11th European Congress on Telepathology and 5th International Congress on Virtual Microscopy Quantitative phase imaging and Raman micro-spectroscopy applied to Malaria Jacques Klossa1&, Benoit Wattelier2, Teddy Happillon3, Dominique Toubas3,4, Lucie de Laulanie2, Valerie Untereiner3, Pierre Bon2, Michel Manfait1 1TRIBVN, 39, rue Louveau, 92320 Châtillon, France 2Phasics, Campus de l'Ecole Polytechnique, 91128 Palaiseau, France 3MEDyC FRE/CNRS 3481, 51096 Reims, France 4CHRU de Reims, Laboratoire de parasitologie-mycologie, 51100 Reims, France &Corresponding author: [email protected] SFR CAP-Santé Structure Champagne - Ardenne Fédérative de Recherche Picardie INTRODUCTION TO AUTOMATED MALARIA DIAGNOSIS Malaria is due to parasitism of red blood cells (RBC) by protozoan parasites of the genus Plasmodium. Three main parameters have to be determined for patient treatment: parasite species, the rate of infected blood cells (parasitemia), and development stage. 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 This study, using the IHMO prototype, presents the first stage of proof of concept for achieving the 3 main diagnostic tasks: i) Parasite detection, ii) Parasite classification for appropriate treatment, iii) Parasitemia evaluation including life cycle stage identification and after treatment monitoring. 50 45 40 35 30 25 20 15 10 5 12000 10000 Heterogeneity Currently, transmitted light microscopy is the gold standard for malaria diagnosis with immuno-chromatography or molecular biology when needed. Microscopy observation needs a specialist and is time consuming (e.g. observation of hundreds fields of view at 100x immersion objective) and automation with slide scanner and image analysis is not straightforward. Therefore we think that for an automated solution, parasite localization could be easily achieved with quantitative phase imaging (QPI). In addition, Raman micro-spectroscopy (RMS) could provide a full molecular signature for parasite species classification. 8000 6000 4000 2000 1000 1500 2000 2500 3000 -1 Wavenumbers (cm ) 0 LABEL FREE TECHNIQUES AND THE MALARIA USE CASE The IHMO multimodal machine (see illustration and window framework) has been used for parasite localization and Raman spectra acquisition. It combines in a single multimodal scanner, i) a transmitted light robotized microscopy platform with two objectives 40x and 150x, ii) a quadriwave quantitative phase imager camera and iii) a Raman micro-spectrometer: laser excitation source 532nm, laser spot size of 1µm on the sample. Unstained blood smear on a slide. After observation (QPI and Raman), the smear is stained to correlate the observations with an optical inspection by a clinician. QUANTITATIVE PHASE IMAGING FOR PARASITE LOCALIZATION AND PARASITEMIA COUNTING What is Quantitative Phase ? e 1 2 3 4 How is Quantitative Phase measured with Quadri-Wave Interferometry (SID4BIO®)? Index of refraction: Vlight=c/n n2 RAMAN MICRO-SPECTROSCOPY FOR PARASITE IDENTIFICATION CCD chip n1 SID4BIO camera Raw spectra (1) need a pre-treatment phase to make them eligible for classification dt=(n2-n1)xe/c Baseline correction: the presence of hemoglobin into the cells implies distortions of the baseline of each spectrum. A function based on a polynomial estimation and correction of the baseline is used for each spectrum (2) Diffractive optics Quantitative Phase=(n2-n1)xe What is seen in Quantitative Phase images? • Morphology changes (thickness) • Protein concentration changes • Tissue physico-chemical property changes Normalization: the normalization function, standard normal variate, is used to eliminate the variation of the absolute values into the spectra, making them comparable avoiding scale differences (3) Bon, Maucort, Wattellier & Monneret ,Opt. Express, 2009, 17, 13080-13094 Quantitative phase image acquisition Phase Profiles Healthy RBC The SID4BIO camera imager is connected to a camera port. Each field of view of the 1” camera at 40x contains about 600 RBC on our specimen. Phase Images Hemoglobin subtraction: a representative spectrum of pure hemoglobin has been estimated and subtracted from each spectrum, using a mean squared based function (4) Classification: The classification of the pretreated spectra is then done with the Hierarchical Clustering Analysis algorithm, based on the Euclidean distances between each spectrum Erythrocyte Plasmodium Cluster 1 Cluster 2 100% 0% (35/35) (0/35) 0% 100% (0/34) (34/34) Erythrocytes Results Parasites The Plasmodium parasite induces a negative phase-shift (white shape inside the RBC on the QP images). Other artefacts inside the RBC appear in black. A high-pass filter is applied to the images to enhance the parasite. A three-level segmentation isolates the medium, the RBC and the parasites. The number of detected parasites versus total RBC number provides the parasitemia count. x-y position of each infected RBC is recorded for further Raman micro-spectroscopy. Infected RBC Phase image analysis Dendrogram of infected and sane red blood cells classification WORKFLOW COMPARISON TASKS MANUAL DIAGNOSIS microscope and complementary techniques STAINED SMEAR AUTOMATED SCANNING still many techniques and quite tricky 100x scan LABEL FREE AUTOMATED SCANNING one single specimen preparation, easy 40x scan Careful microscope reviewing at 100x on thick/thin peripheral blood smear Automated image acquisition: quite tricky at 100x: could need slow Quantitative phase imaging automated scan at 40x and parasite z stack acquisition detection through phase image analysis PARASITE CLASSIFICATION Microscope screening and complementary techniques: immuno-chromatography and molecular biology Automated morphological image pre-classification would need manual confirmation on virtual slide and microscope review of poorly imaged cells Raman micro-spectroscopy spectra provides highly sensible and specific molecular signature and classification can be achieved using quite simple techniques PARASITEMIA COUNTING 2 x 20’ reviewing on thin films and complementary techniques Automated high power image acquisition can lead to miss some parasites. Still needs complementary techniques One single specimen preparation and fully automated data acquisition and classification PARASITE DETECTION CONCLUSION AND ROADMAP This study proved the ability of QPI to detect RBC and infected RBC at low magnification without any previous staining. Each detected infected RBC can then be easily confirmed with Raman micro-spectroscopy spectra acquisition. The results are promising and suggest that a standardized integrated multimodal microscopy solution can easily be automated. The proposed concept appears to be powerful, however it needs further studies which are on our roadmap, I) to validate QPI early stage parasite detection in comparison to current manual microscopy and II) to assess Raman micro-spectroscopy classification power for early stage parasite characterization and for species classification.
© Copyright 2018 ExploreDoc