Artificial Intelligence Applications to Improve the Treatment of Locally Advanced Non-Small Cell Lung Cancers

Abstract

Locally advanced non-small cell lung cancer patients represent around one third of newly diagnosed lung cancer patients. There remains a large unmet need to find treatment strategies that can improve the survival of these patients while minimizing therapeutical side effects. Increasing the availability of patients' data (imaging, electronic health records, patients' reported outcomes, and genomics) will enable the application of AI algorithms to improve therapy selections. In this review, we discuss how artificial intelligence (AI) can be integral to improving clinical decision support systems. To realize this, a roadmap for AI must be defined. We define six milestones involving a broad spectrum of stakeholders, from physicians to patients, that we feel are necessary for an optimal transition of AI into the clinic.

Keywords: artificial intelligence; clinical decision aids; deep learning; lung cancers; radiomics.

Figure 1

An overview of the critical points and key enabling technologies for the application of AI technologies to improve the current IGART workflow in the management of LA-NSCLC patients. Each of the critical points includes different stakeholders who need to be involved for a productive and collaborative environment. The structure of this review follows the progression of this figure.

Figure 2

(a) GAN framework showing CBCT to CT image translation. The generator tries to learn the data distribution of CBCT and CT images and uses this learned representation to convert a CBCT image into a “fake” CT image. The generator learns this representation by trying to fool the discriminator while it compares real and “fake” CTs and learns to differentiate between them. (b) Workflow for adaptive radiotherapy enabled by dose calculation from CBCT images. Prior to dose recalculation, a CBCT image is converted to a synthetic CT (sCT), providing correct Hounsfield unit (HU) values and removing artifacts while preserving the anatomy.

Figure 2

(a) GAN framework showing CBCT to CT image translation. The generator tries to learn the data distribution of CBCT and CT images and uses this learned representation to convert a CBCT image into a “fake” CT image. The generator learns this representation by trying to fool the discriminator while it compares real and “fake” CTs and learns to differentiate between them. (b) Workflow for adaptive radiotherapy enabled by dose calculation from CBCT images. Prior to dose recalculation, a CBCT image is converted to a synthetic CT (sCT), providing correct Hounsfield unit (HU) values and removing artifacts while preserving the anatomy.

Figure 3

A response-based ART system using a three-component deep reinforcement learning (DRL). Automated dose decisions given by DQN (green dots) vs. clinical decision (blue dots) with RMSE 0.76 Gy. Dots in green and red represent indications of “good” and “bad”. Reproduced with permission from Tseng, Med Phys, 2017.

Figure 4

Overview of the interconnections among the developments of an iCCDS and the traditional software lifecycle schema.