SINGAPORE, 22 December 2020 (Tuesday) – Singapore is ranked second^[1] among developed nations with the highest proportion of diabetics in its population. An average of four^[2] diabetic-related foot amputations are performed every day and 1,500^[3] of such amputations are carried out annually. 25%^[4] of diabetic patients will develop foot ulcers in their lifetime, and 50% of those patients suffer from peripheral arterial disease.

Poor blood circulation and nerve injury arising from diabetes can result in foot complications. If left undetected or poorly managed, severe cases can lead to foot amputations. As such, regular foot screening, which includes the detection of foot deformities, sensory loss, and impaired blood flow, is crucial to managing diabetic conditions. Annual foot screening is recommended and with increased frequency if early problems are detected. Technological innovation could help automate and simplify the process to improve the patient experience and compliance to screening.

To contribute to the war on diabetes in Singapore, Team Live-Betics, comprising a group of students from the Singapore Management University (SMU) School of Information Systems who were taking a course which incorporates the SMU-X experiential learning framework of applying rigorous academic processes to address real-world issues, had embarked on a project with the clinicians from the National University Health System (NUHS) to develop a prototype which would enable diabetic patients to undertake foot screening more regularly and conveniently.

PODIA – a prototype for automated diabetic foot screening

Currently, the frequency of foot screenings varies from once every three months to once a year, depending on the patients’ foot risk category. This frequency is low in comparison to the spread of foot infections that could happen in a matter of weeks. In addition, the current foot screening process takes an average of 20-30 minutes depending on whether foot complications are detected, excluding travelling time to polyclinics or limited community screening facilities, and waiting for the screening to take place. Elderly patients and especially those with mobility difficulties would also need caregiver assistance for these visits.

Guided by SMU Associate Professor of Information Systems Tan Hwee Pink and Dr Chua Horng Ruey, a Senior Consultant with the Division of Nephrology at the National University Hospital (NUH), Team Live-Betics has developed a prototype called PODIA to explore the feasibility of early diagnosis of diabetic foot conditions using Red-Green-Blue (RGB) imaging and thermal imaging technology.

Internet-of-Things (IoT) applications and camera sensors are deployed to take feet images during foot screening. The feet images are then relayed automatically to the secured Amazon Web Services (AWS) cloud servers for analysis. On the backend, computer vision technologies are applied on the RGB images taken of the patient’s feet, to detect deformities. For thermal images captured, Artificial Intelligence (AI) algorithms are applied to observe perfusion levels (i.e. blood circulation) in the foot in its resting state - before and after pressure is applied. Ulcers and wounds can also be detected from the thermal images obtained. Authorised healthcare professionals can access the foot screening results through a clinician dashboard.

With PODIA, a patient is able to complete self-assisted foot screening by interacting with the Live-Betics web application which has videos with audio input to guide users through the process. AI technology is also used to provide real-time instructions to users. This reduces the healthcare manpower needed to perform the tests required and speeds up the screening process.

PODIA is compact and portable. The minimal space it occupies, makes deployment to community settings and GP clinics feasible. This way, foot screenings can be accessed with greater ease particularly as they are brought nearer to the homes of diabetic patients.

Clinical trial and preliminary findings

A clinical trial was conducted over two weeks with 20 diabetic patients at NUH and another seven healthy volunteers. The preliminary findings are as follow:

(i) Patients with diabetes mellitus tend to have a much higher mean foot temperature [26.1°C to 31.1°C] across the four sole segments, broadly divided by the main arterial supply, namely the medial plantar artery (MPA), lateral plantar artery (LPA), medial calcaneal artery (MCA), and lateral calcaneal artery (LCA) territories (Taylor & Palmer, 1987), than those without diabetes [22.3°C to 28.3°C]. On average, it is about 2.59°C higher for diabetic patients.

(ii) Ulcers and wounds can be detected in thermal images as hotspots (yellow/green areas demarcated by a red box). These hotspots tend to be at least 1.5°C higher than the mean feet segment temperature.

(iii) Based on the clinical data collected, patients with Charcot foot - a condition in which the weakening of the bones in the foot due to significant nerve damage results in foot deformity - have an average foot temperature of about 26°C to 27°C in normal areas as compared to 29°C to 30°C  in the affected areas.

(iv) There is a positive correlation between the presence of diabetic neuropathy (i.e. nerve injury) and higher feet segment temperatures. Those with neuropathy tend to have a much higher mean foot temperature [26.1°C to 31.1°C] across feet segments than those without neuropathy [25.0°C to 29.0°C].

Benefits of PODIA

Based on the preliminary findings, PODIA offers the following benefits to various stakeholders:

a. Patient

• Improves accessibility by allowing more frequent AI foot screenings done in community settings.
• Spends less time on foot screening (10-15 mins) compared to the current process (20-30 mins)
• Facilitates early detection of diabetic foot disease by improved accessibility and uptake of screening, and prompts early medical attention to prevent further foot complications

b. Nurse

• Adds objective information to support the clinical findings from routine foot examination by nurses; the data derived from machine learning algorithms provide a deeper insight of the patient’s feet conditions
• Shortens the time taken for foot screening with automation and optimises manpower by reallocating resources to higher-risk patients with complex diabetic foot conditions

c. Healthcare System

• Provides an added layer of high quality foot screening service to cope with the increasing diabetic population in Singapore and the region, especially in places with limited healthcare manpower or expertise in foot screening
• Facilitates remote disease monitoring nearer to home and improves community awareness of the importance of diabetic feet care

Looking ahead

NUHS plans to further enhance the aesthetics and functionalities of PODIA and conduct clinical trials with a larger pool of patients using PODIA. With improved device automation and validated accuracy of its artificial intelligence algorithms, the research team will look into the deployment of PODIA for highly accessible diabetic foot screening in the community.

(Please see Annex A for images of PODIA, and Annex B for quotes by SMU Assoc Prof Tan, Dr Chua and Team Live-Betics)

*** END ***

Footnotes:

[1] LAI, L. (2015). Singapore 'has 2nd-highest proportion of diabetics'. The Straits Times. Retrieved from https://www.straitstimes.com/singapore/health/singapore-has-2nd-highest-proportion-of-diabetics

[2]Agency for Care Effectiveness (ACE). (2019). Retrieved from https://www.ace-hta.gov.sg/public-data/our-guidance/Foot%20assessment%20in%20people%20with%20diabetes%20(June%202019).pdf.

[3] Amputation of limbs regarded as a last resort. Singapore General Hospital. (2018). Retrieved from https://www.sgh.com.sg/news/patient-care/amputation-of-limbs-regarded-as-a-last-resort.https://www.sgh.com.sg/news/patient-care/amputation-of-limbs-regarded-as-a-last-resort

[4] Tresierra-Ayala, M., & García Rojas, A. (2017). Association between peripheral arterial disease and diabetic foot ulcers in patients with diabetes mellitus type 2. Medicina Universitaria, 19(76), 123-126. https://doi.org/10.1016/j.rmu.2017.07.002