The use of deep learning models to predict progression-free survival in patients with neuroendocrine tumors

Plain language summary - The use of deep learning models to predict progression-free survival in patients with neuroendocrine tumorsNeuroendocrine tumors (NET) are slow-growing cancers. How well cancers respond to treatment is usually measured using 'Response Evaluation Criteria in Solid Tumors (RECIST)', which is based on measuring the size of tumors. RECIST is not well suited for assessing NETs as these tumors often grow slowly and rarely shrink significantly, so it is difficult to tell whether a treatment has any effect. A better way of measuring how well NETs are responding to treatment is needed, to ensure that patients receive the right treatment as early as possible.The RAISE project aimed to use a type of artificial intelligence (AI) called 'deep learning' to examine images of NETs, taken from patients in a clinical trial of treatment with lanreotide, to help predict how they might respond to treatment. These images were analyzed by the deep learning AI to see if there are any features of tumors, other than shape or size, that may help to predict response to treatment.The project showed that this technology can detect features in images of NETs, other than the shape and size of tumors, that are useful for predicting how well a treatment might work for an individual patient. However, this technology could not improve prediction of how well a treatment would work at an earlier stage compared with other currently used indicators.Overall, further research and work is needed to improve this technology. However, these results show that deep learning may have the potential to improve prediction of treatment response in patients with NETs. Aim: The RAISE project assessed whether deep learning could improve early progression-free survival (PFS) prediction in patients with neuroendocrine tumors. Patients & methods: Deep learning models extracted features from CT scans from patients in CLARINET (NCT00353496) (n = 138/204). A Cox model assessed PFS prediction when combining deep learning with the sum of longest diameter ratio (SLDr) and logarithmically transformed CgA concentration (logCgA), versus SLDr and logCgA alone. Results: Deep learning models extracted features other than lesion shape to predict PFS at week 72. No increase in performance was achieved with deep learning versus SLDr and logCgA models alone. Conclusion: Deep learning models extracted relevant features to predict PFS, but did not improve early prediction based on SLDr and logCgA.