Customized Deep Learning Classifier for Detection of Acute Lymphoblastic Leukemia Using Blood Smear Images.

Healthcare (Basel, Switzerland) 2022 AI 7 Explanations View Original
Original Paper (PDF)

Unable to display PDF. Download it here or view on PMC.

Plain-English Explanations
Pages 1-2
Why Early Detection of Acute Lymphoblastic Leukemia Matters

<

p

>

A

c

u

t

e

l

y

m

p

h

o

b

l

a

s

t

i

c

l

e

u

k

e

m

i

a

(

A

L

L

)

i

s

a

b

l

o

o

d

c

a

n

c

e

r

c

a

u

s

e

d

b

y

t

h

e

b

o

n

e

m

a

r

r

o

w

o

v

e

r

p

r

o

d

u

c

i

n

g

i

m

m

a

t

u

r

e

l

y

m

p

h

o

c

y

t

e

s

,

k

n

o

w

n

a

s

b

l

a

s

t

c

e

l

l

s

.

T

h

e

s

e

a

b

n

o

r

m

a

l

w

h

i

t

e

b

l

o

o

d

c

e

l

l

s

c

r

o

w

d

o

u

t

h

e

a

l

t

h

y

c

e

l

l

s

,

d

i

s

r

u

p

t

t

h

e

i

m

m

u

n

e

s

y

s

t

e

m

,

a

n

d

c

a

n

s

p

r

e

a

d

t

o

o

r

g

a

n

s

i

n

c

l

u

d

i

n

g

t

h

e

k

i

d

n

e

y

,

l

i

v

e

r

,

a

n

d

b

r

a

i

n

.

T

h

e

W

o

r

l

d

H

e

a

l

t

h

O

r

g

a

n

i

z

a

t

i

o

n

r

e

p

o

r

t

e

d

o

v

e

r

4

3

7

,

0

0

0

n

e

w

l

e

u

k

e

m

i

a

c

a

s

e

s

a

n

d

3

0

3

,

0

0

0

d

e

a

t

h

s

g

l

o

b

a

l

l

y

i

n

2

0

2

2

.

<

/

p

>

<

p

>

A

L

L

i

s

a

m

o

n

g

t

h

e

m

o

s

t

c

o

m

m

o

n

c

a

n

c

e

r

s

i

n

c

h

i

l

d

r

e

n

,

w

h

e

r

e

o

u

t

c

o

m

e

s

a

r

e

r

e

l

a

t

i

v

e

l

y

f

a

v

o

r

a

b

l

e

i

f

c

a

u

g

h

t

e

a

r

l

y

.

I

n

a

d

u

l

t

s

o

v

e

r

6

5

,

t

h

e

d

i

s

e

a

s

e

i

s

m

o

r

e

a

g

g

r

e

s

s

i

v

e

a

n

d

p

r

o

g

n

o

s

i

s

w

o

r

s

e

n

s

s

i

g

n

i

f

i

c

a

n

t

l

y

w

i

t

h

l

a

t

e

d

i

a

g

n

o

s

i

s

.

E

a

r

l

y

d

e

t

e

c

t

i

o

n

i

s

t

h

e

r

e

f

o

r

e

c

r

i

t

i

c

a

l

f

o

r

i

m

p

r

o

v

i

n

g

s

u

r

v

i

v

a

l

r

a

t

e

s

a

c

r

o

s

s

a

l

l

a

g

e

g

r

o

u

p

s

.

<

/

p

>

<

p

>

T

r

a

d

i

t

i

o

n

a

l

d

i

a

g

n

o

s

i

s

r

e

l

i

e

s

o

n

b

o

n

e

m

a

r

r

o

w

b

i

o

p

s

y

a

n

d

m

a

n

u

a

l

m

i

c

r

o

s

c

o

p

i

c

r

e

v

i

e

w

o

f

p

e

r

i

p

h

e

r

a

l

b

l

o

o

d

s

m

e

a

r

s

-

a

p

r

o

c

e

s

s

t

h

a

t

i

s

l

a

b

o

r

-

i

n

t

e

n

s

i

v

e

,

t

i

m

e

-

c

o

n

s

u

m

i

n

g

,

a

n

d

p

r

o

n

e

t

o

h

u

m

a

n

e

r

r

o

r

.

A

u

t

o

m

a

t

i

n

g

t

h

i

s

p

r

o

c

e

s

s

w

i

t

h

d

e

e

p

l

e

a

r

n

i

n

g

o

f

f

e

r

s

a

p

a

t

h

t

o

f

a

s

t

e

r

,

m

o

r

e

c

o

n

s

i

s

t

e

n

t

s

c

r

e

e

n

i

n

g

t

h

a

t

c

a

n

s

u

p

p

o

r

t

c

l

i

n

i

c

i

a

n

s

d

u

r

i

n

g

t

h

e

c

o

m

p

l

e

t

e

b

l

o

o

d

c

o

u

n

t

(

C

B

C

)

a

n

d

p

e

r

i

p

h

e

r

a

l

b

l

o

o

d

e

x

a

m

i

n

a

t

i

o

n

w

o

r

k

f

l

o

w

.

<

/

p

>

TL;DR: ALL is a serious blood cancer where early detection dramatically improves outcomes, but current diagnostic methods are slow and error-prone - creating a clear need for automated AI-based screening tools.
Pages 1-2
ALLNet: A Custom CNN Built Specifically for Leukemia Detection

<

p

>

T

h

i

s

s

t

u

d

y

p

r

e

s

e

n

t

s

A

L

L

N

e

t

,

a

c

u

s

t

o

m

c

o

n

v

o

l

u

t

i

o

n

a

l

n

e

u

r

a

l

n

e

t

w

o

r

k

(

C

N

N

)

d

e

s

i

g

n

e

d

f

r

o

m

t

h

e

g

r

o

u

n

d

u

p

t

o

c

l

a

s

s

i

f

y

b

l

o

o

d

s

m

e

a

r

i

m

a

g

e

s

a

s

e

i

t

h

e

r

l

e

u

k

e

m

i

c

b

l

a

s

t

c

e

l

l

s

o

r

h

e

a

l

t

h

y

w

h

i

t

e

b

l

o

o

d

c

e

l

l

s

.

R

a

t

h

e

r

t

h

a

n

a

d

a

p

t

i

n

g

a

g

e

n

e

r

a

l

-

p

u

r

p

o

s

e

a

r

c

h

i

t

e

c

t

u

r

e

,

t

h

e

t

e

a

m

t

a

i

l

o

r

e

d

e

v

e

r

y

l

a

y

e

r

o

f

t

h

e

n

e

t

w

o

r

k

s

p

e

c

i

f

i

c

a

l

l

y

t

o

t

h

e

v

i

s

u

a

l

c

h

a

r

a

c

t

e

r

i

s

t

i

c

s

o

f

p

e

r

i

p

h

e

r

a

l

b

l

o

o

d

s

m

e

a

r

m

i

c

r

o

s

c

o

p

y

i

m

a

g

e

s

.

<

/

p

>

<

p

>

T

h

e

m

o

d

e

l

w

a

s

t

r

a

i

n

e

d

a

n

d

e

v

a

l

u

a

t

e

d

o

n

t

h

e

C

_

N

M

C

_

2

0

1

9

d

a

t

a

s

e

t

-

t

h

e

p

u

b

l

i

c

l

y

a

v

a

i

l

a

b

l

e

A

L

L

C

h

a

l

l

e

n

g

e

D

a

t

a

s

e

t

f

r

o

m

I

S

B

I

2

0

1

9

-

w

h

i

c

h

c

o

n

t

a

i

n

s

e

x

p

e

r

t

-

a

n

n

o

t

a

t

e

d

i

m

a

g

e

s

f

r

o

m

7

3

p

a

t

i

e

n

t

s

.

T

h

i

s

d

a

t

a

s

e

t

w

a

s

c

h

o

s

e

n

b

e

c

a

u

s

e

c

l

a

s

s

i

f

i

c

a

t

i

o

n

l

a

b

e

l

s

w

e

r

e

v

e

r

i

f

i

e

d

b

y

e

x

p

e

r

t

o

n

c

o

l

o

g

i

s

t

s

,

m

a

k

i

n

g

i

t

a

r

e

l

i

a

b

l

e

b

e

n

c

h

m

a

r

k

f

o

r

d

e

v

e

l

o

p

i

n

g

d

i

a

g

n

o

s

t

i

c

t

o

o

l

s

.

<

/

p

>

<

p

>

T

h

e

c

o

r

e

h

y

p

o

t

h

e

s

i

s

i

s

t

h

a

t

d

e

e

p

l

e

a

r

n

i

n

g

c

a

n

e

l

i

m

i

n

a

t

e

t

h

e

n

e

e

d

f

o

r

h

a

n

d

c

r

a

f

t

e

d

f

e

a

t

u

r

e

e

n

g

i

n

e

e

r

i

n

g

,

a

l

l

o

w

i

n

g

t

h

e

n

e

t

w

o

r

k

t

o

a

u

t

o

m

a

t

i

c

a

l

l

y

l

e

a

r

n

t

h

e

d

i

s

t

i

n

g

u

i

s

h

i

n

g

v

i

s

u

a

l

p

a

t

t

e

r

n

s

b

e

t

w

e

e

n

b

l

a

s

t

c

e

l

l

s

a

n

d

h

e

a

l

t

h

y

l

y

m

p

h

o

c

y

t

e

s

d

i

r

e

c

t

l

y

f

r

o

m

p

i

x

e

l

d

a

t

a

.

T

h

i

s

e

n

d

-

t

o

-

e

n

d

a

p

p

r

o

a

c

h

r

e

d

u

c

e

s

d

e

p

e

n

d

e

n

c

e

o

n

d

o

m

a

i

n

-

s

p

e

c

i

f

i

c

f

e

a

t

u

r

e

s

e

l

e

c

t

i

o

n

w

h

i

l

e

i

m

p

r

o

v

i

n

g

s

c

a

l

a

b

i

l

i

t

y

a

n

d

r

e

p

r

o

d

u

c

i

b

i

l

i

t

y

.

<

/

p

>

TL;DR: ALLNet is a purpose-built CNN classifier that analyzes blood smear microscopy images to distinguish leukemic blast cells from healthy white blood cells, trained on a rigorously annotated public dataset.
Pages 4-6
Dataset Preparation: HSI Color Segmentation and Class Balancing

<

p

>

T

h

e

C

_

N

M

C

_

2

0

1

9

d

a

t

a

s

e

t

p

r

o

v

i

d

e

d

1

0

,

6

6

1

i

m

a

g

e

s

f

r

o

m

7

3

p

a

r

t

i

c

i

p

a

n

t

s

-

7

,

2

7

2

b

l

a

s

t

c

e

l

l

i

m

a

g

e

s

a

n

d

3

,

3

8

9

h

e

a

l

t

h

y

c

e

l

l

i

m

a

g

e

s

-

a

l

l

s

t

a

n

d

a

r

d

i

z

e

d

a

t

4

5

0

x

4

5

0

x

3

p

i

x

e

l

s

w

i

t

h

b

l

a

c

k

p

a

d

d

i

n

g

a

r

o

u

n

d

t

h

e

r

e

g

i

o

n

o

f

i

n

t

e

r

e

s

t

.

T

h

e

o

r

i

g

i

n

a

l

c

l

a

s

s

i

m

b

a

l

a

n

c

e

(

r

o

u

g

h

l

y

2

:

1

b

l

a

s

t

-

t

o

-

h

e

a

l

t

h

y

r

a

t

i

o

)

r

e

q

u

i

r

e

d

c

o

r

r

e

c

t

i

o

n

b

e

f

o

r

e

t

r

a

i

n

i

n

g

.

<

/

p

>

<

p

>

I

m

a

g

e

p

r

e

p

r

o

c

e

s

s

i

n

g

u

s

e

d

t

h

e

H

S

I

(

H

u

e

-

S

a

t

u

r

a

t

i

o

n

-

I

n

t

e

n

s

i

t

y

)

c

o

l

o

r

s

p

a

c

e

,

w

h

i

c

h

p

r

o

v

i

d

e

s

b

e

t

t

e

r

c

o

n

t

r

a

s

t

b

e

t

w

e

e

n

w

h

i

t

e

b

l

o

o

d

c

e

l

l

s

a

n

d

s

u

r

r

o

u

n

d

i

n

g

b

l

o

o

d

c

o

m

p

o

n

e

n

t

s

t

h

a

n

s

t

a

n

d

a

r

d

R

G

B

.

T

h

e

s

a

t

u

r

a

t

i

o

n

c

h

a

n

n

e

l

w

a

s

i

s

o

l

a

t

e

d

t

o

e

m

p

h

a

s

i

z

e

c

e

l

l

c

o

l

o

r

i

n

t

e

n

s

i

t

y

,

t

h

e

n

a

t

h

r

e

s

h

o

l

d

i

n

g

s

t

e

p

c

o

n

v

e

r

t

e

d

t

h

e

i

m

a

g

e

t

o

b

i

n

a

r

y

f

o

r

m

-

p

i

x

e

l

s

w

i

t

h

v

a

l

u

e

s

b

e

t

w

e

e

n

1

8

0

a

n

d

2

5

5

b

e

c

a

m

e

w

h

i

t

e

,

t

h

e

r

e

s

t

b

l

a

c

k

.

T

h

e

r

e

s

u

l

t

i

n

g

b

i

n

a

r

y

m

a

s

k

w

a

s

m

u

l

t

i

p

l

i

e

d

w

i

t

h

t

h

e

o

r

i

g

i

n

a

l

i

m

a

g

e

t

o

i

s

o

l

a

t

e

t

h

e

w

h

i

t

e

b

l

o

o

d

c

e

l

l

r

e

g

i

o

n

o

f

i

n

t

e

r

e

s

t

.

<

/

p

>

<

p

>

T

o

a

d

d

r

e

s

s

c

l

a

s

s

i

m

b

a

l

a

n

c

e

a

n

d

i

n

c

r

e

a

s

e

d

a

t

a

s

e

t

r

o

b

u

s

t

n

e

s

s

,

d

a

t

a

a

u

g

m

e

n

t

a

t

i

o

n

w

a

s

a

p

p

l

i

e

d

:

v

e

r

t

i

c

a

l

a

n

d

h

o

r

i

z

o

n

t

a

l

f

l

i

p

p

i

n

g

,

c

l

o

c

k

w

i

s

e

a

n

d

c

o

u

n

t

e

r

-

c

l

o

c

k

w

i

s

e

r

o

t

a

t

i

o

n

,

r

a

n

d

o

m

b

r

i

g

h

t

n

e

s

s

a

d

j

u

s

t

m

e

n

t

,

a

n

d

r

a

n

d

o

m

G

a

u

s

s

i

a

n

b

l

u

r

w

i

t

h

s

a

l

t

-

a

n

d

-

p

e

p

p

e

r

n

o

i

s

e

.

T

h

e

a

u

g

m

e

n

t

e

d

d

a

t

a

s

e

t

r

e

a

c

h

e

d

1

2

,

0

0

0

i

m

a

g

e

s

w

i

t

h

e

x

a

c

t

l

y

6

,

0

0

0

s

a

m

p

l

e

s

p

e

r

c

l

a

s

s

,

e

l

i

m

i

n

a

t

i

n

g

t

h

e

r

i

s

k

o

f

b

i

a

s

t

o

w

a

r

d

t

h

e

m

a

j

o

r

i

t

y

c

l

a

s

s

d

u

r

i

n

g

t

r

a

i

n

i

n

g

.

<

/

p

>

TL;DR: The team used HSI color space segmentation to isolate white blood cells in blood smear images, then applied five augmentation techniques to balance the dataset at 12,000 images before training.
Pages 7-9
ALLNet Architecture: Four Convolutional Blocks with Regularization

<

p

>

A

L

L

N

e

t

c

o

n

s

i

s

t

s

o

f

f

o

u

r

c

o

n

v

o

l

u

t

i

o

n

a

l

(

C

O

N

V

)

l

a

y

e

r

s

,

e

a

c

h

f

o

l

l

o

w

e

d

b

y

a

m

a

x

-

p

o

o

l

i

n

g

l

a

y

e

r

,

t

h

e

n

t

h

r

e

e

f

u

l

l

y

c

o

n

n

e

c

t

e

d

(

F

C

)

l

a

y

e

r

s

e

n

d

i

n

g

i

n

a

2

-

n

o

d

e

s

o

f

t

m

a

x

o

u

t

p

u

t

.

T

h

e

c

o

n

v

o

l

u

t

i

o

n

a

l

l

a

y

e

r

s

p

r

o

g

r

e

s

s

i

v

e

l

y

i

n

c

r

e

a

s

e

i

n

f

i

l

t

e

r

d

e

p

t

h

:

6

4

,

2

5

6

,

3

8

4

,

a

n

d

5

1

2

f

i

l

t

e

r

s

.

M

a

x

p

o

o

l

i

n

g

a

f

t

e

r

e

a

c

h

C

O

N

V

l

a

y

e

r

r

e

d

u

c

e

s

s

p

a

t

i

a

l

d

i

m

e

n

s

i

o

n

s

w

h

i

l

e

r

e

t

a

i

n

i

n

g

t

h

e

m

o

s

t

a

c

t

i

v

a

t

e

d

f

e

a

t

u

r

e

s

,

h

e

l

p

i

n

g

p

r

e

v

e

n

t

o

v

e

r

f

i

t

t

i

n

g

a

n

d

r

e

d

u

c

i

n

g

c

o

m

p

u

t

a

t

i

o

n

a

l

l

o

a

d

.

<

/

p

>

<

p

>

B

a

t

c

h

n

o

r

m

a

l

i

z

a

t

i

o

n

i

s

a

p

p

l

i

e

d

a

f

t

e

r

t

h

e

t

h

i

r

d

a

n

d

f

o

u

r

t

h

c

o

n

v

o

l

u

t

i

o

n

a

l

b

l

o

c

k

s

t

o

s

t

a

b

i

l

i

z

e

t

r

a

i

n

i

n

g

a

n

d

a

l

l

o

w

h

i

g

h

e

r

l

e

a

r

n

i

n

g

r

a

t

e

s

.

I

t

n

o

r

m

a

l

i

z

e

s

t

h

e

o

u

t

p

u

t

o

f

e

a

c

h

l

a

y

e

r

u

s

i

n

g

t

h

e

m

i

n

i

-

b

a

t

c

h

m

e

a

n

a

n

d

v

a

r

i

a

n

c

e

,

k

e

e

p

i

n

g

a

c

t

i

v

a

t

i

o

n

s

i

n

a

c

o

n

t

r

o

l

l

e

d

r

a

n

g

e

a

n

d

r

e

d

u

c

i

n

g

c

o

v

a

r

i

a

t

e

s

h

i

f

t

b

e

t

w

e

e

n

l

a

y

e

r

s

.

D

r

o

p

o

u

t

i

s

a

p

p

l

i

e

d

a

f

t

e

r

t

h

e

t

h

i

r

d

C

O

N

V

b

l

o

c

k

,

t

h

e

f

o

u

r

t

h

C

O

N

V

b

l

o

c

k

,

a

n

d

b

o

t

h

d

e

n

s

e

l

a

y

e

r

s

t

o

r

a

n

d

o

m

l

y

d

e

a

c

t

i

v

a

t

e

n

e

u

r

o

n

s

d

u

r

i

n

g

t

r

a

i

n

i

n

g

,

f

o

r

c

i

n

g

t

h

e

n

e

t

w

o

r

k

t

o

l

e

a

r

n

m

o

r

e

d

i

s

t

r

i

b

u

t

e

d

a

n

d

r

o

b

u

s

t

r

e

p

r

e

s

e

n

t

a

t

i

o

n

s

.

<

/

p

>

<

p

>

T

h

e

t

h

r

e

e

f

u

l

l

y

c

o

n

n

e

c

t

e

d

l

a

y

e

r

s

h

a

v

e

4

,

0

9

6

,

4

,

0

9

6

,

a

n

d

2

n

o

d

e

s

r

e

s

p

e

c

t

i

v

e

l

y

.

T

h

e

m

o

d

e

l

u

s

e

s

c

a

t

e

g

o

r

i

c

a

l

c

r

o

s

s

-

e

n

t

r

o

p

y

a

s

t

h

e

l

o

s

s

f

u

n

c

t

i

o

n

a

n

d

A

d

a

m

a

s

t

h

e

o

p

t

i

m

i

z

e

r

-

a

c

o

m

b

i

n

a

t

i

o

n

o

f

R

M

S

P

r

o

p

a

n

d

A

d

a

G

r

a

d

t

h

a

t

a

d

a

p

t

s

l

e

a

r

n

i

n

g

r

a

t

e

s

p

e

r

p

a

r

a

m

e

t

e

r

.

T

o

t

a

l

t

r

a

i

n

a

b

l

e

p

a

r

a

m

e

t

e

r

s

:

a

p

p

r

o

x

i

m

a

t

e

l

y

9

5

m

i

l

l

i

o

n

.

T

r

a

i

n

i

n

g

u

s

e

d

a

b

a

t

c

h

s

i

z

e

o

f

1

6

a

n

d

a

f

i

n

a

l

l

e

a

r

n

i

n

g

r

a

t

e

o

f

1

x

1

0

^

-

5

,

w

i

t

h

6

5

e

p

o

c

h

s

s

e

l

e

c

t

e

d

a

f

t

e

r

f

i

n

d

i

n

g

t

h

a

t

m

o

r

e

e

p

o

c

h

s

c

a

u

s

e

d

o

v

e

r

f

i

t

t

i

n

g

o

n

t

h

e

h

o

l

d

o

u

t

s

e

t

.

<

/

p

>

TL;DR: ALLNet uses 4 convolutional blocks, 3 fully connected layers, batch normalization, and dropout across 95 million parameters - trained with Adam optimizer and cross-entropy loss at a learning rate of 1e-5 for 65 epochs.
Pages 10-11
5-Fold Cross-Validation and Train-Test Split Strategy

<

p

>

T

h

e

f

u

l

l

1

2

,

0

0

0

-

i

m

a

g

e

a

u

g

m

e

n

t

e

d

d

a

t

a

s

e

t

w

a

s

s

p

l

i

t

8

0

-

2

0

i

n

t

o

t

r

a

i

n

i

n

g

a

n

d

h

o

l

d

o

u

t

t

e

s

t

s

e

t

s

.

T

h

e

8

0

%

t

r

a

i

n

i

n

g

p

o

r

t

i

o

n

w

a

s

f

u

r

t

h

e

r

d

i

v

i

d

e

d

i

n

t

o

5

f

o

l

d

s

f

o

r

c

r

o

s

s

-

v

a

l

i

d

a

t

i

o

n

,

w

h

e

r

e

e

a

c

h

f

o

l

d

s

e

r

v

e

d

a

s

a

v

a

l

i

d

a

t

i

o

n

s

e

t

o

n

c

e

w

h

i

l

e

t

h

e

o

t

h

e

r

f

o

u

r

w

e

r

e

u

s

e

d

f

o

r

t

r

a

i

n

i

n

g

.

T

h

i

s

a

p

p

r

o

a

c

h

g

u

a

r

d

s

a

g

a

i

n

s

t

o

v

e

r

f

i

t

t

i

n

g

b

y

e

n

s

u

r

i

n

g

t

h

e

m

o

d

e

l

s

e

e

s

m

u

l

t

i

p

l

e

d

i

f

f

e

r

e

n

t

v

a

l

i

d

a

t

i

o

n

s

u

b

s

e

t

s

a

c

r

o

s

s

t

r

a

i

n

i

n

g

.

<

/

p

>

<

p

>

T

h

e

b

e

s

t

-

p

e

r

f

o

r

m

i

n

g

f

o

l

d

w

a

s

t

h

e

n

u

s

e

d

t

o

e

v

a

l

u

a

t

e

p

e

r

f

o

r

m

a

n

c

e

o

n

t

h

e

u

n

t

o

u

c

h

e

d

2

0

%

h

o

l

d

o

u

t

t

e

s

t

s

e

t

c

o

n

t

a

i

n

i

n

g

2

,

1

3

2

i

m

a

g

e

s

.

A

u

g

m

e

n

t

e

d

i

m

a

g

e

s

w

e

r

e

m

i

x

e

d

w

i

t

h

o

r

i

g

i

n

a

l

s

d

u

r

i

n

g

t

r

a

i

n

i

n

g

t

o

b

r

o

a

d

e

n

t

h

e

v

a

r

i

e

t

y

o

f

s

a

m

p

l

e

s

a

n

d

i

m

p

r

o

v

e

g

e

n

e

r

a

l

i

z

a

t

i

o

n

.

E

a

c

h

f

o

l

d

w

a

s

t

r

a

i

n

e

d

f

o

r

6

5

e

p

o

c

h

s

,

a

n

d

t

h

e

a

c

c

u

r

a

c

y

a

n

d

l

o

s

s

c

u

r

v

e

s

w

e

r

e

m

o

n

i

t

o

r

e

d

t

o

c

o

n

f

i

r

m

s

t

a

b

l

e

c

o

n

v

e

r

g

e

n

c

e

w

i

t

h

o

u

t

o

v

e

r

f

i

t

t

i

n

g

.

<

/

p

>

<

p

>

T

h

e

c

o

n

f

u

s

i

o

n

m

a

t

r

i

x

r

e

v

e

a

l

e

d

t

h

a

t

o

f

t

h

e

2

,

1

3

2

t

e

s

t

i

m

a

g

e

s

,

t

h

e

m

o

d

e

l

c

o

r

r

e

c

t

l

y

c

l

a

s

s

i

f

i

e

d

1

,

9

9

6

i

m

a

g

e

s

(

9

3

.

6

%

)

,

w

i

t

h

1

,

4

5

4

c

l

a

s

s

i

f

i

e

d

a

s

b

l

a

s

t

c

e

l

l

s

a

n

d

6

7

7

a

s

h

e

a

l

t

h

y

c

e

l

l

s

.

A

c

r

o

s

s

f

i

v

e

i

n

d

e

p

e

n

d

e

n

t

t

r

a

i

n

i

n

g

r

u

n

s

,

a

c

c

u

r

a

c

y

r

a

n

g

e

d

f

r

o

m

9

4

.

5

%

t

o

9

5

.

5

4

%

,

d

e

m

o

n

s

t

r

a

t

i

n

g

c

o

n

s

i

s

t

e

n

t

a

n

d

r

e

l

i

a

b

l

e

p

e

r

f

o

r

m

a

n

c

e

r

a

t

h

e

r

t

h

a

n

a

s

i

n

g

l

e

l

u

c

k

y

o

u

t

c

o

m

e

.

<

/

p

>

TL;DR: ALLNet was evaluated using 5-fold cross-validation on 80% of the data, then tested on a held-out 20% set of 2,132 images, achieving consistent accuracy across five independent training runs.
Pages 12-13
Performance Metrics: 95.54% Accuracy and 95.91% Sensitivity

<

p

>

A

c

r

o

s

s

f

i

v

e

i

n

d

e

p

e

n

d

e

n

t

t

r

a

i

n

i

n

g

i

n

s

t

a

n

c

e

s

,

A

L

L

N

e

t

a

c

h

i

e

v

e

d

p

e

a

k

a

c

c

u

r

a

c

y

o

f

9

5

.

5

4

%

,

s

p

e

c

i

f

i

c

i

t

y

o

f

9

5

.

8

1

%

,

s

e

n

s

i

t

i

v

i

t

y

(

r

e

c

a

l

l

)

o

f

9

5

.

9

1

%

,

F

1

-

s

c

o

r

e

o

f

9

5

.

4

3

%

,

a

n

d

p

r

e

c

i

s

i

o

n

o

f

9

6

%

.

T

h

e

s

e

r

e

s

u

l

t

s

w

e

r

e

c

o

n

s

i

s

t

e

n

t

a

c

r

o

s

s

a

l

l

f

i

v

e

r

u

n

s

,

w

i

t

h

a

c

c

u

r

a

c

y

r

a

n

g

i

n

g

f

r

o

m

9

4

.

5

%

t

o

9

5

.

5

4

%

a

n

d

F

1

-

s

c

o

r

e

m

a

i

n

t

a

i

n

i

n

g

a

b

o

v

e

9

3

%

i

n

e

v

e

r

y

i

n

s

t

a

n

c

e

-

a

s

i

g

n

o

f

r

e

l

i

a

b

l

e

g

e

n

e

r

a

l

i

z

a

t

i

o

n

r

a

t

h

e

r

t

h

a

n

v

a

r

i

a

n

c

e

-

d

r

i

v

e

n

p

e

a

k

p

e

r

f

o

r

m

a

n

c

e

.

<

/

p

>

<

p

>

T

h

e

M

a

t

t

h

e

w

s

C

o

r

r

e

l

a

t

i

o

n

C

o

e

f

f

i

c

i

e

n

t

(

M

C

C

)

,

a

m

e

t

r

i

c

p

a

r

t

i

c

u

l

a

r

l

y

w

e

l

l

-

s

u

i

t

e

d

f

o

r

b

i

n

a

r

y

c

l

a

s

s

i

f

i

e

r

s

o

n

p

o

t

e

n

t

i

a

l

l

y

i

m

b

a

l

a

n

c

e

d

c

l

a

s

s

e

s

,

r

a

n

g

e

d

f

r

o

m

8

8

.

0

t

o

8

9

.

5

3

a

c

r

o

s

s

i

n

s

t

a

n

c

e

s

.

H

i

g

h

M

C

C

s

c

o

r

e

s

c

o

n

f

i

r

m

t

h

a

t

t

h

e

m

o

d

e

l

p

e

r

f

o

r

m

s

w

e

l

l

i

n

b

o

t

h

t

h

e

b

l

a

s

t

c

e

l

l

a

n

d

h

e

a

l

t

h

y

c

e

l

l

c

l

a

s

s

e

s

,

n

o

t

j

u

s

t

t

h

e

m

a

j

o

r

i

t

y

c

l

a

s

s

.

T

h

i

s

b

a

l

a

n

c

e

b

e

t

w

e

e

n

s

e

n

s

i

t

i

v

i

t

y

a

n

d

s

p

e

c

i

f

i

c

i

t

y

i

s

e

s

p

e

c

i

a

l

l

y

i

m

p

o

r

t

a

n

t

f

o

r

a

p

r

e

l

i

m

i

n

a

r

y

s

c

r

e

e

n

i

n

g

t

o

o

l

w

h

e

r

e

b

o

t

h

f

a

l

s

e

n

e

g

a

t

i

v

e

s

(

m

i

s

s

e

d

c

a

n

c

e

r

)

a

n

d

f

a

l

s

e

p

o

s

i

t

i

v

e

s

(

u

n

n

e

c

e

s

s

a

r

y

f

o

l

l

o

w

-

u

p

)

c

a

r

r

y

c

l

i

n

i

c

a

l

c

o

s

t

s

.

<

/

p

>

<

p

>

W

h

e

n

c

o

m

p

a

r

e

d

t

o

o

t

h

e

r

m

e

t

h

o

d

s

o

n

t

h

e

s

a

m

e

C

_

N

M

C

_

2

0

1

9

d

a

t

a

s

e

t

,

A

L

L

N

e

t

'

s

F

1

-

s

c

o

r

e

o

f

9

5

.

4

3

%

o

u

t

p

e

r

f

o

r

m

s

a

b

a

g

g

i

n

g

e

n

s

e

m

b

l

e

a

p

p

r

o

a

c

h

(

F

1

:

8

8

%

)

,

a

n

e

i

g

h

b

o

r

h

o

o

d

c

o

r

r

e

c

t

i

o

n

a

l

g

o

r

i

t

h

m

u

s

i

n

g

p

r

e

t

r

a

i

n

e

d

r

e

s

i

d

u

a

l

n

e

t

w

o

r

k

s

(

F

1

:

9

2

%

)

,

a

n

d

a

n

a

t

t

e

n

t

i

o

n

-

b

a

s

e

d

C

N

N

(

F

1

:

8

3

%

)

.

T

h

i

s

d

e

m

o

n

s

t

r

a

t

e

s

t

h

a

t

t

h

e

c

u

s

t

o

m

a

r

c

h

i

t

e

c

t

u

r

e

,

t

r

a

i

n

e

d

f

r

o

m

s

c

r

a

t

c

h

o

n

d

o

m

a

i

n

-

s

p

e

c

i

f

i

c

p

r

e

p

r

o

c

e

s

s

i

n

g

,

a

c

h

i

e

v

e

s

c

o

m

p

e

t

i

t

i

v

e

r

e

s

u

l

t

s

w

i

t

h

o

u

t

r

e

q

u

i

r

i

n

g

t

r

a

n

s

f

e

r

l

e

a

r

n

i

n

g

f

r

o

m

l

a

r

g

e

p

r

e

t

r

a

i

n

e

d

m

o

d

e

l

s

.

<

/

p

>

TL;DR: ALLNet achieved 95.54% accuracy, 95.91% sensitivity, and 95.43% F1-score on the C_NMC_2019 test set - outperforming ensemble and attention-based approaches on the same dataset.
Pages 13-14
Clinical Potential and Future Directions for Automated ALL Screening

<

p

>

A

L

L

N

e

t

d

e

m

o

n

s

t

r

a

t

e

s

t

h

a

t

a

p

u

r

p

o

s

e

-

b

u

i

l

t

C

N

N

t

r

a

i

n

e

d

o

n

p

r

e

p

r

o

c

e

s

s

e

d

b

l

o

o

d

s

m

e

a

r

i

m

a

g

e

s

c

a

n

r

e

l

i

a

b

l

y

d

e

t

e

c

t

A

L

L

a

t

a

c

c

u

r

a

c

y

l

e

v

e

l

s

s

u

i

t

a

b

l

e

f

o

r

c

l

i

n

i

c

a

l

p

r

e

-

s

c

r

e

e

n

i

n

g

.

T

h

e

m

o

d

e

l

d

o

e

s

n

o

t

r

e

q

u

i

r

e

m

a

n

u

a

l

f

e

a

t

u

r

e

e

n

g

i

n

e

e

r

i

n

g

a

n

d

p

r

o

c

e

s

s

e

s

t

h

e

f

u

l

l

i

m

a

g

e

r

a

t

h

e

r

t

h

a

n

a

f

e

a

t

u

r

e

v

e

c

t

o

r

s

u

b

s

e

t

,

w

h

i

c

h

r

e

d

u

c

e

s

t

h

e

r

i

s

k

o

f

m

i

s

s

i

n

g

r

e

l

e

v

a

n

t

v

i

s

u

a

l

i

n

f

o

r

m

a

t

i

o

n

d

u

r

i

n

g

c

l

a

s

s

i

f

i

c

a

t

i

o

n

.

<

/

p

>

<

p

>

T

h

e

s

y

s

t

e

m

i

s

d

e

s

i

g

n

e

d

t

o

i

n

t

e

g

r

a

t

e

i

n

t

o

t

h

e

e

x

i

s

t

i

n

g

c

o

m

p

l

e

t

e

b

l

o

o

d

c

o

u

n

t

(

C

B

C

)

a

n

d

p

e

r

i

p

h

e

r

a

l

b

l

o

o

d

s

m

e

a

r

t

e

s

t

i

n

g

w

o

r

k

f

l

o

w

a

s

a

n

a

u

t

o

m

a

t

e

d

f

i

r

s

t

-

p

a

s

s

f

i

l

t

e

r

.

B

y

f

l

a

g

g

i

n

g

s

u

s

p

i

c

i

o

u

s

s

a

m

p

l

e

s

f

o

r

p

a

t

h

o

l

o

g

i

s

t

r

e

v

i

e

w

r

a

t

h

e

r

t

h

a

n

r

e

p

l

a

c

i

n

g

h

u

m

a

n

j

u

d

g

m

e

n

t

,

A

L

L

N

e

t

c

o

u

l

d

r

e

d

u

c

e

d

i

a

g

n

o

s

t

i

c

d

e

l

a

y

s

a

n

d

e

r

r

o

r

r

a

t

e

s

i

n

s

e

t

t

i

n

g

s

w

i

t

h

h

i

g

h

t

h

r

o

u

g

h

p

u

t

o

r

l

i

m

i

t

e

d

s

p

e

c

i

a

l

i

s

t

a

v

a

i

l

a

b

i

l

i

t

y

.

<

/

p

>

<

p

>

F

u

t

u

r

e

w

o

r

k

i

d

e

n

t

i

f

i

e

d

b

y

t

h

e

a

u

t

h

o

r

s

i

n

c

l

u

d

e

s

e

x

p

a

n

d

i

n

g

t

h

e

t

r

a

i

n

i

n

g

d

a

t

a

s

e

t

w

i

t

h

n

o

i

s

i

e

r

r

e

a

l

-

w

o

r

l

d

c

l

i

n

i

c

a

l

i

m

a

g

e

s

t

h

a

t

r

e

q

u

i

r

e

l

e

s

s

p

r

e

p

r

o

c

e

s

s

i

n

g

,

i

n

t

e

g

r

a

t

i

n

g

e

x

p

l

a

i

n

a

b

i

l

i

t

y

m

e

t

h

o

d

s

s

u

c

h

a

s

G

r

a

d

-

C

A

M

t

o

p

r

o

v

i

d

e

v

i

s

u

a

l

j

u

s

t

i

f

i

c

a

t

i

o

n

s

f

o

r

e

a

c

h

c

l

a

s

s

i

f

i

c

a

t

i

o

n

,

a

n

d

e

x

p

l

o

r

i

n

g

a

l

t

e

r

n

a

t

i

v

e

a

r

c

h

i

t

e

c

t

u

r

e

s

i

n

c

l

u

d

i

n

g

Y

O

L

O

v

4

,

R

e

s

N

e

t

,

a

n

d

A

l

e

x

N

e

t

t

h

a

t

m

a

y

o

f

f

e

r

f

u

r

t

h

e

r

p

e

r

f

o

r

m

a

n

c

e

g

a

i

n

s

o

n

l

e

u

k

e

m

i

a

d

e

t

e

c

t

i

o

n

t

a

s

k

s

.

<

/

p

>

TL;DR: ALLNet is ready for evaluation as a clinical pre-screening tool, with future work focused on noisier training data, explainability integration, and comparisons with YOLO and ResNet architectures.
Citation: Open Access, 2022. Available at: PMC9601337.