Prognostic Value of Deep Learning-Mediated Treatment Monitoring in Lung Cancer Patients Receiving Immunotherapy

Abstract

BackgroundCheckpoint inhibitors provided sustained clinical benefit to metastatic lung cancer patients. Nonetheless, prognostic markers in metastatic settings are still under research. Imaging offers distinctive advantages, providing whole-body information non-invasively, while routinely available in most clinics. We hypothesized that more prognostic information can be extracted by employing artificial intelligence (AI) for treatment monitoring, superior to 2D tumor growth criteria.MethodsA cohort of 152 stage-IV non-small-cell lung cancer patients (NSCLC) (73 discovery, 79 test, 903CTs), who received nivolumab were retrospectively collected. We trained a neural network to identify morphological changes on chest CT acquired during patients’ follow-ups. A classifier was employed to link imaging features learned by the network with overall survival.ResultsOur results showed significant performance in the independent test set to predict 1-year overall survival from the date of image acquisition, with an average area under the curve (AUC) of 0.69 (p < 0.01), up to AUC 0.75 (p < 0.01) in the first 3 to 5 months of treatment, and 0.67 AUC (p = 0.01) for durable clinical benefit (6 months progression-free survival). We found the AI-derived survival score to be independent of clinical, radiological, PDL1, and histopathological factors. Visual analysis of AI-generated prognostic heatmaps revealed relative prognostic importance of morphological nodal changes in the mediastinum, supraclavicular, and hilar regions, lung and bone metastases, as well as pleural effusions, atelectasis, and consolidations.ConclusionsOur results demonstrate that deep learning can quantify tumor- and non–tumor-related morphological changes important for prognostication on serial imaging. Further investigation should focus on the implementation of this technique beyond thoracic imaging.

Development of a Prognostic AI-Monitor for Metastatic Urothelial Cancer Patients Receiving Immunotherapy

“…In-depth analysis revealed stark differences in the prognostic value of morphological changes depending on the time point of treatment, with the first 9 months of treatment being the most predictive and significant AUCs > 0.70, peaking to over 0.80 for both abdominal imaging, and thoracic imaging. Similar findings were observed in our previous study on NSCLC (8), where the changes recorded by the algorithm in the first 3 to 5 months of treatment were observed to have a higher prognostic value. In the present study, we extended the system to include both the thorax and abdomen, and trained with far larger datasets both in terms of pre-training as well as survival association.…”

“…The tracker module was designed following the research of Balakrishnan et al (11) and Zhao et al (12), as well as our previous work on chest imaging in NSCLC (8). The network receives two images as input (i.e., a moving and fixed one) concatenated along the channel axis.…”

“…Our goal was for the AI to learn to match corresponding imaging landmarks, and to cover a large set of possible variations while, simultaneously, containing its scope (or prior knowledge) to landmarks in the abdominal area. For the chest AI tracker module, we leveraged the trained weights from the NSCLC study (8). The code of both tracker and localizer have been added to the department AI repository 2 .…”

“…Novel techniques of computational imaging and artificial intelligence (AI) can be the basis for quantitative methods for treatment monitoring (8). Specifically, AI algorithms can be seen as methods to capture, measure, and quantify complex highly-variable anatomical phenomena for prognostic purposes, in a robust and time-efficient manner.…”

“…We term our AI system the Prognostic AI-monitor (PAM). Recently, a similar pilot approach was tested in a study on chest imaging of a NSCLC cohort (8), demonstrating accurate response prediction and a correlation with overall survival. In this study, we aim to extend the model to thoracoabdominal imaging, and validate it on a cohort of metastatic urothelial cancer patients treated with anti-PD1/PDL1.…”