Smartphone-based optical devices

Modification of the functionalities of smartphones allows for the rapid development of smartphone-based optical devices to visualize and quantify parasite-derived morphological products in samples. Their portability makes them suitable for parasitology field surveys. In the past decade, numerous studies have been carried out using smartphone-based optical devices of which some are limited to proof-of-concept while others have been validated through a field evaluation (Bogoch et al., Reference Bogoch, Andrews, Speich, Utzinger, Ame, Ali and Keiser2013; Bogoch et al., Reference Bogoch, Coulibaly, Andrews, Speich, Keiser, Stothard, N'goran and Utzinger2014b; Switz et al., Reference Switz, D'ambrosio and Fletcher2014; Ephraim et al., Reference Ephraim, Duah, Cybulski, Prakash, D'ambrosio, Fletcher, Keiser, Andrews and Bogoch2015; Coulibaly et al., Reference Coulibaly, Ouattara, D'ambrosio, Fletcher, Keiser, Utzinger, N'goran, Andrews and Bogoch2016b; Sowerby et al., Reference Sowerby, Crump, Johnstone, Krause and Hill2016; Bogoch et al., Reference Bogoch, Koydemir, Tseng, Ephraim, Duah, Tee, Andrews and Ozcan2017; Koydemir et al., Reference Koydemir, Coulibaly, Tseng, Bogoch and Ozcan2019; Yang et al., Reference Yang, Bakhtari, Langdon-Embry, Redwood, Lapierre, Rakotomanga, Rafalimanantsoa, Santos, Vigan-Womas and Knoblauch2019; Armstrong et al., Reference Armstrong, Harris, D'ambrosio, Coulibaly, Essien-Baidoo, Ephraim, Andrews, Bogoch and Fletcher2022). Table 2 recapitulates the diagnostic characteristics of smartphone-based optical devices for the detection of STHs and Schistosoma spp. and the corresponding TRL assigned by the authors of this review.

In 4 published studies, 5 smartphone-based optical devices for the detection of helminth eggs in urine and/or stool were reported as experimental proof-of-concept studies (Bogoch et al., Reference Bogoch, Andrews, Speich, Utzinger, Ame, Ali and Keiser2013; Switz et al., Reference Switz, D'ambrosio and Fletcher2014; Sowerby et al., Reference Sowerby, Crump, Johnstone, Krause and Hill2016; Yang et al., Reference Yang, Bakhtari, Langdon-Embry, Redwood, Lapierre, Rakotomanga, Rafalimanantsoa, Santos, Vigan-Womas and Knoblauch2019). First, Bogoch et al. carried out a study within an ongoing clinical trial from which 199 Kato–Katz (KK) thick smears were analysed for the presence of STH eggs using a smartphone-based microscope with conventional light microscopy as a reference (Bogoch et al., Reference Bogoch, Andrews, Speich, Utzinger, Ame, Ali and Keiser2013). The device used for this study consisted of a 3 mm ball lens that was mounted in alignment with the lens of an iPhone 4S using a double-sided tape. The light source was a separate handheld incandescent lamp powered by 1 AA battery. For this design, a moderate sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) for the detection of STH eggs were observed (Table 2). Compared to conventional light microscopy, the authors observed a higher sensitivity of 93.3% (66.0–99.7) for moderate and high infection intensities of A. lumbricoides. According to the authors, this device is a promising point-of-care diagnostic tool for helminthic infections. However, this design lacks a functional sample stage which makes slide scanning unfeasible in practice.

Second, Switz et al. carried out a proof-of-concept study using a smartphone microscope to image the eggs of A. lumbricoides in stool samples over a wide field of view (FOV) (Switz et al., Reference Switz, D'ambrosio and Fletcher2014). The smartphone microscope consisted of an iPhone 4S coupled to a reversed iPhone 4S lens and embodied by a frame with a screw-based focusing system, an x–y stage and a light-emitting diode (LED) light source. The system set-up provided high-resolution images with significantly large FOV. While the authors conclude that the availability and use of this device in resource-limited settings will increase, its robustness and diagnostic performance at patient level are still unknown and require more testing.

Third, Sowerby et al. carried out a proof-of-concept study that uses the flotation technique to accumulate A. lumbricoides ova in a single FOV for imaging using a smartphone-based microscope (Sowerby et al., Reference Sowerby, Crump, Johnstone, Krause and Hill2016). Stool was collected from a person known to be positive for ascariasis. The smartphone-based microscope consisted of a Nokia Lumia 1020 smartphone, an external double convex objective lens, a cassette containing a sample well and a cassette rack with a light source – all coupled together such that the sample well, external lens and the lens of the smartphone are aligned. Several images of a single FOV at different focal lengths were captured for image reconstruction and visualization of A. lumbricoides eggs. According to the authors this device has potential as a point-of-care device for the detection of helminth species upon further development. The sample preparation method and the processing of images through reconstruction clearly do not support the design as a point-of-care test (Peeling and Mabey, Reference Peeling and Mabey2010; Osredkar, Reference Osredkar2017), especially with stool material. Furthermore, functionalities of point-of-care devices were not examined for this prototype.

The fourth proof-of-concept by Yang et al. focused on the development and evaluation of a smartphone-based microscope with and without applying the Kankanet AI algorithm, for the detection of A. lumbricoides, T. trichiura and hookworm eggs. KK, spontaneous sedimentation technique in tube, and merthiolate–iodine–formaldehyde slides were analysed with standard microscopy as a reference (Yang et al., Reference Yang, Bakhtari, Langdon-Embry, Redwood, Lapierre, Rakotomanga, Rafalimanantsoa, Santos, Vigan-Womas and Knoblauch2019). This study was embedded in an ongoing epidemiology survey. The smartphone based-microscope setup consisted of a smartphone mounted on a tripod and connected to a USB video class microscope mounted on a translational stage that moves in the x–y direction. The number of samples included in the study was not stated and evaluation of the device prototype with Kankanet was based on the detection of eggs at an image level. Sensitivity and specificity were determined based on 124 images captured with the device setup as shown in Table 2. The Kankanet AI yielded a lower sensitivity and specificity for the detection of these parasite eggs compared to the microscopist (see Table 2). According to the authors, the setup with the Kankanet AI, if further developed, can achieve a performance comparable to microscopy, and the device could serve as a point-of-care diagnostic procedure. To support this claim, a more thorough performance evaluation of this tool should be carried out at a sample level.

In addition to the aforementioned proof-of-concept studies, 5 different smartphone-based optical devices were field validated for the detection of helminth eggs in urine and/or stool with 2 of the devices evaluated twice in different settings (Bogoch et al., Reference Bogoch, Coulibaly, Andrews, Speich, Keiser, Stothard, N'goran and Utzinger2014b; Ephraim et al., Reference Ephraim, Duah, Cybulski, Prakash, D'ambrosio, Fletcher, Keiser, Andrews and Bogoch2015; Coulibaly et al., Reference Coulibaly, Ouattara, D'ambrosio, Fletcher, Keiser, Utzinger, N'goran, Andrews and Bogoch2016b; Bogoch et al., Reference Bogoch, Koydemir, Tseng, Ephraim, Duah, Tee, Andrews and Ozcan2017; Koydemir et al., Reference Koydemir, Coulibaly, Tseng, Bogoch and Ozcan2019; Armstrong et al., Reference Armstrong, Harris, D'ambrosio, Coulibaly, Essien-Baidoo, Ephraim, Andrews, Bogoch and Fletcher2022). First, a cohort validation study by Bogoch et al. was integrated into an ongoing study in South Côte d'Ivoire. In this study, a single KK thick smear per sample from a total of 164 stool samples were examined with a smartphone-based microscope to detect the eggs of T. trichiura and S. mansoni, with conventional light microscopy as the reference (Bogoch et al., Reference Bogoch, Coulibaly, Andrews, Speich, Keiser, Stothard, N'goran and Utzinger2014b). The smartphone-based microscope consisted of a 3 mm glass ball lens mounted on an S4 iPhone with the support of 1/16 thick foam tape. The sensitivity, specificity, PPV and NPV were reported as shown in Table 2. The authors found this device not ready for use in clinical or epidemiological settings due to its low sensitivity.

The second validation study, carried out by Ephraim et al., was embedded within an ongoing epidemiology survey of schistosomiasis and soil-transmitted helminthiasis in the Central Region of Ghana. A total of 49 participants were randomly selected to provide urine samples for examination of S. haematobium eggs using 2 devices, a smartphone-mounted Foldscope and a reversed-lens CellScope, with conventional microscopy as the reference (Ephraim et al., Reference Ephraim, Duah, Cybulski, Prakash, D'ambrosio, Fletcher, Keiser, Andrews and Bogoch2015). Both devices were iPhone S5-based. However, the Foldscope consisted of an LED light source attached to the phone lens with the help of a paper-based attachment, a magnet and tape, whereas the CellScope consisted of a 3D-printed plastic embodiment with an integrated lens, fixed to the iPhone's lens. The urine samples (10 mL) were processed by centrifugation and examined by both devices and conventional light microscopy. The sensitivity and specificity of the smartphone-mounted Foldscope and the reversed-lens CellScope are shown in Table 2. The authors concluded that these devices could be of use in epidemiological and clinical settings upon further technical adjustments. The design prototype reported in this study did not have a sample stage without which its application in epidemiological surveys for mass disease screening is not feasible. Coulibaly et al. further evaluated the performance of the CellScope in a cross-sectional exploratory study for the diagnosis of schistosomiasis, with conventional light microscopy as a reference (Coulibaly et al., Reference Coulibaly, Ouattara, D'ambrosio, Fletcher, Keiser, Utzinger, N'goran, Andrews and Bogoch2016b). This study examined filtered urine slides and single KK thick smears from 226 participants with sensitivity and specificity reported in Table 2. The sensitivity of the CellScope was lower in low infection intensity samples (<40 eggs 10 mL⁻¹) and strongly correlated with light microscopy in terms of S. haematobium egg count estimated with a Pearson's correlation coefficient of 0.92. Based on the outcome, the authors concluded that this device may be of use in areas with moderate and high intensity of infection. However, the lack of a sample stage makes diagnosis with this prototype impractical.

The fourth validation study also by Bogoch et al. explored an ongoing epidemiological survey, of which 60 urine specimens were randomly selected and examined for S. haematobium eggs using a smartphone-based microscope with conventional microscopy-based centrifugation as the reference standard (Bogoch et al., Reference Bogoch, Koydemir, Tseng, Ephraim, Duah, Tee, Andrews and Ozcan2017). The device was a custom-made 3D-printed optomechanical attachment with an inbuilt lens mounted in alignment with the lens of a Nokia Lumia 1020 smartphone. It also contained 2 LED light sources, 2 polymer diffusers and an x–y stage for manual adjustment of sample slides. The sensitivity, specificity, PPV and NPV for the detection of S. haematobium are shown in Table 2. The sensitivity of the device when separately analysed for low- and high-intensity infections was 65.7% (95% CI 47.7–80.3) and 100.0% (95% CI 59.8–100.0), respectively. Based on these findings, the authors suggested this prototype for use in high infection intensity settings. Similarly, Koydemir et al. evaluated the smartphone-based microscope prototype previously validated by Bogoch et al. (Reference Bogoch, Koydemir, Tseng, Ephraim, Duah, Tee, Andrews and Ozcan2017) in an ongoing epidemiology study in a different setting. Urine and KK thick smear slides were imaged for manual identification of S. mansoni and S. haematobium eggs and detection was at an image level (Koydemir et al., Reference Koydemir, Coulibaly, Tseng, Bogoch and Ozcan2019). The number of urine and stool samples used in this study was not stated. The authors concluded that, this device may be a valuable diagnostic tool in endemic settings only after carrying out rigorous field validation. However, a pilot study to show promising diagnostic performance of this prototype at patient level would concretize this conclusion.

The sixth study by Armstrong et al. evaluated the SchistoScope; a modified version of the LoaScope [developed previously to detect Loa loa microfilaria (Kamgno et al., Reference Kamgno, Pion, Chesnais, Bakalar, D'ambrosio, Mackenzie, Nana-Djeunga, Gounoue-Kamkumo, Njitchouang, Nwane, Tchatchueng-Mbouga, Wanji, Stolk, Fletcher, Klion, Nutman and Boussinesq2017)] as a point-of-care device to automate quantitative detection of S. haematobium eggs with microscopy as a reference (Armstrong et al., Reference Armstrong, Harris, D'ambrosio, Coulibaly, Essien-Baidoo, Ephraim, Andrews, Bogoch and Fletcher2022). The prototype consists of an iPhone 8 with a reverse lens technology (previously described in this section) and a specialized cartridge to concentrate S. haematobium eggs in a single FOV. A total of 205 urine samples were collected to determine the sensitivity and specificity of the SchistoScope with AI and manual detection (Table 3). The authors concluded that this tool has a potential to provide quantitative diagnosis of S. haematobium. However, the cartridge used in this study for sample preparation resulted in approximately 79% of eggs lost during the proposed sample preparation method and therefore requires improvements.

Non-smartphone-based optical devices

Of the 8 non-smartphone-based optical devices reviewed here, six were limited to proof-of-concept (Linder et al., Reference Linder, Grote, Varjo, Linder, Lebbad, Lundin, Diwan, Hannuksela and Lundin2013; Holmstrom et al., Reference Holmstrom, Linder, Ngasala, Martensson, Linder, Lundin, Moilanen, Suutala, Diwan and Lundin2017; Sukas et al., Reference Sukas, Van Dorst, Kryj, Lagatie, De Malsche and Stuyver2019; Agbana et al., Reference Agbana, Nijman, Hoeber, Van Grootheest, Van Diepen, Van Lieshout, Diehl, Verhaegen, Vdovine and Coté2020; Dacal et al., Reference Dacal, Bermejo-Peláez, Lin, Álamo, Cuadrado, Martínez, Mousa, Postigo, Soto, Sukosd, Vladimirov, Mwandawiro, Gichuki, Williams, Muñoz, Kepha and Luengo-Oroz2021; Oyibo et al., Reference Oyibo, Jujjavarapu, Meulah, Agbana, Braakman, Van Diepen, Bengtson, Van Lieshout, Oyibo, Vdovine and Diehl2022) as reported by the authors. The first study by Linder et al. demonstrated the use of a mini-microscope (On-Chip) to image and detect the eggs of S. haematobium with AI (Linder et al., Reference Linder, Grote, Varjo, Linder, Lebbad, Lundin, Diwan, Hannuksela and Lundin2013). In this study, pooled urine samples from a pre-existing quality and assurance survey were used. The mini-microscope consisted of a webcam image sensor chip and an LED as a light source, all connected to computer software for image capture. Two AI algorithms were deployed for the detection of S. haematobium eggs in urine-captured images (see Table 3). Stool samples were also imaged, and eggs of T. trichiura and S. mansoni were morphologically distinguished by a technician. Based on the outcome, the authors concluded that this setup could be exploited for building simpler imaging low-cost devices for diagnosis of schistosomiasis and STHs. However, in this study, automated detection was applied for urine parasite detection only, and not for stool parasites.

The second study by Holstrom et al. reported the imaging performance of a digital microscope and a deep learning-based image analysis algorithm, with human experts as a comparator, for the identification of helminth eggs in urine and stool samples (Holmstrom et al., Reference Holmstrom, Linder, Ngasala, Martensson, Linder, Lundin, Moilanen, Suutala, Diwan and Lundin2017). The design of the digital microscope included a white LED with a bandpass filter and an ultraviolet LED to support both bright field and fluorescent imaging. This digital microscope was constructed from mobile phone-based camera components, such as an external motor unit attached for precise manual adjustment of the glass slide, and it was connected to a laptop with the AI algorithm for image processing. Sixteen positive stool slides and 5 positive urine slides were prepared, and 7385 images were captured with the portable microscope. Of this total number of images captured, 410 were labelled as positive for helminth eggs by human experts and were used for subsequent training and evaluation of the AI algorithm. For detection on an image level for eggs of A. lumbricoides, T. trichiura and hookworm by the algorithm, the sensitivity and PPV were computed (see Table 3). Based on the authors, the device produced sufficient image quality for manual and automatic detection of STH and S. haematobium eggs using a deep learning algorithm. Further validation of this setup at patient level is required before implementation in field settings can be considered.

The third study developed and tested a lab-on-a-disk (LOD) digital platform, to image and manually count STHs and S. mansoni eggs in a single FOV with Mini-FLOTAC and McMaster as reference tests (Sukas et al., Reference Sukas, Van Dorst, Kryj, Lagatie, De Malsche and Stuyver2019). Stool samples were processed by floatation and eggs collected in a single FOV. The complete design included a setup for imaging, a minicentrifuge and a sample disk where the sample is loaded. The imaging setup consisted of a macro lens, an adapter, a 10 ×  objective lens and a halogen light source. These parts were mounted in that order to a Sony α5100 camera and the imaging zone of the sample disk, in a vertical axis. The Sony camera was connected to a Samsung tablet for the camera control and image capture. Seven positive stool samples were examined using the 3 methods. The Mini-FLOTAC counts correlated with the LOD counts from the whole chamber of the disk (correlation coefficient 0.91). In addition, samples with low infection intensity of STHs (~30–100 egg per gram) determined with Mini-FLOTAC and McMaster were also detected as egg positive by the LOD device but with a lower egg count. According to the authors, this platform produced quality images to properly detect and quantify eggs of parasites in stool. Nonetheless, the sample disk needs further improvement since eggs were lost along the channel path leading to the collection zone where quantification in a single FOV was aimed.

The fourth study by Agbana et al. reported the use of a novel lens-free imaging system in combination with flow cytometry and AI (SODOS) to image and detect S. haematobium eggs in comparison with conventional microscopy (Agbana et al., Reference Agbana, Nijman, Hoeber, Van Grootheest, Van Diepen, Van Lieshout, Diehl, Verhaegen, Vdovine and Coté2020). The SODOS device consists of an imaging sensor, a micro-channel flow cell and a monochromatic light source aligned on the same optical axis. It also contains an automatic urine injection system with a calibrated syringe for pumping the sample into the flow cell and a cup for collecting the outlet of the flow cell. To evaluate the device performance, urine samples collected from uninfected volunteers were spiked with S. haematobium eggs (17 eggs mL⁻¹). Per run, 10.7 mL urine was imaged by the device, followed by image analysis using the AI software programme. An accuracy of 96.8% and a sensitivity and specificity for the AI software programme were determined (see Table 3). According to the authors, the SODOS may serve as a future tool to complement existing techniques for the diagnosis of urinary schistosomiasis if the sensitivity is further improved. Furthermore, this study was limited to spiked urine samples and validation using clinical samples is required.

The study by Dacal et al. reported the use of a mobile microscope assisted by AI for the detection of T. trichiura and A. lumbricoides in stool samples using microscopy as a reference (Dacal et al., Reference Dacal, Bermejo-Peláez, Lin, Álamo, Cuadrado, Martínez, Mousa, Postigo, Soto, Sukosd, Vladimirov, Mwandawiro, Gichuki, Williams, Muñoz, Kepha and Luengo-Oroz2021). The design of the digital mobile microscope consisted of a smartphone coupled to a conventional microscope (Leica DM-200) with the help of an adapter such that the smartphone's camera aligns with the optics of the microscope. For this study, 51 KK slides were prepared from 12 samples, imaged and uploaded into a telemedicine platform for remote image annotation, training and validation of AI algorithm. The sensitivity and specificity of their setup at image level were computed (see Table 3). According to the authors, the proposed design could serve a great deal to reduce time, distance and expertise needed for coproscopic analysis of STHs detection thereby making accurate diagnosis feasible. Nevertheless, this design still requires an expert to manually operate the attached microscope for sample imaging thereby sharing similar challenges as conventional microscopy.

The sixth proof-of-concept study by Oyibo et al. reported the SchistoScope; an automated microscope builds from off-the-shelf components for use in resource-limited settings for the detection of S. haematobium eggs with AI (Oyibo et al., Reference Oyibo, Jujjavarapu, Meulah, Agbana, Braakman, Van Diepen, Bengtson, Van Lieshout, Oyibo, Vdovine and Diehl2022). The prototype consists of an optical system with an automated 3-axis scanner (XYZ) to automate sample scanning process and an onboard computer connected to a Raspberry Pi HQ camera. Both urine samples from uninfected persons spiked with eggs from gut tissue of hamsters, clinical urine and stool samples were used to test the prototype. This tool captured quality images with distinguishable features of S. haematobium, S. mansoni and hookworm eggs. A dice similarity coefficient (DSQ) of 0.44 was obtained when evaluating the egg count outcome from the automated SchistoScope compared to manual counts. Based on the authors, the SchistoScope could be a suitable tool for use in the monitoring and evaluation of schistosomiasis control programmes in endemic settings. However, we think that the SchistoScope's performance in automated egg quantification based on the reported DSQ limits its potential application in monitoring and evaluation of schistosomiasis control programmes.

In 2 non-smartphone-based validation studies, a commercially available research-grade handheld mobile light microscope, Newton Nm1, was evaluated for the diagnosis of STH infections with conventional light microscopy as the reference standard (Bogoch et al., Reference Bogoch, Andrews, Speich, Ame, Ali, Stothard, Utzinger and Keiser2014a). In this study, 91 KK slides were randomly selected from a pack already analysed with conventional microscopy. These slides were re-examined using the Newton Nm1 microscope of which, 39.6 and 70.3% were positive for A. lumbricoides and T. trichiura, respectively, whereas 38.5% and 76.9% were positive for A. lumbricoides and T. trichiura, respectively, using conventional microscopy. No other known and highly sensitive reference test was used to determine whether samples were truly or falsely positives. Based on these findings, the authors concluded that the Newton Nm1 microscope may serve as a useful tool to detect and quantify STH eggs in endemic settings. Interestingly, Bogoch et al. (Reference Bogoch, Coulibaly, Andrews, Speich, Keiser, Stothard, N'goran and Utzinger2014b), and Coulibaly et al. (Reference Coulibaly, Ouattara, D'ambrosio, Fletcher, Keiser, Utzinger, N'goran, Andrews and Bogoch2016b), again evaluated the Newton Nm1 microscope for the detection of helminth eggs with conventional light microscopy (single KK and urine filtration) as a reference standard. From 164 prepared stool slides and 180 urine slides, Bogoch et al. determined the sensitivity and specificity of the device for the detection of Schistosoma and T. trichiura (Bogoch et al., Reference Bogoch, Coulibaly, Andrews, Speich, Keiser, Stothard, N'goran and Utzinger2014b). A sufficient diagnostic yield was obtained (see Table 3). Furthermore, Coulibaly et al. conducted a cross-sectional study and used urine and KK slides prepared from urine and stool samples collected from 226 individuals (Coulibaly et al., Reference Coulibaly, Ouattara, D'ambrosio, Fletcher, Keiser, Utzinger, N'goran, Andrews and Bogoch2016b). The Newton Nm1 handheld microscope produced a good sensitivity and an excellent specificity (see Table 3). The authors of these studies concluded that the Newton Nm1 microscope may serve as a valuable tool in clinical and public health settings to detect these infections. Nonetheless, both studies used conventional microscopy as the reference with limited sensitivity. Therefore, thorough evaluation with more sensitive diagnostics is needed to validate this tool. Furthermore, the Newton Nm1 microscope for the detection of these parasites is not fully automated.

The last study evaluated, for the first time, a new and remote diagnostic technique used in veterinary medicine called FECPAK compared to a single, double and quadruplicate KK method for analysing STHs eggs in human stool (Moser et al., Reference Moser, Barenbold, Mirams, Cools, Vlaminck, Ali, Ame, Hattendorf, Vounatsou, Levecke and Keiser2018). The FECPAK system is made up of a filtration unit, a micro-I cassette with sample loading wells that fit into the micro-I cassette, all connected to a computer for imaging of samples in a single FOV. For this study, 590 positive and 25 negative samples for any of the STHs were collected and analysed by the 4 methods. Images captured with the FECPAK were analysed by a microscopist. The sensitivity and specificity of the FECPAK device were determined for the detection of A. lumbricoides, T. trichiura and hookworm (Table 3). The sensitivity of FECPAK was found to be lower than single, double and quadruple KK method for the detection of A. lumbricoides and hookworm (see Table 3). The FECPAK underestimated the prevalence of these helminth infections compared to the single, double and quadruplicate KK method. Although FECPAK was originally developed for animal helminths diagnosis, authors concluded that it might be an interesting tool for both epidemiological and clinical human studies. Further evaluation studies of FECPAK for diagnosis of human STHs would help to further develop the tool.