participants for the study included a really small number of lung cancer positive patients, and the actual clinical prevalence of positive encounters for national lung screening trials would be extremely rare.…”

“…Throughout many studies in the application of AI against human and algorithmic performance, a large risk that remains present in these studies would be that the models have limited to no clinical context, including the patient's medical history or prior laboratory findings, which can be critical in finding the proper treatment plans [18]. When it comes to the processing of data, there are a lot of times in which the systems connecting and supplying each other information automatically, are distributed through multiple heterogeneous and semantically incompatible systems, which leads to interoperability problems and inconsistencies [17]. When subsystems of a medical center are inconsistent with each other, and the data transferred between systems gets processed at each stage, there is increased potential for data loss and context misinterpretation.…”

Towards Machine Learning-Aided Lung Cancer Clinical Routines: Approaches and Open Challenges

“…Lack of large datasets remains a problem for healthcare solutions since models cannot learn how to deal with complex data due to insufficient training samples [ 205 , 206 ]. The difficult access to the medical data has slowed the progress of CAD development.…”

A Vision-Based System for Stage Classification of Parkinsonian Gait Using Machine Learning and Synthetic Data

“…The use of synthetic data has also been applied in the biomedical field, where augmented data has been developed to increase the number and variability of examples [55]. Furthermore, different studies with a similar goal as this paper's have also focused on generating synthetic data to incorporate it to the training set and avoid class imbalance problems [56][57][58].…”