Elastic Deformation of Optical Coherence Tomography Images of Diabetic Macular Edema for Deep-Learning Models Training: How Far to Go?

• Published in: IEEE journal of translational engineering in health and medicine Vol. 11; pp. 487 - 494
• Main Authors: Bar-David, Daniel, Bar-David, Laura, Shapira, Yinon, Leibu, Rina, Dori, Dalia, Gebara, Aseel, Schneor, Ronit, Fischer, Anath, Soudry, Shiri
• Format: Journal Article
• Language: English
• Published: United States IEEE 01.01.2023; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Algorithms; Data augmentation; Datasets; Deep Learning; Deformation; Diabetes; Diabetes Mellitus; Diabetic Retinopathy - diagnosis; DME; Edema; Elastic deformation; Fluids; Humans; Image enhancement; Machine learning; Macular Edema - diagnostic imaging; OCT; Ophthalmology; Optical Coherence Tomography; Retina; Tomography; Tomography, Optical Coherence - methods; Training; elastic deformation; OCT; deep learning; Data augmentation; DME
• ISSN: 2168-2372; 2168-2372
• DOI: 10.1109/JTEHM.2023.3294904

- Objective: To explore the clinical validity of elastic deformation of optical coherence tomography (OCT) images for data augmentation in the development of deep-learning model for detection of diabetic macular edema (DME). Methods: Prospective evaluation of OCT images of DME (n = 320) subject to elastic transformation, with the deformation intensity represented by (<inline-formula> <tex-math notation="LaTeX">\sigma </tex-math></inline-formula>). Three sets of images, each comprising 100 pairs of scans (100 original & 100 modified), were grouped according to the range of (<inline-formula> <tex-math notation="LaTeX">\sigma </tex-math></inline-formula>), including low-, medium- and high-degree of augmentation; (<inline-formula> <tex-math notation="LaTeX">\sigma </tex-math></inline-formula> = 1-6), (<inline-formula> <tex-math notation="LaTeX">\sigma </tex-math></inline-formula> = 7-12), and (<inline-formula> <tex-math notation="LaTeX">\sigma </tex-math></inline-formula> = 13-18), respectively. Three retina specialists evaluated all datasets in a blinded manner and designated each image as 'original' versus 'modified'. The rate of assignment of 'original' value to modified images (false-negative) was determined for each grader in each dataset. Results: The false-negative rates ranged between 71-77% for the low-, 63-76% for the medium-, and 50-75% for the high-augmentation categories. The corresponding rates of correct identification of original images ranged between 75-85% (<inline-formula> <tex-math notation="LaTeX">\text{p}> </tex-math></inline-formula>0.05) in the low-, 73-85% (<inline-formula> <tex-math notation="LaTeX">\text{p}> </tex-math></inline-formula>0.05 for graders 1 & 2, p = 0.01 for grader 3) in the medium-, and 81-91% (<inline-formula> <tex-math notation="LaTeX">\text{p} < 0.005 </tex-math></inline-formula>) in the high-augmentation categories. In the subcategory (<inline-formula> <tex-math notation="LaTeX">\sigma </tex-math></inline-formula> = 7-9) the false-negative rates were 93-83%, whereas the rates of correctly identifying original images ranged between 89-99% (<inline-formula> <tex-math notation="LaTeX">\text{p}> </tex-math></inline-formula>0.05 for all graders). Conclusions: Deformation of low-medium intensity (<inline-formula> <tex-math notation="LaTeX">\sigma </tex-math></inline-formula> = 1-9) may be applied without compromising OCT image representativeness in DME. Clinical and Translational Impact Statement-Elastic deformation may efficiently augment the size, robustness, and diversity of training datasets without altering their clinical value, enhancing the development of high-accuracy algorithms for automated interpretation of OCT images.