Abstract—Small bowel obstruction is a common acute abdomen condition which can play a role in death. Small bowel strangulation (SBS) refers to a small bowel obstruction associated with bowel ischemia. Patients with SBS need emergency surgery because the ischemic small bowel can become necrotic in a short time, causing sepsis and death. Nowadays the “gold standard” for diagnosing SBS is via computed tomography (CT) scanned images. However, an easier way to detect SBS is desired among emergency medicine physicians. Thus, we tried to develop a rapid test using circulating cell-free DNA (ccfDNA). ccfDNA is part of the DNA from a cell that died because of apoptosis or necrosis. The size of ccfDNA varies depending on the origination of cell death. If a patient has SBS, long-size ccfDNA would appear in the peripheral blood. We used data including the concentration of ccfDNA in the blood of certain patients as training data to make a support vector machine, a decision tree, and a learned random forest. We evaluated these classifiers using leave-one-out cross-validation. These machine-learning methods performed well. In addition, the decision tree and random forest results indicate that long-size ccfDNA is important for classifying SBS. In this paper, we demonstrate that machine learning can be an alternative method for detecting SBS and that the concentration of ccfDNA, especially long ccfDNA, contributes to detecting SBS.
Index Terms—Bioinformatics, circulating cell-free DNA, machine learning, small bowel obstruction.
Kazutaka Nishiwaki, and Hayato Ohwada are with Tokyo University of Science, Noda city, Chiba, 278-8510 Japan (e-mail: 7416634@ed.tus.ac.jp, ohwada@rs.tus.ac.jp).
Takeshi Yamada, Takuma Iwai, Goro Takahashi, Kouki Takeda, and Eiji Uchida are with Nippon Medical School, Bunkyo-ku, Tokyo, 113-3603 Japan. (e-mail: y-tak@nms.ac.jp).
Cite: Kazutaka Nishiwaki, Takeshi Yamada, Takuma Iwai, Goro Takahashi, Eiji Uchida and Hayato Ohwada, "Detecting Small Bowel Strangulation using Circulating Cell-Free DNA with Machine Learning," International Journal of Machine Learning and Computing vol. 7, no. 3, pp. 35-39, 2017.