1 of 49

2 of 49

FotoFinder �Artificial Intelligence

A new era in dermatology

3 of 49

AIMEE

Artificial Intelligence Mole Examination and Evaluation

AI as a Class II Medical Device according to EU MDR

4 of 49

FotoFinder Systems GmbH

4

21.09.2023

  • Improved workflow in dermoscopy
  • AIMEE AI Assistant speeds up lesion assessment, supports decision making and enhances patient experience
  • AIMEE quickly provides an AI Score - directly visualized in the dermoscopic image.

New: FotoFinder AIMEE (Artificial Intelligence Mole Examination & Evaluation)

5 of 49

FotoFinder Systems GmbH

5

21.09.2023

AIMEE Scoring

6 of 49

What is AI good for? �What is the added value?How does it change �the clinical practice in 5 or 10 years?

FotoFinder Systems GmbH

6

21.09.2023

7 of 49

8 of 49

AI for the pre-assessment of skin lesions

The Basis: “Deep Learning“

FotoFinder Systems GmbH

8

21.09.2023

9 of 49

FotoFinder Systems GmbH

9

21.09.2023

From early AI to Deep Learning

1950´s

1960´s

1970´s

1980´s

1990´s

2000´s

2010´s

2020´s

Early Artificial Intelligence

causes enthusiasm.

Machine Learning

picks up pace

Breakthroughs in Deep Learning

boost the AI boom.

Artificial Intelligence

Since an early wave of optimism in the fifties, smaller sub-areas of Artificial Intelligence have triggered increasing disruptions.

First Machine Learning, later Deep Learning, a sub-area of Machine Learning.

10 of 49

FotoFinder Systems GmbH

10

21.09.2023

Analysis based on

pattern recognition algorithms

Image taking

Score

Input

Feature extraction

Classification

Malignant

Benign

Machine Learning

Output

Old standard: Machine Learning

11 of 49

FotoFinder Systems GmbH

11

21.09.2023

Melanoma

Pixel-based analysis through convolutional neural network

„Memory“ melanoma

Nevus

Pixel-based analysis through convolutional neural network

„Memory“ naevus

New standard: Deep Learning

12 of 49

FotoFinder Systems GmbH

12

21.09.2023

FotoFinder Classifier

Image taking

AI Score AIMEE�Moleanalyzer pro

Input

Feature extraction + Classification

Malignant

Benign

Deep Learning

Output

New standard: Deep Learning

13 of 49

FotoFinder Systems GmbH

13

21.09.2023

AUC 0.945

AUC 0.738

A prospective test data set with 1.981 dermoscopic images shows the superiority of Moleanalyzer pro (black) compared to its predecessor Tuebinger Moleanalyzer 3 (grey), given their AUC (area under the curve) values.

Moleanalyzer pro versus previous algorithm

14 of 49

What makes FotoFinder AI unique for the pre-assessment of skin lesions ?

FotoFinder Systems GmbH

14

21.09.2023

15 of 49

FotoFinder Systems GmbH

15

21.09.2023

350.000 +

A huge base of training images from global AI collaborators

FotoFinder AI benefits from a continuously increasing number of histologically confirmed images of various ethnical skin types.

16 of 49

FotoFinder Systems GmbH

16

21.09.2023

  • International AI development team
  • Parallel development of binary and multi-class classifier
  • Ensemble learning
  • Superior accuracy

FotoFinder Classifier

17 of 49

FotoFinder Systems GmbH

17

21.09.2023

2017

AI on the same level as dermatologists

2018

2023

Study „Man against machine reloaded“: FotoFinder AI exceeds dermatologists

Study „Man against machine“:

FotoFinder AI exceeds dermatologists

AI in the course of time

2019

Study „Human with machine“:

Dermatologists can improve their performance using AI

18 of 49

Superior accuracy – �verified by clinical studies

  • FotoFinder Moleanalyzer pro: the first algorithm confirmed by a representative study
  • 2018: „Man against machine“, University of Heidelberg, Holger Haenssle, Christine Fink et al., Annals of Oncology
  • 2019: „Man against machine reloaded“, University of Heidelberg, Holger Haenssle, Christine Fink et al., Annals of Oncology
  • Result: FotoFinder AI exceeds expert average
  • Further studies on FotoFinder AI are continuously published, most recently "Human with machine", Winkler et al., 2023
  • Study Overview: https://www.fotofinder.de/en/technology/artificial-intelligence/scientific-studies-on-fotofinder-ai

21.09.2023

FotoFinder Systems GmbH

18

VS

19 of 49

FotoFinder AI - Studies

20 of 49

FotoFinder Systems GmbH

20

21.09.2023

Haenssle et al. (2019); Annals of Oncology

Melanoma

PECA

DF

BCC

Man against machine reloaded: Broad Spectrum

21 of 49

FotoFinder Systems GmbH

21

21.09.2023

Sensitivity 95.0%

Specificity 76.7%

Haenssle et al. (2019); Annals of Oncology

Broad Spectrum

  • Broad Spectrum: Wide range of benign and malignant skin lesions tested in the study Haenssle et al. (2019)
  • Validation of FotoFinder AI by the means of three independent and comprehensive test data sets
  • FotoFinder AI The study confirms high specificity and sensitivity values

22 of 49

FotoFinder Systems GmbH

22

21.09.2023

Sensitivity 95.0%

Specificity 76.7%

According to the study, FotoFinder AI achieves a better result than the average value generated from the test results of the dermatologists.

Broad Spectrum

FotoFinder AI curve

Individual results of dermatologists

Median result of dermatologists

Operation point FotoFinder AI

Haenssle et al. (2019); Annals of Oncology

23 of 49

FotoFinder Systems GmbH

23

21.09.2023

Broad Spectrum

Based on specificity and sensitivity of FotoFinder AI, the AUC (area under the curve) value is high (91.8%).

Haenssle et al. (2019); Annals of Oncology

Area under the curve 91.8%

24 of 49

25 of 49

26 of 49

27 of 49

FotoFinder Systems GmbH

27

21.09.2023

AI

Responsible

Use of AI�

Early diagnosis requires

follow-up!

28 of 49

FotoFinder Systems GmbH

28

21.09.2023

AI Score + Expert opinion

H. Peter Soyer, MD, FACD

Rainer Hofmann-Wellenhof, MD

Andreas Blum, MD

International

FotoFinder expert team

29 of 49

Intelligent tools implemented

In Bodymapping

To detect new & modified lesions

30 of 49

THE MOST EFFICIENT WORK FLOW �THE 2 STEP METHOD OF DERMOSCOPY

STEP 1: CLINICAL DIFFERENTIAL DIAGNOSIS

STEP 2: DERMOSCOPY DIFFERENTIAL DIAGNOSIS

31 of 49

FotoFinder Systems GmbH

31

21.09.2023

ATBM master room requirements

No direct sunlight!

32 of 49

FotoFinder Systems GmbH

32

21.09.2023

The benchmark for �video dermoscopy worldwide

33 of 49

34 of 49

Baseline

6 months

Female

64 y.o

History of melanoma 0.2 mm (3 years ago)

History of organ transplantation

Melanoma in situ

35 of 49

36 of 49

37 of 49

38 of 49

Do we contribute to life saving of melanoma patients?

Does Dermoscopy & Early Diagnosis have a general positive impact?

Are skin cancer mortality rates decreasing thanks to Dermoscopy?

39 of 49

Australian Cancer Atlas

Melanoma incidence

40 of 49

41 of 49

This correlation, unique to melanoma, has prompted speculation, including by Queensland epidemiologist, David Whiteman, that “‘overdiagnosis’ is [currently] a price paid to improve survival.

Rosendahl,C , Clark, S. (2023). General practice and melanoma management in Australia: controversies and implications for generalist GP training. Medical Journal of Australia, 218(9), 397–398. https://doi.org/10.5694/mja2.51928

42 of 49

2021

2022

54-year-old male

43 of 49

LM Melanoma in situ associated with a compound naevus

44 of 49

2018

2023

45 of 49

LM Melanoma in situ (no associated naevus)

46 of 49

TBP software detects new lesion

Male age 41, > 100 naevi, father has had multiple melanomas, mother is Chinese

Baseline 2017

Sep 2022

47 of 49

Nodular subtype melanoma Breslow 1.9mm (no associated naevus)

48 of 49

FotoFinder Systems GmbH

48

21.09.2023

AI

Responsible

Use of AI�

Early diagnosis requires

follow-up!

49 of 49

FotoFinder Systems GmbH

49

21.09.2023

Visualizes.