Development of an explainable artificial intelligence model for Asian vascular wound images.

Chronic wounds contribute to significant healthcare and economic burden worldwide. Wound assessment remains challenging given its complex and dynamic nature. The use of artificial intelligence (AI) and machine learning methods in wound analysis is promising. Explainable modelling can help its integration and acceptance in healthcare systems. We aim to develop an explainable AI model for analysing vascular wound images among an Asian population. Two thousand nine hundred and fifty-seven wound images from a vascular wound image registry from a tertiary institution in Singapore were utilized. The dataset was split into training, validation and test sets. Wound images were classified into four types (neuroischaemic ulcer [NIU], surgical site infections [SSI], venous leg ulcers [VLU], pressure ulcer [PU]), measured with automatic estimation of width, length and depth and segmented into 18 wound and peri-wound features. Data pre-processing was performed using oversampling and augmentation techniques. Convolutional and deep learning models were utilized for model development. The model was evaluated with accuracy, F1 score and receiver operating characteristic (ROC) curves. Explainability methods were used to interpret AI decision reasoning. A web browser application was developed to demonstrate results of the wound AI model with explainability. After development, the model was tested on additional 15 476 unlabelled images to evaluate effectiveness. After the development on the training and validation dataset, the model performance on unseen labelled images in the test set achieved an AUROC of 0.99 for wound classification with mean accuracy of 95.9%. For wound measurements, the model achieved AUROC of 0.97 with mean accuracy of 85.0% for depth classification, and AUROC of 0.92 with mean accuracy of 87.1% for width and length determination. For wound segmentation, an AUROC of 0.95 and mean accuracy of 87.8% was achieved. Testing on unlabelled images, the model confidence score for wound classification was 82.8% with an explainability score of 60.6%. Confidence score was 87.6% for depth classification with 68.0% explainability score, while width and length measurement obtained 93.0% accuracy score with 76.6% explainability. Confidence score for wound segmentation was 83.9%, while explainability was 72.1%. Using explainable AI models, we have developed an algorithm and application for analysis of vascular wound images from an Asian population with accuracy and explainability. With further development, it can be utilized as a clinical decision support system and integrated into existing healthcare electronic systems.

Explainable Risk Prediction of Post-Stroke Adverse Mental Outcomes Using Machine Learning Techniques in a Population of 1780 Patients.

Post-stroke depression and anxiety, collectively known as post-stroke adverse mental outcome (PSAMO) are common sequelae of stroke. About 30% of stroke survivors develop depression and about 20% develop anxiety. Stroke survivors with PSAMO have poorer health outcomes with higher mortality and greater functional disability. In this study, we aimed to develop a machine learning (ML) model to predict the risk of PSAMO. We retrospectively studied 1780 patients with stroke who were divided into PSAMO vs. no PSAMO groups based on results of validated depression and anxiety questionnaires. The features collected included demographic and sociological data, quality of life scores, stroke-related information, medical and medication history, and comorbidities. Recursive feature elimination was used to select features to input in parallel to eight ML algorithms to train and test the model. Bayesian optimization was used for hyperparameter tuning. Shapley additive explanations (SHAP), an explainable AI (XAI) method, was applied to interpret the model. The best performing ML algorithm was gradient-boosted tree, which attained 74.7% binary classification accuracy. Feature importance calculated by SHAP produced a list of ranked important features that contributed to the prediction, which were consistent with findings of prior clinical studies. Some of these factors were modifiable, and potentially amenable to intervention at early stages of stroke to reduce the incidence of PSAMO.

A prospective study on the wound healing and quality of life outcomes of patients with venous leg ulcers in Singapore-Interim analysis at 6 month follow up.

Venous leg ulceration results in significant morbidity. However, the majority of studies conducted are on Western populations. This study aims to evaluate the wound healing and quality of life for patients with venous leg ulcers (VLUs) in a Southeast Asian population. This is a multi-centre prospective cohort study from Nov 2019 to Nov 2021. All patients were started on 2- or 4-layer compression bandage and were reviewed weekly or fortnightly. Our outcomes were wound healing, factors predictive of wound healing and the EuroQol 5-dimensional 5-level (EQ-5D-5L) health states. Within our cohort, there were 255 patients with VLU. Mean age was 65.2 ± 11.6 years. Incidence of diabetes mellitus was 42.0%. Median duration of ulcer at baseline was 0.30 years (interquartile range 0.136-0.834). Overall, the median time to wound healing was 4.5 months (95% confidence interval [CI]: 3.77-5.43). The incidence of complete wound healing at 3- and 6-month was 47.0% and 60.9%, respectively. The duration of the wound at baseline was independently associated with worse wound healing (Hazard ratio 0.94, 95% CI: 0.89-0.99, P = .014). Patients with healed VLU had a significantly higher incidence of perfect EQ-5D-5L health states at 6 months (57.8% vs 13.8%, P < .001). We intend to present longer term results in subsequent publications.

Clinical validation of a machine-learning-based handheld 3-dimensional infrared wound imaging device in venous leg ulcers.

Chronic venous insufficiency is a chronic disease of the venous system with a prevalence of 25% to 40% in females and 10% to 20% in males. Venous leg ulcers (VLUs) result from venous insufficiency. VLUs have a prevalence of 0.18% to 1% with a 1-year recurrence of 25% to 50%, bearing significant socioeconomic burden. It is therefore important for regular assessment and monitoring of VLUs to prevent worsening. Our study aims to assess the intra- and inter-rater reliability of a machine learning-based handheld 3-dimensional infrared wound imaging device (WoundAide [WA] imaging system, Konica Minolta Inc, Tokyo, Japan) compared with traditional measurements by trained wound nurse. This is a prospective cross-sectional study on 52 patients with VLUs from September 2019 to January 2021 using three WA imaging systems. Baseline patient profile and clinical demographics were collected. Basic wound parameters (length, width and area) were collected for both traditional measurements and measurements taken by the WA imaging systems. Intra- and inter-rater reliability was analysed using intra-class correlation statistics. A total of 222 wound images from 52 patients were assessed. There is excellent intra-rater reliability of the WA imaging system on three different image captures of the same wound (intra-rater reliability ranging 0.978-0.992). In addition, there is excellent inter-rater reliability between the three WA imaging systems for length (0.987), width (0.990) and area (0.995). Good inter-rater reliability for length and width (range 0.875-0.900) and excellent inter-rater reliability (range 0.932-0.950) were obtained between wound nurse measurement and each of the WA imaging system. In conclusion, high intra- and inter-rater reliability was obtained for the WA imaging systems. We also obtained high inter-rater reliability of WA measurements against traditional wound measurement. The WA imaging system is a useful clinical adjunct in the monitoring of VLU wound documentation.

Clinical validation of an artificial intelligence-enabled wound imaging mobile application in diabetic foot ulcers.

There is a lifetime risk of 15% to 25% of development of diabetic foot ulcers (DFUs) in patients with diabetes mellitus. DFUs need to be followed up on and assessed for development of complications and/or resolution, which was traditionally performed using manual measurement. Our study aims to compare the intra- and inter-rater reliability of an artificial intelligence-enabled wound imaging mobile application (CARES4WOUNDS [C4W] system, Tetsuyu, Singapore) with traditional measurement. This is a prospective cross-sectional study on 28 patients with DFUs from June 2020 to January 2021. The main wound parameters assessed were length and width. For traditional manual measurement, area was calculated by overlaying traced wound on graphical paper. Intra- and inter-rater reliability was analysed using intra-class correlation statistics. A value of <0.5, 0.5-0.75, 0.75-0.9, and >0.9 indicates poor, moderate, good, and excellent reliability, respectively. Seventy-five wound episodes from 28 patients were collected and a total of 547 wound images were analysed in this study. The median wound area during the first clinic consultation and all wound episodes was 3.75 cm2 (interquartile range [IQR] 1.40-16.50) and 3.10 cm2 (IQR 0.60-14.84), respectively. There is excellent intra-rater reliability of C4W on three different image captures of the same wound (intra-rater reliability ranging 0.933-0.994). There is also excellent inter-rater reliability between three C4W devices for length (0.947), width (0.923), and area (0.965). Good inter-rater reliability for length, width, and area (range 0.825-0.934) was obtained between wound nurse measurement and each of the C4W devices. In conclusion, we obtained good inter-rater and intra-rater reliability of C4W measurements against traditional wound measurement. The C4W is a useful adjunct in monitoring DFU wound progress.

Clinical and economic outcomes of a multidisciplinary team approach in a lower extremity amputation prevention programme for diabetic foot ulcer care in an Asian population: A case-control study.

Present guidelines recommend a multidisciplinary team (MDT) approach to diabetic foot ulcer (DFU) care, but relevant data from Asia are lacking. We aim to evaluate the clinical and economic outcomes of an MDT approach in a lower extremity amputation prevention programme (LEAPP) for DFU care in an Asian population. We performed a case-control study of 84 patients with DFU between January 2017 and October 2017 (retrospective control) vs 117 patients with DFU between December 2017 and July 2018 (prospective LEAPP cohort). Comparing the clinical outcomes between the retrospective cohort and the LEAPP cohort, there was a significant decrease in mean time from referral to index clinic visit (38.6 vs 9.5 days, P < .001), increase in outpatient podiatry follow-up (33% vs 76%, P < .001), decrease in 1-year minor amputation rate (14% vs 3%, P = .007), and decrease in 1-year major amputation rate (9% vs 3%, P = .05). Simulation of cost avoidance demonstrated an annualised cost avoidance of USD $1.86m (SGD $2.5m) for patients within the LEAPP cohort. In conclusion, similar to the data from Western societies, an MDT approach in an Asian population, via a LEAPP for patients with DFU, demonstrated a significant reduction in minor and major amputation rates, with annualised cost avoidance of USD $1.86m.

Risk Prediction of Diabetic Foot Amputation Using Machine Learning and Explainable Artificial Intelligence.

BACKGROUND: Diabetic foot ulcers (DFUs) are serious complications of diabetes which can lead to lower extremity amputations (LEAs). Risk prediction models can identify high-risk patients who can benefit from early intervention. Machine learning (ML) methods have shown promising utility in medical applications. Explainable modeling can help its integration and acceptance. This study aims to develop a risk prediction model using ML algorithms with explainability for LEA in DFU patients. METHODS: This study is a retrospective review of 2559 inpatient DFU episodes in a tertiary institution from 2012 to 2017. Fifty-one features including patient demographics, comorbidities, medication, wound characteristics, and laboratory results were reviewed. Outcome measures were the risk of major LEA, minor LEA and any LEA. Machine learning models were developed for each outcome, with model performance evaluated using receiver operating characteristic (ROC) curves, balanced-accuracy and F1-score. SHapley Additive exPlanations (SHAP) was applied to interpret the model for explainability. RESULTS: Model performance for prediction of major, minor, and any LEA event achieved ROC of 0.820, 0.637, and 0.756, respectively, with XGBoost, XGBoost, and Gradient Boosted Trees algorithms demonstrating best results for each model, respectively. Using SHAP, key features that contributed to the predictions were identified for explainability. Total white cell (TWC) count, comorbidity score and red blood cell count contributed highest weightage to major LEA event. Total white cell, eosinophils, and necrotic eschar in the wound contributed most to any LEA event. CONCLUSIONS: Machine learning algorithms performed well in predicting the risk of LEA in a patient with DFU. Explainability can help provide clinical insights and identify at-risk patients for early intervention.

Getting a Foothold on Diabetic Foot Disease-Outcomes of a Multidisciplinary Clinical Pathway for Inpatient Diabetic Foot Care: A 17-Year Institutional Review.

INTRODUCTION: Diabetes foot disease (DFD) contributes to poor quality of life, clinical and economic burden. Multidisciplinary diabetes foot teams provide prompt access to specialist teams thereby improving limb salvage. We present a 17-year review of an inpatient multidisciplinary clinical care path (MCCP) for DFD in Singapore. METHODS: This was a retrospective cohort study of patients admitted for DFD and enrolled in our MCCP to a 1700-bed university hospital from 2005 to 2021. RESULTS: There were 9279 patients admitted with DFD with a mean of 545 (±119) admissions per year. The mean age was 64 (±13.3) years, 61% were Chinese, 18% Malay and 17% Indian. There was a higher proportion of Malay (18%) and Indian (17%) patients compared to the country's ethnic composition. A third of the patients had end stage renal disease and prior contralateral minor amputation. There was a reduction in inpatient major lower extremity amputation (LEA) from 18.2% in 2005 to 5.4% in 2021 (odds ratio 0.26, 95% confidence interval 0.16-0.40, P < .001) which was the lowest since pathway inception. Mean time from admission to first surgical intervention was 2.8 days and mean time from decision for revascularization to procedure was 4.8 days. The major-to-minor amputation rate reduced from 1.09 in 2005 to 0.18 in 2021, reflecting diabetic limb salvage efforts. Mean and median length of stay (LOS) for patients in the pathway was 8.2 (±14.9) and 5 (IQR = 3) days, respectively. There was a gradual trend of increase in the mean LOS from 2005 to 2021. Inpatient mortality and readmission rate was stable at 1% and 11%. CONCLUSION: Since the institution of a MCCP, there was a significant improvement in major LEA rate. An inpatient multidisciplinary diabetic foot care path helped to improve care for patients with DFD.

Influence of blood pressure on retinal vascular caliber in young children.

OBJECTIVE: Prior studies have reported that blood pressure (BP) has a significant influence on retinal vascular caliber both in adults and children aged 6 years and older. This study aimed to examine the association between BP and retinal vascular caliber in Singapore Chinese preschoolers 4 to 5 years of age. DESIGN: Population-based, cross-sectional study. PARTICIPANTS: A total of 385 eligible and healthy Singapore Chinese children 4 to 5 years of age who were recruited in The Strabismus, Amblyopia and Refractive Error Study in Singaporean Chinese Preschoolers from May 2006 through October 2008 underwent BP measurements and retinal photography. METHODS: According to standard protocols, BP was measured with an automatic Omron sphygmomanometer (Omron HEM 705 LP, Omron Healthcare, Inc., Bannockburn, IL) and a retinal photograph was obtained with a Canon 45° digital retinal camera (Model CR6-NM45, Canon, Inc., Tokyo, Japan) after pupil dilation. Anthropometric and optical biometric measurements such as height, weight, and axial length were obtained also. Information regarding sociodemographic status and child birth information was supplied by parents in either English or Chinese questionnaires. MAIN OUTCOME MEASURES: The computer imaging program was used to measure the caliber of all retinal arterioles and venules located in zone B. The central retinal arteriolar equivalent and central retinal venular equivalent were estimated by using a revised Knudtson-Parr-Hubbard formula. RESULTS: The mean retinal arteriolar and venular calibers were 156.19 µm and 220.01 µm in boys and 161.97 µm and 224.22 µm in girls. Higher systolic BP was associated with narrower retinal arterioles. After adjusting for age, gender, father's education, body mass index, birth weight, axial length, and caliber of the fellow retinal vessel, each 10-mmHg increase in systolic BP was associated with 2.00 µm (95% confidence interval, 0.39-3.61; P = 0.02) of retinal arteriolar narrowing and 2.51 µm (95% confidence interval, 0.35-4.68; P = 0.02) of retinal venular widening. However, neither diastolic BP nor mean arterial BP was associated with retinal arteriolar or venular caliber. CONCLUSIONS: In very young children 4 to 5 years of age, higher systolic BP was associated with narrower retinal arterioles and wider retinal venules. This suggests that elevated BP may affect the retinal microvasculature from early childhood.

Automated detection of glaucoma using optical coherence tomography angiogram images.

Glaucoma is a malady that occurs due to the buildup of fluid pressure in the inner eye. Detection of glaucoma at an early stage is crucial as by 2040, 111.8 million people are expected to be afflicted with glaucoma globally. Feature extraction methods prove to be promising in the diagnosis of glaucoma. In this study, we have used optical coherence tomography angiogram (OCTA) images for automated glaucoma detection. Ocular sinister (OS) from the left eye while ocular dexter (OD) were obtained from right eye of subjects. We have used OS macular, OS disc, OD macular and OD disc images. In this work, local phase quantization (LPQ) technique was applied to extract the features. Information fusion and principal component analysis (PCA) are used to combine and reduce the features. Our method achieved the highest accuracy of 94.3% using LPQ coupled with PCA for right eye optic disc images with AdaBoost classifier. The proposed technique can aid clinicians in glaucoma detection at an early stage. The developed model is ready to be tested with more images before deploying for clinical application.