Complication Prediction after Esophagectomy with Machine Learning

Abstract

Esophageal cancer can be treated effectively with esophagectomy; however, the postoperative complication rate is high. In this paper, we study to what extent machine learning methods can predict anastomotic leakage and pneumonia up to two days in advance. We use a dataset with 417 patients who underwent esophagectomy between 2011 and 2021. The dataset contains multimodal temporal information, specifically, laboratory results, vital signs, thorax images, and preoperative patient characteristics. The best models scored mean test set AUROCs of 0.87 and 0.82 for leakage 1 and 2 days ahead, respectively. For pneumonia, this was 0.74 and 0.61 for 1 and 2 days ahead, respectively. We conclude that machine learning models can effectively predict anastomotic leakage and pneumonia after esophagectomy.

Keywords: clinical decision support; esophagectomy; multimodal machine learning; temporal learning.

Figure A1

(Boxplots comparing laboratory values and vital signs (rows) with three complication types (columns). Amylase values were log-10 transformed.

Figure 1

Histograms illustrating the distribution of postoperative complications. Pneumonia tends to occur early, while anastomotic leakage incidence is more evenly distributed. Complications lead to a longer hospital stay.

Figure 2

Boxplots comparing three laboratory results (rows) with three complication types (columns). Amylase values were log-10 transformed.

Figure 3

Sequence length distributions before and after data preparation.

Figure 4

Overview of unimodal (temporal) models and how these were fused in multimodal models. Laboratory results: C-reactive protein (CRP), amylase (AMY), leucocytes (LEUC). Vital signs: temperature (Temp), heart rate (HR), respiratory rate (RR), systolic blood pressure (BP). Late fusion: Concatenation of unimodal predictions. Mid-fusion: Concatenation of unimodal hidden layer features. LSTM: long short-term memory unit.