Smart or Simple Solution? How Artificial Intelligence Facilitates Clinical Decision for Radiation Oncologists in Preoperative Chemoradiotherapy for Locally Advanced Rectal Cancer

Abstract

Preoperative concurrent chemoradiotherapy (CCRT) is one of the standard treatments for patients with locally advanced rectal cancer (LARC). The research prioritizing neoadjuvant CCRT in 1990s adopted 3-dimentional conformal radiotherapy (3DCRT) as treatment technique. While technique progresses, intensity-modulated radiation therapy (IMRT) has become more widely used in irradiating rectal cancer with evidence suggesting less toxicities than 3DCRT. Artificial-intelligence (AI)-based clinical decision support system (CDSS) is designed to combine relevant guidelines, medical journals, textbooks and patients’ features and then recommend treatment options according to the level of evidence. Because of different purpose, study protocol, and gap in RT technique in various researches, how an AI-based CDSS provides suggestion with incorporation of medical literature and clinical condition could significantly improve clinical practice. Therefore, we reported our experience in using AI-based CDSS for neoadjuvant CCRT in LARC. We retrospectively included patients aged 18 years and older with stage T2-4N1 or T3-4N0 lower rectal carcinoma, who received preoperative CCRT from 2018 January to December at a single institute. All clinical data and attributes were input into the CDSS with de-identification process. A study panel including original radiation oncologists would review RT received by each patient and the advice from the AI-based CDS system. Literature provided from the CDSS and the clinicians, searching from the PubMed and Cochrane Library databases, would also be investigated. Twenty-seven patients with LARC receiving neoadjuvant CCRT were recruited. All patients completed the neoadjuvant therapy without severe toxicities and received IMRT for their preoperative RT. For 23 patients with T2-3 rectal cancer, AI suggested 3DCRT as “Recommended” and IMRT as “For Consideration”; while other 4 patients with T4 rectal cancer, AI suggested IMRT as “Recommended.” The AI system provided 3 landmark studies, which confirmed the importance of neoadjuvant CCRT in LARC, adopted 3DCRT as the RT technique. The radiation oncologists provides more articles, including 3 radio-physics papers, 6 retrospective clinical studies and 1 conclusive meta-analysis, publishing from 2011 to 2018. Although without advantage in overall survival, significant benefit in less gastrointestinal toxicity was noted in IMRT, comparing to 3DCRT. AI-based CDSS could rapidly provide evidence-based recommendation by incorporating patient data and updating literature. Studies suggest this system could facilitate clinical practice, yet opposing opinions exist. From our experience, it helps physicians for most practice; however, in complicated clinical setting, well-experienced physicians are still required to solve dilemma.