Authors’ reply to Kapoor and Mahajan, Fazal et al., and Gupta and Rangarajan

Abstract

We thank Kapoor and Mahajan,[1] Fazal et al.,[2] and Gupta and Rangarajan[3] for their keen interest, valuable appreciation, and insightful comments on our article, “Artificial intelligence-based prediction of oral mucositis in patients with head-and-neck cancer: A prospective observational study utilizing a thermographic approach.”[4] We agree with Fazal et al.[1] that, when assessing oral mucositis, it is imperative to give due consideration to the patient’s clinical history and conduct a thorough physical examination. These aspects hold paramount importance in the evaluation process. At the Homi Bhabha Cancer Hospital, Tata Memorial Center, Sangrur, India, we maintain patient records both in the electronic medical record (EMR) system and physical files, while also conducting regular patient examinations. We completely acknowledge the validity of the comment by Fazal et al.[1] that doctors invest significant time in evaluating medical images, and the automation of thermal image processing with the help of artificial intelligence would reduce computational time.[5,6] In the future, more efforts should be made to improve the computational algorithms for larger datasets. We agree with the comments by Kapoor and Mahajan[2] that radiation-induced mucositis takes a minimum of 5–14 days to evolve, and thus, the data acquisition must be done within that specific time slot. Thermal imaging of patients with head-and-neck cancer was conducted over a four-week period as part of a preliminary (pilot) study. Our study[4] was cross-sectional, but thermal data were acquired every week; hence, in many cases, the thermal data were possibly from the same patient in consecutive weeks, but we did not document the data (details) on a weekly basis, which could have provided better clarity to the thermal data. Obtaining real-time data is an extremely time-consuming process, given the concurrent focus on uninterrupted treatment for patients with head-and-neck cancer during the data acquisition phase. The aim of our study[4] was to check the predictability of artificial intelligence-based thermal imaging for oral mucositis. We did not document the clinical aspects. To clarify, we included all cases that received a curative radical radiation dose of 70 Gy; we did not include any patients who received palliative radiotherapy. We agreed with the observation of Gupta and Rangarajan[3] that a larger sample size could have made the deep learning method more sensitive.[7] Real-time thermal data acquisition is a time-consuming process, and data acquisition is still ongoing. In the future, more efforts will be made to improve the computational algorithm on larger thermal datasets that contain images from patients with all grades of mucositis. We thank Gupta and Rangarajan[3] for their recommendations. We will go through the Medical Image Computing and Computer Assisted Intervention Society (MICCAI) and Checklist for Artificial Intelligence in Medical Imaging (CLAIM) checklists and ensure that we incorporate them in future studies.[8,9] We express our heartfelt gratitude for their valuable suggestions, which we hold in the highest regard. The inputs are immensely appreciated, and we are deeply thankful for the contributions. Financial support and sponsorship Nil. Conflicts of interest There are no conflicts of interest.