TY - GEN

T1 - Computerised Methods for Monitoring Diabetic Foot Ulcers on Plantar Foot

T2 - A Feasibility Study

AU - Goyal, Manu

AU - Reeves, Neil D.

AU - Rajbhandari, Satyan

AU - Yap, Moi Hoon

PY - 2022/7/27

Y1 - 2022/7/27

N2 - Recognition and analysis of Diabetic Foot Ulcers (DFU) by computerised methods has been an emerging research area with the evolution of image processing and machine learning algorithms. Precise documentation of wound size over time allows clinicians to gauge responses to treatment, improving healing rates by modifying interventions as required. One of the major issues in the analysis of DFU is non-standardised foot images captured with cameras including factors such as distance of the camera from the foot and orientation of the image. Designing a computerised solution to determine site of DFU and measurements of area for remote assessment and monitoring represents a significant challenge due to the variables involved. In this work, we propose a new computerised solution with the combination of image processing and deep learning algorithms to estimate the site and predict the progress (based on estimated area index) of the DFU irrespective of distance and orientation of the plantar foot. First we segment the foot region and align the foot by fixing the orientation of a series of longitudinal images. Then we localise the region of interest of DFUs and find its relative size to the foot area. We introduce a distribution analysis to determine the site of DFUs. Finally, we introduce an area index () to predict the healing progress of DFU at different time intervals (t). We demonstrate the feasibility of our proposed method on 154 longitudinal DFUs of plantar foot. We achieved 92.3% on site estimation and 84.7% on healing progress prediction.

AB - Recognition and analysis of Diabetic Foot Ulcers (DFU) by computerised methods has been an emerging research area with the evolution of image processing and machine learning algorithms. Precise documentation of wound size over time allows clinicians to gauge responses to treatment, improving healing rates by modifying interventions as required. One of the major issues in the analysis of DFU is non-standardised foot images captured with cameras including factors such as distance of the camera from the foot and orientation of the image. Designing a computerised solution to determine site of DFU and measurements of area for remote assessment and monitoring represents a significant challenge due to the variables involved. In this work, we propose a new computerised solution with the combination of image processing and deep learning algorithms to estimate the site and predict the progress (based on estimated area index) of the DFU irrespective of distance and orientation of the plantar foot. First we segment the foot region and align the foot by fixing the orientation of a series of longitudinal images. Then we localise the region of interest of DFUs and find its relative size to the foot area. We introduce a distribution analysis to determine the site of DFUs. Finally, we introduce an area index () to predict the healing progress of DFU at different time intervals (t). We demonstrate the feasibility of our proposed method on 154 longitudinal DFUs of plantar foot. We achieved 92.3% on site estimation and 84.7% on healing progress prediction.

U2 - 10.1007/978-3-031-12053-4_15

DO - 10.1007/978-3-031-12053-4_15

M3 - Conference contribution/Paper

SN - 9783031120527

T3 - Lecture Notes in Computer Science

SP - 199

EP - 211

BT - Medical Image Understanding and Analysis

A2 - Yang, Guang

A2 - Aviles-Rivero, Angelica

A2 - Roberts, Michael

A2 - Schönlieb, Carola-Bibiane

PB - Springer-Verlag

CY - Cham

ER -