The Use of Mobile Thermal Imaging and Deep Learning for Prediction of Surgical Site Infection

Abstract

The ability to detect surgical site infections (SSI) is a critical need for healthcare worldwide, but is especially important in low-income countries, where there is limited access to health facilities and trained clinical staff. In this paper, we present a new method of predicting SSI using a thermal image collected with a smart phone. Machine learning algorithms were developed using images collected as part of a clinical study that included 530 women in rural Rwanda who underwent cesarean section surgery. Thermal images were collected approximately 10 days after surgery, in conjunction with an examination by a trained doctor to determine the status of the wound (infected or not). Of the 530 women, 30 were found to have infected wounds. The data were used to develop two Convolutional Neural Net (CNN) models, with special care taken to avoid overfitting and address the problem of class imbalance in binary classification. The first model, a 6-layer naïve CNN model, demonstrated a median accuracy of AUC=0.84 with sensitivity=71% and specificity=87%. The transfer learning CNN model demonstrated a median accuracy of AUC=0.90 with sensitivity =95% and specificity=84%. To our knowledge, this is the first successful demonstration of a machine learning algorithm to predict surgical infection using thermal images alone. Clinical Relevance-This work establishes a promising new method for automated detection of surgical site infection.

Risk predictions of surgical wound complications based on a machine learning algorithm: A systematic review

“…Within this inquiry, ML was strategically applied to diverse surgical wound care. Six studies have delved into the nuanced evaluation of specific surgical procedures, encompassing (1) Cardiothoracic, (2) Caesarean, (3) total abdominal colectomy, (4) Burn plastic surgery, (5) head and neck cancer and (6) facial plastic surgery 17–22 . Moreover, a tripartite evaluation was conducted in one study, appraising wounds resulting from (1) laparotomy, (2) minimal invasive surgery and (3) hernia repair 23 .…”

“…The salient application of ML has been discerned in the meticulous assessment of SSIs, with seven studies employing this methodology for prognostication and evaluation 17–19,22–25 . Two studies have extended the use of ML for the nuanced assessment of burn grade diagnosis and wound classification 20,21 .…”

Prevalence of post‐caesarean section surgical site infections in Rwanda: A systematic review and meta‐analysis

“…Nine studies were cohort studies, 1,9,15–17,19–21,24 six were randomized controlled trials 10,13,18,22,25,26 and two were cross‐sectional studies 14,23 . Five studies had sample sizes above 650 participants 9,13,16,21,26 while 12 studies had sample sizes below 650 participants 1,10,14,15,17–20,22–25 . Eight studies were published from 2021 to 2023 9,13,14,16,17,20,21,24 while nine studies were published before 2021 (from 2020 to 2014) 1,10,15,18,19,22,23,25,26 .…”

Image-based surgical site infection algorithms to support home-based post-cesarean monitoring: Lessons from Rwanda

“…We have demonstrated that home-based follow-up by CHWs is feasible and acceptable [5,6], but our attempts for simple clinical screenings or sending pictures to general practitioners has resulted in low SSI diagnostic accuracy [5][6][7][8]. In contrast, our machine learning image-based diagnostic algorithms are yielding higher accuracy: 81.3% sensitivity and 65.3% specificity for visible image algorithms [9], increasing to 97% sensitivity and 87% specificity when using computer vision techniques and color calibration in the image processing pipeline, and 95% specificity and 84% specificity for thermal image algorithms [10]. We believe these algorithms hold huge potential for remote SSI diagnostics by health workers in the community in lowand middle-income countries (LMICs), and here we offer some lessons learned to improve others' efforts as they work towards this goal.…”