Mobile Imaging, Sensing and Diagnostics

Transcription

1 Mobile Imaging, Sensing and Diagnostics Aydogan Ozcan, Ph.D. Electrical Engineering Department & Bioengineering Department & California NanoSystems Institute University of California, Los Angeles (UCLA)

2 Aydogan Ozcan, Ph.D. UCLA

3 A different view of the Moore s Law

4 Smart-phones as super-computers

5 Cell phones are now everywhere: A great potential for Telemedicine needs Afghanistan 40 Canada 68 U.K. 97 Russia 162 Mexico 78 U.S. 97 Chad 24 India 45 China 56 Bolivia 72 Brazil 90 Argentina 129 Angola 44 South Africa 94 World Bank >7 billion cell phones are being used worldwide. ~15 billion cell phones have been sold so far. UAE 232 Indonesia 69 Australia 111 > 75% are in developing countries. (International Telecommunication Union)

6 Providing a new platform for telemedicine Lensfree Microscope Tomographic Microscope Cellphone Fluorescent Microscope Lensfree Super-resolution Microscope Cellphone Holographic Microscope Albumin Tester Imaging Flow-Cytometer Ozcan Research UCLA

7 Providing a new platform for telemedicine E. coli sensor Allergen detector Heavy metal detector Google Glass based Diagnostics Diagnostic test reader (HIV, malaria, etc.) Blood Analyzer Blood Analyzer

8 Providing a new platform for telemedicine E. coli sensor Allergen detector Heavy metal detector Chlorophyll Measurement Diagnostic test reader (HIV, malaria, etc.) Blood Analyzer Blood Analyzer

9 The µ-internet Big Data New opportunities in micro-analysis, medical diagnostics and epidemiology 9

10 Providing a new platform for telemedicine LUCAS: Lensfree (On-Chip Imaging) Compact, Cost-effective High-throughput Highly sensitive (Incoherent Holography) Tolerant to Misalignments LUCAS: Lensless, Ultra wide-field Cell monitoring Array platform based on Shadow imaging 10

11 Imaging of shadows Cell Shadows imaged using lensfree on-chip imaging LED CMOS/CCD sensor-array H<1-2 µm contact on-chip imaging H ~ 10-1,000 µm holographic on-chip imaging 11

12 (e) (a) (a) Lymphocyte (c) (b) (b) Lymphocyte Neutrophil (d) (e) (c) Neutrophil (d) Neutrophil Neutrophil

13 Cell Density: 102K [Cells/uL] Reconstruction 10X

14 Lensfree imaging of histopathology slides Breast Cancer Tissue 14

15 microcontroller 23 LEDs 23 fibers colour filter sample tray CMOS sensor

16 LUCAS 2.5 (NA=0.65) 5 Megapixel CMOS imager 10 µm 10 µm 10 µm radians radians radians Plasmodium falciparum

17 Lensfree imaging of histopathology slides Science Translational Medicine - AAAS (2014) 17

18 Science Translational Medicine - AAAS (2014)

19 (a) Lensfree FOV Lensfree reconstructed amplitude (colored) Pap Smear (b) 50um (c) 50um (d) (e) 40X FOV Microscope 40X 0.75NA 1mm 50um 50um Lensfree reconstructed amplitude Abnormal Pap Smear (f) (g) (h) (i) 50um 20um 20um 20um (j) (k) (l) (m) Microscope 40X 0.75NA 50um 20um 20um 20um 19

20

21 A different view of the Moore s Law

22

23 ACS Nano, 2013

24 Imaging and Sizing of Single DNA Molecules on a Mobile-Phone ACS Nano, 2014

25 Quantification of Giardia lamblia cysts using mobile-phone based fluorescent microscopy and machine learning Lab Chip, 2015

26 Lab Chip, 2015

27 ACS Nano 2015 Cellphone-Based Hand-Held Micro-Plate Reader for Point-of-Care ELISA Testing

28

29 ACS Nano 2015

30 ACS Nano 2015

31 Mumps Measles HSV 1 HSV 2 N Analysis Type Curve Fitting M.L Curve Fitting M.L Curve Fitting M.L Curve Fitting M.L Specificity 92.70% ± 0.49% 97.37% ± 2.63% 92.09% ± 0.77% 94.56% ± 3.43% % ± 0.00% % ± 0.00% % ± 0.00% % ± 0.00% Sensitivity % ± 0.00% % ± 0.00% 98.39% ± 0.42% 99.47% ± 0.59% 97.32% ± 0.32% % ± 0.00% % ± 0.00% % ± 0.00% Agreement 97.21% ± 0.41% 99.61% ± 0.39% 96.08% ± 0.32% 98.56% ± 0.53% 97.16% ± 0.55% 99.42% ± 0.43% 98.66% ± 0.22% 99.41% ± 0.44% ACS Nano 2015

32 Providing a new platform for telemedicine E. coli sensor Allergen detector Heavy metal detector Google Glass based Diagnostics Diagnostic test reader (HIV, malaria, etc.) Blood Analyzer Blood Analyzer

33 The µ-internet Big Data - An opportunity or a challenge? 33

34 Imaging technologies experience an exponential growth, but human experts and diagnosticians do NOT! Diagnosticians 34

35 Diagnosis of Malaria using Crowd-Sourced Games

36 Diagnosis of Malaria using Crowd-Sourced Games More than 80 countries are playing BioGames, with >3 Million cell diagnoses so far.

37 Each gamer is a Repeater Decoder: Maximum a Posteriori Probability (MAP) 37

38 Decoding of Diagnosis 38

39 Non-expert human crowd can collectively diagnose malaria infected cells Experiment Description of Test Images Gamers Positive RBCs Negative RBCs Control Images Accuracy SE SP PPV NPV test RBC images crowd-sourced to human gamers 5055 test RBC images presented to a boosted set of classifiers, trained on 1266 RBC images 459 low-confidence test images taken from the results of experiment 2 Hybrid diagnosis results using experiments 2 & test RBC images crowd-sourced to human gamers % 95.12% 99.41% 94.32% 99.50% NA (Training Images) 96.26% 69.64% 99.00% 87.70% 96.95% % 97.81% 91.89% 94.70% 96.59% % 89.38% 99.43% 94.18% 98.91% % 97.81% 99.05% 96.68% 99.38% 39

40 Non-expert human crowd can collectively diagnose malaria infected cells Term Acronym Definition TTTT + TTTT Accuracy ACC TTTT + TTTT + FFFF + FFFF TTTT Sensitivity or True Positive Rate SE or TPR TTTT + FFFF 40

41 Experts disagree with each other 41

42 Experts are not self-consistent 42

43 Digital Super Pathologist & Training of experts Digital Super Pathologist Training of Experts 43

44 Big Data Management: Crowd + Machine Crowd: Professionals + Non-experts (Smart and Cost-effective Telemedicine) 44

45 Distributed Medical Image Analysis and Diagnosis through Crowd-Sourced Games 45

46 Training on Malaria Diagnosis 46

47 Korea reached >1600 middle school students 47

48 Korea reached >1600 middle school students 48

49 Korea reached >1600 middle school students 49

50 Online educational content maintains student interest and trains students about global health topics 8 sessions: Video Lecture & Quiz followed by BioGames malaria training session 91% satisfied with the program 98% want to participate in future programs

51 Korea reached >1600 middle school students 51

52 Distributed Medical Image Analysis and Diagnosis through Crowd-Sourced Games

53 Ozcan Research Group Members & Funding 53

54 You can move from cell level diagnosis to slide level diagnosis Optimize diagnosis time based on the statistics of the individual diagnostician 54