1 / 38

Results and challenges running cloud based

diagnostics

Dr.sc.ing. Dmitrijs Bļizņuks

Riga Technical University

Non-contact Skin Cancer

Diagnostic System

DOC Riga 2019

2 / 38

Outlines

• Skin cancer non-contact diagnostics

• Achievements

• Modular architecture: Pros & Cons

• Diagnostic algorithms

3 / 38

Why skin cancer?

4 / 38

Age-adjusted death rates for the 10 leading causes of death in United States

SOURCE: NCHS, National Vital Statistics System, Mortality.

5 / 38

6 / 38

7 / 38https://www.cancerresearchuk.org/health-professional/cancer-statistics/statistics-by-cancer-type/melanoma-skin-cancer/survival

95%

5%Survival rates

8 / 38

Wavelengths’ skin penetration depth

9 / 38

Place your screenshot here

Cloud service

Handheld device

10 / 38

• More than 1500 lesions tested

totally

• More than 1.5 years of testing

in Latvian Oncology Clinics

• Tested by General Practice

doctor

• Found 2 melanomas during two

months

• Ongoing tests in Hungary and

Bulgary

11 / 38

10+ years of research in biophotonics

5 Dr.phys.

University of Latvia

Experienced doctors of the Oncology Center of Latvia

Dermato-oncologists

Hardware and software design3 Dr.Sc.ing.

Riga Technical University

12 / 38

✓ Non-invasive✓ Portable✓ Quantitative✓ Reliable (Sp=98%)✓ Uses various spectra✓ Image analysis within seconds✓ Wireless✓ Image evaluation in time (Archive)✓ Affordable

Main users

General Practice doctors

13 / 3813

14 / 3814

First page from

doctor’s view

15 / 3815

Acquired images

viewing and assigning

to the patient

16 / 3816

Patient control

17 / 3817

Patient’s skin

measurements

18 / 3818

Servers monitoring

and control

19 / 3819

Server’s load

monitoring

20 / 38

21 / 38

Cloud based diagnostics (SaaS)

Could be used as a standalone service for various image analysis

Calculation node 1

Checkyourskin.eu

Device 1Doctor 1Doctor 1

Patient 1Patient 1

Internet

Calculation

process 1-1

Calculation

process 1-m

Calculation node n

Calculation

process n-1

Calculation

process n-m

...

Patient nPatient n

Device nDoctor nDoctor n

Patient 1Patient 1

Patient nPatient n

22 / 38

23 / 38

24 / 38

Optical Density spectra of skin lesions

D.Jakovels et al. Application of principal component analysis to multispectral imaging data for evaluation of pigmented skin

lesions, Proc. SPIE 9032, Biophotonics—Riga 2013, 903204 (November 18, 2013)

25 / 38

Specific stair-like inner structure to keep light polarization

26 / 38

4 layer PCBs

5 separately controlled LED groups

4 LEDs in each group

up to 500mA to each diode

27 / 38

Raspberry Pi

Compute 3

module

USB WiFi

module

USB “IDS”

camera

USB 3G/4G

module

Central control module

Illumination

control (STM32)

LED current

drivers

Five groups of

4x LEDs each

USB

USB USB

SPI

LED supply voltage +0.3V

Current and voltage

control for each LED

Photodiode

Trigger

Mea

sure

d in

ten

sity

LED control board

LED ring

LED supply

voltage step-down

Analog

voltage

setpoint

Li-ion

battery

Power and

charging

Temperature

sensor

User buttonsDigital inputs

Analog

voltage

setpoint

Digital-to-

analog

convertor

SPI

Resistance

Modular architecture

• LED illumination,

• LED control,

• Camera,

• Central module,

• Wireless connectivity.

3G → 4G

WiFi 2.4GHz → 5GHz

Camera 1 Mpix / 5MPix

28 / 38

l=535nm

l=660nm

l=950nm

Nevus

p’ karte

Melanoma

p’ karte

29 / 38

Searching for marker

30 / 38

Stabilizing by marker

31 / 38

Removing hair

32 / 38

Searching for healthy skin

33 / 38

34 / 38

35 / 38

MELANOMA VS SEB.KERATOSIS

Melanoma

Seborejas keratoze

36 / 38

0.0

50.0

100.0

150.0

200.0

250.0

-2.0 0.0 2.0 4.0Me

an

AF

in

ten

sity fro

m s

ele

cte

d le

sio

n a

rea

, a

.u.

Mean p'max from selected lesion area (estimated from 100 maximal values of selected area), a.u.

Seborrheic Keratosis(n=13)

Hyperkeratosis (n=9)

Melanocytic Nevus(n=23)

Melanoma (n=3)

I. Lihacova, K. Bolochko, E. V. Plorina, M. Lange, A. Lihachev, D. Bliznuks, and A. Derjabo, "A method for skin malformation classification by combining

multispectral and skin autofluorescence imaging," SPIE Proc 10685, 1068535 (2018).

37 / 38

38 / 38

Acknowledgements

This work has been supported by European Regional Development Fund

project ‘Portable Device for Non-contact Early Diagnostics of Skin Cancer’

under grant agreement # 1.1.1.1/16/A/197

Dmitrijs Bļizņuks

Dmitrijs.Bliznuks@rtu.lv