{"id":13922,"date":"2025-02-27T12:45:37","date_gmt":"2025-02-27T10:45:37","guid":{"rendered":"https:\/\/centerforsurgicalscience.dk\/?page_id=13922"},"modified":"2025-03-05T20:35:36","modified_gmt":"2025-03-05T18:35:36","slug":"our-ai-model","status":"publish","type":"page","link":"https:\/\/centerforsurgicalscience.dk\/da\/our-ai-model\/","title":{"rendered":"Our AI model"},"content":{"rendered":"
\n\t\t\t\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
<\/div>\n\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t

OUR
AI PREDICTION MODEL<\/h2>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
<\/div>\n\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t
\n\t\t\t\t\t\t\t
<\/div>\n\t\t\t\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t
\n\t\t\t

FROM CONCEPT TO CLINIC<\/h2>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t
\n\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t
\n\t\t\t

AID-SURG TO AID-PRO-CRC<\/h2>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
<\/div>\n\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
<\/div>\n\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
<\/div>\n\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\t\"\"\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
<\/div>\n\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t

HOW DOES THE AI MODEL WORK?<\/h2>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t

The AIDPRO manual CRC is a clinical decision support tool designed for colorectal surgeons managing perioperative care for patients undergoing elective curative surgery for colorectal cancer. It uses a prediction model trained on national Danish data to estimate a patient\u2019s one-year mortality risk and categorizes them into risk strata to guide perioperative treatment planning. While the tool provides recommendations, surgeons retain full decision-making authority and can override suggestions when necessary.<\/span><\/p>

\u00a0<\/p>

The algorithm was developed using extensive patient data from national health registries, with a structured selection process ensuring relevant clinical variables are included. Performance <\/span>validation was conducted to ensure accuracy and reliability. The tool is accessed through a secure platform, requiring manual data input by authorized colorectal surgeons.<\/span><\/p>

\u00a0<\/p>

A dedicated graphical user interface (GUI) called MDTplus has been designed to integrate with the AIDPRO manual CRC. MDTplus serves as the user interface, providing a secure and structured environment for surgeons to input patient data and receive risk estimations. The system operates as a web, ensuring seamless accessibility during multidisciplinary team meetings and preoperative consultations. The platform is designed to support, rather than replace, clinical judgment in colorectal cancer surgery planning.<\/span><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t

\n\t\t\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
<\/div>\n\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t

A HUGE ECONOMIC IMPACT<\/h2>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t

The economic cost of cancer is increasing, and major drivers include frailty and postoperative complications. Temporary results of a modeling study, found that implementing personalized perioperative bundle care based on an AI-based decision-support tool is a dominant strategy compared to the current standard of care, increasing health related quality of life and introducing savings, in the first year after surgery, driven by fewer complications. In fact, we showed in a cost-effectiveness analysis cost savings of 2. 847.59 USD (2.810.32 to 2. 884.00) per patient during the first year following surgery!

Thus, the net effect of the implementation of complication preventive strategies, such as the present study, will free clinical labor resources within the healthcare system, allowing staff to solve other tasks. The analysis did not include indirect effects, such as the expected positive effect on return to intended oncological treatment, which is expected to improve health-related quality-of-life, through improved long-term oncological outcomes.
<\/span><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t

\n\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
<\/div>\n\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\t\"\"\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
<\/div>\n\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t
\n\t\t\t

THE STORY<\/h2>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\t\"\"\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t

Despite significant advances during the last decades, mortality and morbidity following surgery for colorectal cancer remains high (1). In the past, focus has primarily been on implementing guidelines and national quality programs, overlooking the patient-level risk factors such as high age, comorbidities, and frailty, which also play a crucial role in determining outcomes following cancer treatment (2).<\/span><\/p>

\u00a0<\/p>

Better patient selection and individual treatment is therefore necessary to specifically tailor and target the phenotype of the patient and the tumor through differential, perioperative treatments aiming at optimizing postoperative outcomes. At Center for Surgical Science (CSS), Department of Surgery, Zealand University Hospital (ZUH), we have successfully established an artificial intelligence (AI)-based risk prediction model, which have shown a positive effect not only for the individual patients with colorectal cancer, but also on the Danish socio-economy (publication in submission).<\/span><\/p>

\u00a0<\/p>

The prediction model is built on data from all patients undergoing curative intended surgery for colorectal cancer in Denmark from 2011-19. This includes standardized and harmonized data sources from Danish Colorectal Cancer Group (DCCG) database (<\/span>www.DCCG.dk<\/a>), Danish National Patient Registry (DNPR), Danish Register of Laboratory Results for Research database (RLRR) and Medicinal

<\/span><\/p>

products Statistics database (MED). The model is one of the first nationwide tests of independent public-developed and public-driven clinical research health technology aiming at improving patient outcome. The data-driven AI-model was implemented as an evidence-based decision-support tool The prediction model was implemented as a pilot study, AID-SURG, at Department of Surgery on February 1, 2023, to stratify patients with colorectal cancer into four risk profiles based on their predicted 1-year mortality. According to the estimated risk and the clinical assessment, each patient is allocated to a specific treatment trajectory with tailored evidence-based perioperative interventions in addition to standard perioperative treatment.<\/span><\/p>

\u00a0<\/p>

In this phase 2 clinical trial, the incidence proportion of Comprehensive Complication Index (CCI) > 20 was reduced from 28% in the control group to 19.1% in the treatment group, adjusted odds ratio (OR) of 0.63 (95% CI: 0.42-0.92, p=0.02). The incidence of any medical complication was 23.7% in the treatment group and 37.3% in the control group; OR of 0.53 (95% CI: 0.36-0.76, p<0.001). Thus, our data show that individualized treatment strategies, before, during, and after surgery, based on the individual CRC patient’s risk of 1-year mortality determined by the AI model, significantly reduced the occurrence of complications and readmission after CRC surgery.<\/span><\/p>



<\/span><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t

\n\t\t\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t
\n\t\t\t

CURRENT STATUS & FUTURE IMPLEMENTATION PLANS<\/h2>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\t\"\"\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
<\/div>\n\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
<\/div>\n\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t

By implementing the AI-prediction model (AID-SURG pilot study) at Department of Surgery, ZUH, individualized perioperative treatment strategies, have been introduced to all patients diagnosed with colorectal cancer at ZUH. The model has shown promising results and robust feasibility in a single CRC center in Denmark. However, it is crucial to investigate and evaluate the effectiveness of decision-making support with AI-based risk stratification prediction models on a national level in a randomized setup to validate the usefulness of the device.\u00a0<\/span><\/p>

The plan is therefore in the spring of 2025 to implement AI-prediction model at 7 different surgical departments across Denmark as the AIDPRO-CRC study. The purpose of the AIDPRO-CRC is to establish level one evidence for AI-augmented decision support in selecting patients for perioperative optimization pathways, by directly comparing the ability of an AI-algorithm against clinical experts in risk-stratification of patients.<\/span><\/p>

As prehabilitation differs between hospitals, the first step in initiating the AIDPRO-CRC study is to support a systematic uniform prehabilitation effort across participating hospitals. Based on the experience gained from the AID-SURG study, and the national guidelines from DCCG, a description of the prehabilitation effort have been created in collaboration with a panel of multidisciplinary partners from each site. This step has been followed by close support in implementing prehabilitation, ensuring that each site have implemented systematic prehabilitation before testing the AI-algorithm.<\/p>

Moreover, the plan is to create a robust platform for integrating translational data in the future selection of patients for oncological pathways aiming towards benefitting long term oncological outcomes for the individual patient, by integrating both oncological and surgical decisions.<\/span><\/p>

\u00a0<\/span><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t

\n\t\t\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t
\n\t\t\t

OUR AI MODEL IN THE MEDIAS<\/h2>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t

Due to its significant impact on health economics and patient well-being, our AI prediction model has attracted considerable attention from politcians, the media and the public.<\/p>\n

Below, you can find (Danish) articles related to the model:<\/p>\n