Leja, Marcis, Kortelainen, Juha M., Polaka, Inese, Turppa, Emmi, Mitrovics, Jan, Padilla, Marta, Mochalski, Pawel, Shuster, Gregory, Pohle, Roland, Kashanin, Dmitry, Klemm, Richard, Ikonen, Veikko, Mezmale, Linda, Broza, Yoav Y., Shani, Gidi, Haick, Hossam, Kloper, Viki, Milyutin, Yana, Abboud, Manal, Saliba, Walaa, Bdarneh, Shifaa, Khateb, Salam, Gharra, Alaa, Zuri, Liat, Vasiljevs, Edgars, Lauka, Lelde, Gasenko, Evita, Skapars, Roberts, Sivins, Armands, Bogdanova, Inga, Isajevs, Sergejs, Kikuste, Ilze, Vanags, Aigars, Tolmanis, Ivars, Kojalo, Ilona, Veliks, Viktors, Jaeschke, Carsten, Fleischer, Max, Sramek, Maria, nav Gils, Mark, Kulju, Minna, and Miettinen, Janika
Background Detection of disease by means of volatile organic compounds from breath samples using sensors is an attractive approach to fast, noninvasive and inexpensive diagnostics. However, these techniques are still limited to applications within the laboratory settings. Here, we report on the development and use of a fast, portable, and IoT-connected point-of-care device (so-called, SniffPhone) to detect and classify gastric cancer to potentially provide new qualitative solutions for cancer screening. Methods A validation study of patients with gastric cancer, patients with high-risk precancerous gastric lesions, and controls was conducted with 2 SniffPhone devices. Linear discriminant analysis (LDA) was used as a classifying model of the sensing signals obatined from the examined groups. For the testing step, an additional device was added. The study group included 274 patients: 94 with gastric cancer, 67 who were in the high-risk group, and 113 controls. Results The results of the test set showed a clear discrimination between patients with gastric cancer and controls using the 2-device LDA model (area under the curve, 93.8%; sensitivity, 100%; specificity, 87.5%; overall accuracy, 91.1%), and acceptable results were also achieved for patients with high-risk lesions (the corresponding values for dysplasia were 84.9%, 45.2%, 87.5%, and 65.9%, respectively). The test-phase analysis showed lower accuracies, though still clinically useful. Conclusion Our results demonstrate that a portable breath sensor device could be useful in point-of-care settings. It shows a promise for detection of gastric cancer as well as for other types of disease. Lay summary A portable sensor-based breath analyzer for detection of gastric cancer can be used in point-of-care settings. The results are transferrable between devices via advanced IoT technology. Both the hardware and software of the reported breath analyzer could be easily modified to enable detection and monitirng of other disease states.