Will AI Make IO Better?

Artificial intelligence (AI) is the future of clinical data management with the potential to aid in the intersection between complex data and clinical decision making. Given AIs ability to rapidly analyze big data, interventional radiologists (IR) can design more individualized staging systems by collaborating to develop large highly organized multi-institutional registries thus improving how medicine is practiced, according to an article in the Journal of Vascular and Interventional Radiology

In this Research in Translation article, the authors used hepatic interventional oncology for the application of AI to IR.

One type of AI is machine learning (ML). In ML a computer learns a task from data although it is not explicitly programmed to do so. A specific type of ML is artificial neural networks (ANNs), which are useful for modeling complex nonlinear relationships between inputs and outputs. An example would be using an ANN to predict tumor response to transarterial chemoembolization based on baseline patient characteristics.

Using ANNs with a large number of hidden layers between the input and output is called “deep learning.” One type of ANN is a convolutional neural network (CNN) which automatically classifies malignant and benign hepatic masses on a patient’s magnetic resonance (MR) imaging examination aiding in decision support in radiologic diagnosis. CNNs can be used to classify hepatic masses on ultrasound, computed tomography (CT), or MR imaging.

The authors suggest that prediction accuracies “could possibly be improved by using MR imaging as the data input (owing to higher level of intrinsic soft tissue contrast resolution), including volumetric three-dimensional input (to more completely analyze the full mass), expanding the dataset to include information from multi-institutional registries (which can allow for more stringent inclusion criteria of pathologic proof), and incorporating clinical data.”

CNNs also allow physicians to understand the reasoning behind decisions and predict when they may fail.

“By integrating ML into diagnosis, treatment, and management, AI can empower physicians to provide the highest quality personalized care in an efficient manner that meets the demands of modern clinical practice. Whereas medicine has traditionally focused on incremental hypothesis-based research, AI allows for a new paradigm where ‘big data’ can be rapidly analyzed, uncovering new insights that may otherwise have required decades of prospective trials,” the authors concluded. “Moreover, a more objective ML-based analysis can decrease bias in clinical decision making. Current staging systems rely on a small number of predictive features to allocate patients to different therapy arms. An AI-based approach can drastically improve allocation by integrating the patient’s entire clinical data into decision making.”

 

--Kelsey Moroz

 

Reference 

Letzen B, Wang CJ, Chapiro J. The role of artificial intelligence in interventional oncology: a primer. 2018; epub ahead of print. DOI: https://doi.org/10.101