Connect the Dots: Towards a data-driven and multidisciplinary approach for the management of anastomotic leakage

An anastomosis is created to restore gastrointestinal continuity after a sick part within this this track is removed. Failure of this anastomosis is called colorectal anastomotic leakage (CAL) in colorectal surgery and a postoperative pancreatic fistula (POPF) in pancreatic surgery. Both a CAL and POPF are major impact complications on both the short- and long-term outcome, as well as oncological outcomes. Several mitigations strategies have been proposed to diminish the incidence or severity of anastomotic leakage. A diverting stoma is a commonly used mitigation strategy in colorectal surgery. However, it is not fully understood which patients benefit most from a diverting stoma. A diverting stoma is a commonly employed mitigation strategy in colorectal surgery. However, it remains unclear which patients benefit the most from a diverting stoma. The integration of AI is becoming increasingly prevalent in the development of prediction models within the healthcare sector. In Chapter 6, numerous challenges associated with the implementation of AI were thoroughly examined, encompassing aspects from initial development to practical application. The chapter offered insightful recommendations concerning the utilization of AI for predicting postoperative complications following major abdominal surgery. The ability to POPF accurately holds significant potential in assisting surgeons with making individualized treatment decisions. Chapter 7 delved into a systematic review, assessing the predictive capabilities of radiomic features in anticipating POPF. The evaluation encompassed considerations of methodological quality, reporting transparency, and the risk of bias. The findings underscored the promising potential of radiomic features in accurately predicting POPF. However, it was noted that the existing evidence supporting their clinical application was still lacking. Moreover, the operating room stands out as a data-rich environment where AI can play a pivotal role. Chapter 8 provided a comprehensive review of current AI applications in perioperative care, highlighting their focus on supporting decision-making during the perioperative phase and enhancing surgical skills and safety. This section emphasized the paramount importance of standardizing data, clinically validating AI systems, meticulously evaluating the implementation process, and establishing ethical and legal frameworks. Part III: Development of prediction models In chapter 10, a groundbreaking model, the Radiomics preoperative-Fistula Risk Score (RAD-FRS), was introduced for predicting postoperative pancreatic fistula (POPF). This novel model derived radiomic features from preoperative CT scans of patients following pancreatoduodenectomy. Successfully developed and internally validated in the Netherlands, it underwent external validation in Italy, demonstrating a robust performance with an area under the curve of 0.81 in the external validation cohort. The potential integration of this model with hospital CT reporting systems could enhance patient counseling before surgery.