Improving the assessment of tympanic membrane injuries using adjustable Flood Fill algorithm

Background. Accurate and rapid assessment of tympanic membrane perforations is essential in clinical otolaryngology, particularly in resource-constrained environments. The purpose of the work is to develop a computer system powered by highly adjustable segmentation tools that can be quickly applied to improve the assessment, reduce the time needed for obtaining consistent, objective and precise results when dealing with endoscopic images of different quality, contrast, amount of light and other variabilities. Materials and methods. To achieve this, a specialized web-based application was developed, utilizing a modified adaptive Flood Fill segmentation algorithm. The experiments were conducted using the developed system and a dataset of endoscopic images with various types of perforations, examined by expert to ensure the quality and applicability of the proposed models. Results. Experimental results demonstrated that the adaptive segmentation algorithm provided reliable and consistent identification of damaged areas, significantly reducing diagnostic variability and probable inaccuracies. Adjustability of algorithm parameters allowed accurate segmentation even under suboptimal imaging conditions (low contrast, blurred images), addressing typical clinical challenges. The practical usability and intuitive design of the application minimized training requirements, facilitating potential rapid clinical deployment and ability to assess perforations quicker than with traditional visual approach. Conclusions. The proposed model, method and web-based application effectively meet the goal of the work — improving diagnostic reliability and the clinical workflow, saving clinicians’ time. The developed tool has substantial potential to enhance patient outcomes in emergency and resource-limited settings. Further improvement through advanced image processing techniques could increase accuracy and applicability even more broadly across other medical domains.