AI-Driven Image Analysis in Central Nervous System Tumors-Traditional Machine Learning, Deep Learning and Hybrid Models

Abstract

The interpretation of imaging in medicine in general and in oncology specifically remains problematic due to several limitations which include the need to incorporate detailed clinical history, patient and disease-specific history, clinical exam features, previous and ongoing treatment, and account for the dependency on reproducible human interpretation of multiple factors with incomplete data linkage. To standardize reporting, minimize bias, expedite management, and improve outcomes, the use of Artificial Intelligence (AI) has gained significant prominence in imaging analysis. In oncology, AI methods have as a result been explored in most cancer types with ongoing progress in employing AI towards imaging for oncology treatment, assessing treatment response, and understanding and communicating prognosis. Challenges remain with limited available data sets, variability in imaging changes over time augmented by a growing heterogeneity in analysis approaches. We review the imaging analysis workflow and examine how hand-crafted features also referred to as traditional Machine Learning (ML), Deep Learning (DL) approaches, and hybrid analyses, are being employed in AI-driven imaging analysis in central nervous system tumors. ML, DL, and hybrid approaches coexist, and their combination may produce superior results although data in this space is as yet novel, and conclusions and pitfalls have yet to be fully explored. We note the growing technical complexities that may become increasingly separated from the clinic and enforce the acute need for clinician engagement to guide progress and ensure that conclusions derived from AI-driven imaging analysis reflect that same level of scrutiny lent to other avenues of clinical research.

GradWise: A Novel Application of a Rank-Based Weighted Hybrid Filter and Embedded Feature Selection Method for Glioma Grading with Clinical and Molecular Characteristics

“…Typically, the diagnosis of glioma is made by obtaining tissue for pathological examination with molecular alterations being increasingly important for CNS tumor classification [ 6 , 7 , 8 ]. The isocitrate dehydrogenase (IDH) mutation is now more routinely employed as a molecular marker, given its prognostic value [ 4 , 9 , 10 ], but is limited by the associated costs and turnaround time of molecular testing, with p.R132H-specific IDH1 immunohistochemistry costing USD 135, single-gene sequencing costing USD 420, and next-generation sequencing costing USD 1800 [ 9 ] and the time required for analysis ranging from approximately two days for immunohistochemistry to up to 14 days for next-generation sequencing [ 9 ]. IDH mutation vs. IDH wild-type confers superior prognosis, particularly when accompanied by 1p19q co-deletion altering the management of non-GBM gliomas in terms of type and timing of systemic management, while GBMs are treated with standard-of-care concurrent chemo-irradiation irrespective of IDH status.…”

Integrating Artificial Intelligence in Accounting: A Quantitative Economic Perspective for the Future of U.S. Financial Markets

“…In the realm of data analysis and interpretation, AI-driven methods, including machine learning and deep learning, offer a more dynamic and comprehensive approach compared to traditional techniques. Krauze et al (2022) demonstrate this in the context of image analysis, where AIdriven methods provide nuanced insights that traditional methods may miss. Similarly, in accounting, AI-driven tools can analyze vast datasets, offering deeper and more accurate financial insights than traditional methods, which are often limited by manual data processing and interpretation constraints.…”

Artificial Intelligence Applications in Glioma With 1p/19q Co‐Deletion: A Systematic Review

“…The protocol and scanners change between different institutions, which may be limiting factors in the development of a robust ML model. As regard to DL models, image features can be “learned” implicitly through the iterative process of optimizing prediction performance, it takes images as input eliminating the steps of feature engineering and selection, hence potentially mitigating human bias 65 . However, DL is not without drawbacks, it depends on tremendous volume of high quality, well‐annotated data to train a reliable model, 66 such data are difficult to aggregate in large part because of regulatory and privacy concerns across institutions.…”

“…As regard to DL models, image features can be "learned" implicitly through the iterative process of optimizing prediction performance, it takes images as input eliminating the steps of feature engineering and selection, hence potentially mitigating human bias. 65 However, DL is not without drawbacks, it depends on tremendous volume of high quality, well-annotated data to train a reliable model, 66 such data are difficult to aggregate in large part because of regulatory and privacy concerns across institutions. Thus, data-sharing and federated learning may be future directions for overcoming current barriers for supporting clinical implementation of DL models.…”