Evaluating the performance of a low-cost mobile phone attachable microscope in cervical cytology

Abstract

Background: Cervical cancer remains a global health problem especially in remote areas of developing countries which have limited resources for cervical cancer screening. In this study, we evaluated the performance of a lowcost, smartphone attachable paper-based microscope when used for classifying images of cervical cytology. Methods: Cervical cytology samples included: 10 Normal, 10 Low-grade squamous intraepithelial lesion (LSIL), 10 High-grade squamous intraepithelial lesion (HSIL), and 10 Malignant Pap Smears. The agreement between conventional microscopy vs. Foldscope imaging was calculated using a weighted kappa coefficient. A confusion matrix was created with three classes: Normal, LSIL, and HSIL/malignant, to evaluate the performance of the Foldscope by calculating the accuracy, sensitivity, and specificity. Results: We observed a kappa statistic of 0.68 for the agreement. This translates into a substantial agreement between the cytological classifications by the Foldscope vs. conventional microscopy. The accuracy of the Foldscope was 80%, with a sensitivity and specificity of 85 and 90% for the HSIL/Mal category, 80 and 83.3%, for LSIL, and 70 and 96.7% for Normal. Conclusions: This study highlights the usefulness of the Foldscope in cervical cytology, demonstrating it has substantial agreement with conventional microscopy. Its use could improve cytologic interpretations in underserved areas and, thus, improve the quality of cervical cancer screening. Improvements in existing limitations of the device, such as ability to focus, could potentially increase its accuracy.

Impact of mobile devices on cancer diagnosis in cytology

“…Most of the studies in our analysis dealt with static images. 26,27,30,31,34,35,40,42 Three studies evaluated the live transmission of dynamic images 25,36,37 and only one assessed WSI. 32 Static images are relatively easy to transmit whereas dynamic images often demand more bandwidth.…”

“…Comparing Android and iOS with sWSI, both systems were satisfactory, with a nonsignificantly different kappa value (average kappa value for Android: 0.70; average kappa value for iOS: 0.72). Naqvi et al 31 , 24 and 250 images of atypical cells, the final accuracy of the proposed ANN was 90%, with a sensitivity of 89% and a specificity of 100%, increasing overall accuracy by 30% compared to the manual method.…”

“…The 18 included studies comprised 11 full-text articles [25][26][27][28][29][30][31][32][33][34][35] and 7 abstracts/posters. [36][37][38][39][40][41][42] They were published in the time span 2012-2020 and were geographically distributed as follows: 10 from North America, 25,28,31,33,[36][37][38][39][40][41] four from Asia, 26,27,32,35 two from Europe, 29,34 one from South America 30 ; one non stated. 42 These studies involved a total of 684 patients, with the number of subjects per study ranging from 20 to 172.…”

“…Mobile devices in these studies were employed either for screening programs (e.g., Pap tests, oral samples) or for interpreting FNA-derived material. Two studies focused on urinary cytology, 26,34 five dealt with cervical cytology, 31,32,34,40,42 and another two with oral cancer screening. 27,35 Several kinds of smartphones and tablets were used in these works including a Samsung Galaxy Note II, A7, S7, and M20 (Samsung Electronics, Seoul, South Korea), an iPhone 4S with a SkyLight adapter (Apple Inc., Cupertino, CA, USA), an iPhone 6 and 7 (Apple Inc., Cupertino, CA, USA), a Huawei Mate 8 (Huawei Technologies Co. Ltd., Shenzhen, Guangdong, China), a Xiaomi 5S (Xiaomi Inc, Beijing, China), and an iPad Mini 2 (Apple Inc., Cupertino, CA, USA) which were mainly employed for acquiring static images, 26,31,32,34,35,40,42 but also for creating scalable whole slide imaging (sWSI) 32 and analyzing them with artificial neural networks (ANN).…”

“…Two studies focused on urinary cytology, 26,34 five dealt with cervical cytology, 31,32,34,40,42 and another two with oral cancer screening. 27,35 Several kinds of smartphones and tablets were used in these works including a Samsung Galaxy Note II, A7, S7, and M20 (Samsung Electronics, Seoul, South Korea), an iPhone 4S with a SkyLight adapter (Apple Inc., Cupertino, CA, USA), an iPhone 6 and 7 (Apple Inc., Cupertino, CA, USA), a Huawei Mate 8 (Huawei Technologies Co. Ltd., Shenzhen, Guangdong, China), a Xiaomi 5S (Xiaomi Inc, Beijing, China), and an iPad Mini 2 (Apple Inc., Cupertino, CA, USA) which were mainly employed for acquiring static images, 26,31,32,34,35,40,42 but also for creating scalable whole slide imaging (sWSI) 32 and analyzing them with artificial neural networks (ANN). 27 Other six papers focused on the utility of mobile devices for ROSE of FNA cytological specimens, 25,26,30,34,36 relying on static or dynamic images generated by iOS (iPhone 5, iPad) and Android (Samsung Galaxy s7) supported devices and remotely transmitted via applications such as Facetime (Apple Inc., Cupertino, CA, USA) and WhatsApp (Facebook Inc., Menlo Park, CA, USA).…”