Explainable Machine Learning (XML) for multimedia-based healthcare systems: Opportunities, challenges, ethical and future prospects
Explainable Machine Learning for Multimedia Based Healthcare Applications, Springer International Publishing Ag, ss.21-46, 2023
- Yayın Türü: Kitapta Bölüm / Araştırma Kitabı
- Basım Tarihi: 2023
- Doi Numarası: 10.1007/978-3-031-38036-5_2
- Yayınevi: Springer International Publishing Ag
- Sayfa Sayıları: ss.21-46
- Anahtar Kelimeler: Black-box, Computer-based systems, Explainability, Healthcare systems, Internet of Things, Interpretability, Machine learning, Multimedia data
- Recep Tayyip Erdoğan Üniversitesi Adresli: Evet
Özet
Various scientific fields have abruptly shifted in the direction of data-dependent methodologies in recent years. In some instances, concurrent improvements in data processing have made this process possible, and the advancement of technology for network infrastructures. In the healthcare systems, where data abundance has generated a rush of new techniques for effective data collection and processing, this new predicament is particularly noticeable. This presents a unique opportunity to apply machine learning (ML) methodologies to issues where more conventional data analysis methods would falter. Access to multimedia data is now possible thanks to advancements in the Internet of Things, devices, and smartphones. Applications based on ML-based algorithms use data including images, video, audio, and text as input to help the current healthcare diagnosis, prognostic, or receive treatment. However, this scenario also presents some significant problems. Interpretability and explainability of such models is one of the like issues, particularly for sophisticated nonlinear models. If these issues are not resolved in the healthcare system, adoption chances may be severely hampered, in actual use of ML-based approaches for data processing in computer-based systems. Therefore, this chapter presents recent surveys on ML-based explainability and interpretability on multimedia based healthcare applications. The applicability, challenges, and future prospects of these techniques was discussed and addressed, with their significant impact on healthcare systems. The major ethical problems that can arise using ML-based interpretability and explainability were presented. Besides enhancing model interpretability as a standalone objective, the chapter contend that the integration of physicians in the development and implementation of data analysis interpretation approaches. Otherwise, it is doubtful that ML-based models will be included into common clinical and healthcare procedures.