Table 3 English translation of words with the largest positive and negative ranking score in calls predicted as stroke and non-stroke, respectively.

From: A retrospective study on machine learning-assisted stroke recognition for medical helpline calls

 

Stroke predictions

D = 1897

Non-stroke predictions

D = 342,133

 

Word, w (translated)

Occurrences, D(w)

Word, w (translated)

Occurrences, D(w)

1.

Ambulance

1680

Tetanus

4378

2.

Blood clot

895

Pregnant

8749

3.

Left

1108

Cut

7592

4.

Right

1050

Bandage

4561

5.

Double vision

84

Amager (a location)

23,776

6.

The words

344

O’clock

94,436

7.

Suddenly

783

The emergency room

42,809

8.

Arm

709

The police

2903

9.

Side

1,139

Swollen

60,559

10.

Stroke

117

Over-the-counter (OTC)

4641

11.

Double

113

The neck

30,151

12.

Control

134

Fever

112,586

13.

Call

39

Prescription

5450

14.

Numb

94

Centimetre

12,026

15.

Minutes

763

The knee

8875

16.

Difficulties speaking

44

The pharmacy

10,085

17.

Haemorrhagic stroke

133

The stomach

42,105

18.

Hand

297

Psychiatric

3688

19.

The ambulance

521

Pneumonia

7597

20.

Slurred speech

58

Stomach pain

10,551

21.

Blood clots

224

Stool

19,155

22.

Fast

663

The ribs

3928

23.

Express

44

Bleed

10,501

24.

Blood thinner

259

Bleeding

24,313

25.

Incoherent

15

Ribs

2941

26.

Lopsided

211

Broken

19,415

27.

Reduced

528

Inflammation

10,050

28.

Hangs

628

Common cold

8127

29.

Transient

48

Morning or morrow

78,558

30.

Not making sense

14

Swelling

17,762

  1. For this analysis, we used the model with the median F1-score out of 11 randomly seeded runs.