Explainable Image Quality Assessments in Teledermatological Photography

Abstract

Background and Objectives: Image quality is a crucial factor in the effectiveness and efficiency of teledermatological consultations. However, up to 50% of images sent by patients have quality issues, thus increasing the time to diagnosis and treatment. An automated, easily deployable, explainable method for assessing image quality is necessary to improve the current teledermatological consultation flow. We introduce ImageQX, a convolutional neural network for image quality assessment with a learning mechanism for identifying the most common poor image quality explanations: bad framing, bad lighting, blur, low resolution, and distance issues. Methods: ImageQX was trained on 26,635 photographs and validated on 9,874 photographs, each annotated with image quality labels and poor image quality explanations by up to 12 board-certified dermatologists. The photographic images were taken between 2017 and 2019 using a mobile skin disease tracking application accessible worldwide. Results: Our method achieves expert-level performance for both image quality assessment and poor image quality explanation. For image quality assessment, ImageQX obtains a macro F1-score of 0.73 ± 0.01, which places it within standard deviation of the pairwise inter-rater F1-score of 0.77 ± 0.07. For poor image quality explanations, our method obtains F1-scores of between 0.37 ± 0.01 and 0.70 ± 0.01, similar to the inter-rater pairwise F1-score of between 0.24 ± 0.15 and 0.83 ± 0.06. Moreover, with a size of only 15 MB, ImageQX is easily deployable on mobile devices. Conclusion: With an image quality detection performance similar to that of dermatologists, incorporating ImageQX into the teledermatology flow can enable a better, faster flow for remote consultations.

Mobile health technologies in an interventional hybrid study on actinic keratosis: Results from an early phase randomized controlled trial investigating the safety and efficacy of a cytosolic phospholipase A2 inhibitor gel in photodamaged skin

“…Supplemented by questionnaires on patient‐reported side effects, we were able to ensure patient safety using remote assessments which supports the BYOD concept in contemporary clinical research 34,47 . Hybrid trials that incorporate the use of PSPs do recognize and report on the fluctuations in image quality, 48 which can be ameliorated by image quality assessment by dermatologists, as performed in our study, or by the integration of convolutional neural networks for image quality assessment 49 . Despite the many advantages of collecting PSPs, changes in ambient lighting and background were frequently encountered in our study and assumed to pose a potential risk to the standardized monitoring of side effects.…”

“…34,47 Hybrid trials that incorporate the use of PSPs do recognize and report on the fluctuations in image quality, 48 which can be ameliorated by image quality assessment by dermatologists, as performed in our study, or by the integration of convolutional neural networks for image quality assessment. 49 Despite the many advantages of collecting PSPs, changes in ambient…”

Development of SkinTracker, an integrated dermatology mobile app and web portal enabling remote clinical research studies

“…Finally, image quality rating of patient-submitted photographs was conducted using the framework developed by ImageQX (42). ImageQX is a convolutional neural network for image quality assessment specifically developed for teledermatology and includes input from 12 dermatologists and a dataset of 36,509 skin photographs obtained from 2017 to 2019 using a mobile skin disease tracking app incorporating patients internationally (42).…”

Quality Assessment Assistance of Lateral Knee X-rays: A Hybrid Convolutional Neural Network Approach

“…The models are evaluated using accuracy and achieve a score of 0.97 on the validation dataset. Jalaboi et al (2023) [15] focus on the application of a custom CNN architecture "ImageQX". The architecture is trained on 36,509 images, out of which the validation set consists of 9874 photographs.…”

“…= 0.95, ACC = 0.94, F1 = 0.96, Spec. = 0.94, AUC = 0.96 [14] Angiography imagery 200 ACC = 0.97 [15] Teledermatological photography 39,509 F1 = 0.73 ± 0.01 [16] Fundus photography 216 ACC = 0.98, Sens. = 0.99, Spec.…”