Evaluating the Impact of Pseudo-Colour and Coordinate System on the Detection of Medication-induced ECG Changes

Abstract

The electrocardiogram (ECG), consisting of complex signal data representing the heart’s electrical activity, is used for detecting cardiac pathologies. Certain medications can produce a complication known as ‘long QT syndrome’, shown on the ECG as an increased gap between two parts of the waveform. Self-monitoring for this could be lifesaving, as the syndrome can result in sudden death, but detecting it on the ECG is difficult. Here we evaluate whether introducing a pseudo-colour and changing the coordinate system can support lay people in identifying increases in the QT interval. The results show that introducing colour significantly improves accuracy, and that whilst it is easier to detect a difference without colour with Cartesian coordinates, the greatest accuracy is achieved when Polar coordinates are combined with colour. The results show that applying simple visualisation techniques has the potential to improve ECG interpretation accuracy, and support people in monitoring their own ECG.

Full text

Evaluating the Impact of Pseudo-Colour and
Coordinate System on the Detection of
Medication-induced ECG Changes
Alaa Alahmadi1, Alan Davies1, Jennifer Royle2, Markel Vigo1, Caroline Jay1
1School of Computer Science, The University of Manchester
2CRUK Manchester Institute, and The Christie NHS Foundation
Manchester, UK
alaa.alahmadi@postgrad.manchester.ac.uk,Jenny.Royle@digitalecmt.org
[alan.davies-2,markel.vigo,caroline.jay]@manchester.ac.uk
ABSTRACT
The electrocardiogram (ECG), a graphical representation of
the heart’s electrical activity, is used for detecting cardiac
pathologies. Certain medications can produce a complica-
tion known as ‘long QT syndrome’, shown on the ECG as
an increased gap between two parts of the waveform. Self-
monitoring for this could be lifesaving, as the syndrome can
result in sudden death, but detecting it on the ECG is dicult.
Here we evaluate whether using pseudo-colour to highlight
wave length and changing the coordinate system can support
lay people in identifying increases in the QT interval. The
results show that introducing colour signicantly improves
accuracy, and that whilst it is easier to detect a dierence
without colour with Cartesian coordinates, the greatest accu-
racy is achieved when Polar coordinates are combined with
colour. The results show that applying simple visualisation
techniques has the potential to improve ECG interpretation
accuracy, and support people in monitoring their own ECG.
CCS CONCEPTS
•Human-centered computing →Visualization tech-
niques;
KEYWORDS
Visualisation; Visual Perception; ECG; Drug-induced LQTS
Permission to make digital or hard copies of all or part of this work for
personal or classroom use is granted without fee provided that copies
are not made or distributed for prot or commercial advantage and that
copies bear this notice and the full citation on the rst page. Copyrights
for components of this work owned by others than the author(s) must
be honored. Abstracting with credit is permitted. To copy otherwise, or
republish, to post on servers or to redistribute to lists, requires prior specic
permission and/or a fee. Request permissions from permissions@acm.org.
CHI 2019, May 4–9, 2019, Glasgow, Scotland UK
©
2019 Copyright held by the owner/author(s). Publication rights licensed
to ACM.
ACM ISBN 978-1-4503-5970-2/19/05.. .$15.00
https://doi.org/10.1145/3290605.3300353
ACM Reference Format:
Alaa Alahmadi
1
, Alan Davies
1
, Jennifer Royle
2
, Markel Vigo
1
, Caro-
line Jay
1
. 2019. Evaluating the Impact of Pseudo-Colour and Coordi-
nate System on the Detection of Medication-induced ECG Changes.
In CHI Conference on Human Factors in Computing Systems Proceed-
ings (CHI 2019), May 4–9, 2019, Glasgow, Scotland UK. ACM, New
York, NY, USA, 13 pages. https://doi.org/10.1145/3290605.3300353
1 INTRODUCTION
A side eect of commonly prescribed medications including
antihistamines, antibiotics and antidepressants is prolonga-
tion of the QT-interval, or drug-induced Long QT Syndrome
(LQTS) [
9
,
78
]. LQTS is a cardiac abnormality that can in-
crease the risk of the life-threatening arrhythmia torsades
de pointes (TdP), which can lead to loss of consciousness or
sudden death in young, otherwise healthy people [3, 20, 47,
78
]. People may not experience symptoms and an electro-
cardiogram (ECG) is often the only way to identify LQTS
[34, 60, 64].
ECGs are a graphical representation of the electrical ac-
tivity of the heart, widely used in clinical practice to assess
heart function [
62
]. ECG results are displayed as a line on
a graph-like trace, where the ‘waves’ (peaks and troughs)
are labelled with letters and represent dierent stages of the
heartbeat. The duration of the QT-interval (the time period
between the ‘Q’ and ‘T’ waves) represents the activity of the
heart ventricles (Figure 1).
Figure 1: Measurement of the QT-interval on the ECG from
the start of the Q-wave to the end of the T-wave.
CHI 2019 Paper
CHI 2019, May 4–9, 2019, Glasgow, Scotland, UK
Paper 123
Page 1

Frequent ECG monitoring is advised for people at high
risk of acquiring medication-induced LQTS [
38
,
65
]. Recent
technology innovations have made it possible to monitor
ECGs outside of the clinical environment [
55
] but this ap-
proach still relies on clinician interpretation [
4
,
41
]. This not
only increases cost, but also makes it dicult (and some-
times impossible) to manage everyone who is at high risk.
If lay people can interpret their own results, this may lead
to a step-change in the detection and management of this
potentially critical condition. However, ECG interpretation
is known to be complex, even for clinicians [
66
,
72
], and as
such little work has examined self-monitoring.
Assessing the QT-interval, in particular, is known to be
dicult [
66
,
72
], and in a prior study the majority of clini-
cians were not able to recognise it [
72
]. This may be due to
the fact that whilst people nd it easy to perceive quantity
on a vertical scale, they are poor at judging it on a horizontal
scale (see e.g. [
39
,
40
,
53
,
76
]). Artifacts in the ECG signal
can also cause misinterpretation of QT-interval length [2].
Visualisation techniques have the potential to help high-
light abnormalities within the ECG. Here we examine whether
pseudo-colouring—representing continuously varying val-
ues using a sequence of colors [
61
,
74
]—and changing the
coordinate system can support lay people in identifying in-
creases in the QT-interval. Using a psychophysical paradigm
and eye tracking to systematically examine the issue, we nd
that:
(1)
Pseudo-colouring signicantly increases lay individ-
uals’ ability to identify increases in the QT-interval,
even when T-wave morphology is abnormal.
(2)
Coordinate system interacts with colour, such that
people are most accurate in the condition where the
ECG is presented using polar coordinates and pseudo-
colour, and least accurate when presentation occurs
with polar coordinates and no colour.
(3)
According to eye tracking data, pseudo-colour helps
to focus visual attention, and people are most accurate
when using the polar coordinates as this concentrates
colour in the center of the screen.
(4)
People are signicantly more satised when pseudo-
colour is used.
2 BACKGROUND
Previous research providing a foundation for the current
study is described in this section. This covers: (1) ECG in-
terpretation; (2) ECG visualisation methods; and (3) human
perception of visualised data.
ECG Interpretation
An ECG trace is a cyclic time series with each cycle repre-
senting a new heartbeat. The electrical activity is detected
via leads placed on the body, where each lead produces a dif-
ferent electrical ‘view’ of heart activity. In hospital, clinicians
commonly interpret short (10 second) ECGs via 12-leads [
48
].
This is the most comprehensive view, but useful ECG infor-
mation can be gathered from a single lead ‘view’ (often used
by mobile technologies). LQTS is detected by measuring from
the beginning of the Q-wave to the end of the T-wave (iden-
tied using the tangent drawn at the maximum downslope
of the T-wave) [3, 21] as shown in Figure 1.
The standard method for visualising ECG data is a Carte-
sian line graph showing the voltage of the heart on the Y-axis,
and time in milliseconds (msec) on the X-axis [
5
]. A back-
ground grid supports the reader in measuring duration. To
measure the QT-interval, the interpreter counts the small
squares (each representing 40 msec) from the beginning of
the Q-wave to the end of the T-wave [3, 21, 56].
Automated ECG interpretation was introduced in the 1950s
to assist clinicians who had less training in ECG interpreta-
tion [
57
]. It remains far from perfect, and even the best com-
putational methods can produce signicant errors [
30
,
63
].
Research has shown that QT-interval is underestimated or
unreported by computational methods [
18
,
36
,
45
,
58
,
69
,
70
].
The main challenge lies in identifying the end of the T-wave,
especially when the morphology (shape) of the T-wave is
non-standard. [
17
,
21
,
26
,
46
]. This is particularly problem-
atic, as QT-prolonging drugs often aect the morphology of
the T-wave, with some drugs (e.g. quinidine and ranolazine)
causing large T-wave morphology changes [71].
Each person has a unique baseline ECG that reects their
individual heart function: health status, age, gender and eth-
nicity all inuence the ECG in general, and the QT-interval
in particular [
21
,
27
,
43
]. This complicates population-level
computer-derived QT calculations. Abstracting the QT inter-
val numerically also risks masking other potentially abnor-
mal clinically signicant changes in the ECG. For instance,
specic T-wave patterns can aid detection of drug-induced
LQTS [
11
], and large T-U waves are known to precede the life-
threating arrhythmia Torsades de Pointes [
35
]. As such the
ECG morphology continues to provide the richest informa-
tion for recognising LQTS. Current automated methods are
thus a supplement to, rather than a substitute for, the human
eye, and a combination of computer-visualisation methods
with the gold-standard human interpretation remains the
most accurate and reliable method [
14
,
18
,
36
,
45
,
58
,
63
,
69
].
ECG Visualisation
A number of studies have examined the eectiveness of vi-
sualisation techniques in supporting clinician-interpretation
of ECGs. Chiang et al. [
10
] integrated ECG signals from the
periodical and limb leads into two images of electrical heart
function. This enabled clinicians to observe an overall in-
tegral heart view, which aided interpretation when viewed
CHI 2019 Paper
CHI 2019, May 4–9, 2019, Glasgow, Scotland, UK
Paper 123
Page 2

alongside the 12-lead ECG. Kors et al. [
37
] presented a mirror
image that converted the 12 leads to 24, and was shown to
be eective in improving the detection-rate of heart attacks.
Madias et al. [
44
] used a 13th, multi-use lead, which provided
a further ‘view’ of the heart. When exploring large scale ECG
data, a glyph-based interactive system has been shown to
be eective in detecting arrhythmia [
77
]. Vectorial methods
have been used to represent direction and magnitude data
[
16
,
49
] and spatial visualisations have presented the ECG
on a body surface potential map [
42
,
68
]. As these methods
provide data with respect to further dimensions of the heart,
they are useful as a supplement, but do not replace the stan-
dard method [
7
]. Furthermore, previous work in this area
focused on aiding clinicians. Here, the aim is to help lay-
people identify when their ECG is dierent (i.e. has deviated)
from their normal baseline, so they know when to seek help.
Human Perception of Visualised Data
Here, we consider the problem from the perspective of the lay
interpreter, rather than the data, using knowledge of visual
perception to enhance the way the ECG is presented. In
particular, we draw from the eld of pre-attentive processing,
which outlines a set of visual properties known to be detected
rapidly and accurately by the human eye [
50
]. Examples of
pre-attentive properties include colour, form, and spatial
positioning. Using these properties in design can improve
both the eectiveness and the eciency of a visualisation
[24, 25, 74].
Colour is a pre-attentive attribute that is noticed without
conscious eort [
22
,
50
]. Many studies have shown the ef-
fectiveness of using colour to separate visual elements from
their surroundings, saving the user from having to carry out a
linear visual search [
23
,
54
,
74
]. A useful technique is pseudo-
colouring, which represents continuously varying values
using a sequence of colors [
74
]. Pseudo-colouring is com-
monly used in geo- and time-series visualisations [
61
,
74
].
Figure 2 shows an example of using pseudo-colouring to
show changes in temperature over time.
Adnan et al. [
1
] have examined perception of time-series
visualisations. They showed Cartesian coordinates to be most
eective for detecting trends and identifying maximum and
minimum values when used with positional and colour visual
encodings, and Polar coordinates to be most eective for
nding minimum values when using area visual encoding.
The circular layout used in the Polar coordinate system
has also been employed to perceive changes in data over
time. Page et al. [
51
,
52
] proposed an "ECG Clock" generator,
to visualise the changes in QT interval values automatically
generated by a 24-hour Holter ECG monitor. Circular layouts
have been also used to detect symmetrical patterns in data
[28] and to measure symmetry in graphs [75].
3 METHOD
Measuring Visual Perception
To systematically evaluate the eectiveness of pseudo-colour
and coordinate system in supporting lay people’s assessment
of the QT interval, we use methods from psychophysics and
eye-tracking research. Psychophysical experiments investi-
gate the relationship between physical stimuli and human
perception, by varying the properties of a stimulus along
one or more physical dimensions [
67
]. Eye-tracking is used
to quantify visual behavior when performing a given task, to
understand dierences in locus and level of attention [13].
ECG Data Acquisition
The ECG datasets were taken from a clinical trial that as-
sessed the eect of known QT-prolonging medication on
healthy subjects [
32
]. As our work is motivated by support-
ing self-monitoring, we selected data from a single partici-
pant, whose QT-interval was seen to rise to clinically dan-
gerous levels. The subject was a 35-year-old male who had
normal QT-intervals (< 430 milliseconds) prior to taking the
medication "Dofetilide" (an antiarrhythmic drug); he subse-
quently experienced a gradual increase in the QT-interval,
which eventually reached very high levels (QT=579 millisec-
onds). The ECGs sampled all have a regular heart rate (HR=60
BPM) and are from lead-II, which is typically used to mea-
sure the QT-interval. The QT-values of the selected ECGs
were 417, 421, 441, 447, 455, 468, 485, 537, 565 and 579 mil-
liseconds. We categorised these values based on their clinical
signicance: normal (QT < 430); borderline (QT > 430 and <
470); prolonged (QT > 470 and < 500); very prolonged levels
(QT > 500) [
33
]. The open ECG dataset is available from the
PhysioNet database [19].
Visualisation Design
We used a co-design approach, creating the visualisation
techniques with an expert in ECG interpretation (to ensure
accuracy), and rening them with input from lay people. As a
rst step, R-peaks were detected in the raw ECG datasets, and
a dashed vertical line used to show the halfway point of the
R-R interval (Figure 1). This helps to identify the area of inter-
est containing the QT-interval. Note that it is easy to detect
the R wave in the vast majority of ECGs, as (unlike the other
waveforms which vary considerably) it consistently has the
greatest amplitude. We then applied pseudo-colouring, to
shift the ‘work’ of QT interval visual encoding from per-
ceiving distance between two waves, to perceiving colour in
terms of hue and intensity.
As spectrum-approximation sequences in particular help
with reading values [
73
], we used these as a foundation
for the pseudo-colouring technique. Cool spectral colour
codes (purple to blue to green) were used to indicate normal
CHI 2019 Paper
CHI 2019, May 4–9, 2019, Glasgow, Scotland, UK
Paper 123
Page 3

Figure 2: Using pseudo-colouring to represent temperature in time series data over the 12 months of the year [61].
QT-interval ranges, and warm colours (yellow to orange to
red) to show abnormal QT-interval ranges. We applied the
pseudo-colouring sequence in the area between the 0 voltage
baseline up to (or down to) the signal, from the beginning
of the R-wave to the R-R interval halfway point (see e.g.
Figure 3). The pseudo-colouring sequence was mapped to
the ECG signal such that the colour code changed every 40
milliseconds, which is equal to a small square on the standard
ECG background grid.
To understand the impact of coordinate system on ECG
data interpretation, we displayed the ECG signals on Carte-
sian and Polar coordinates with and without pseudo-colouring.
We used R [
31
] with RStudio software version 1.1.447 to cre-
ate the visualisations. Figure 3 shows ECGs with normal and
very prolonged QT intervals with and without the pseudo-
colouring sequence on Cartesian coordinates. Figure 4 shows
the same ECGs, but on Polar coordinates. The ECGs are re-
duced in size for inclusion in the paper. The full size images,
along with the scripts used to create them, can be found in
the supplementary materials, and in our repository1.
Experiment Design
We hypothesized that changes in the T-wave morphology
(e.g. attening of the wave, which can be caused by QT-
prolonging medication [
71
]) might cause misperception of
the QT-interval, and included this as a factor. The study thus
used a counterbalanced within-subjects design with three
independent variables, each with two levels:
(1) Colour-coding: no colouring; pseudo-colouring.
(2) Coordinate system: Cartesian; Polar.
(3) The T-wave morphology: normal; abnormal.
The within-subjects factorial design yielded a total of 8
(2x2x2) experimental conditions for each participant. We
counterbalanced the order of visualisation presentation us-
ing a balanced Latin square to minimize practice eects. We
assessed the eects of T-wave morphology in two separate
tasks (described below). The order of the T-wave morphol-
ogy condition (normal or abnormal) was counterbalanced
across participants. The dependent variables were response
correctness, reaction time, xation location, mean xation
duration, and satisfaction.
1https://github.com/mbchxaa6/ECG_QT_Visualisation.
Participants
Forty two participants (22 males and 20 females) were re-
cruited from a university campus. Eligibility for the study
was determined by asking participants to rate their knowl-
edge of ECGs/ECG interpretation, and including only people
who reported no knowledge at all. Participants consisted
of 34 students and 8 sta. The mean age was 30 (SD=7).
The backgrounds of the participants were Computer Science
(n=27), Education (n=3), Chemical Engineering (n=3), Elec-
trical Engineering (n=3), Mathematics (n=4), History and
Sociology (n=1) and Music/Violin Performance (n=1). Their
sight was normal or corrected-to-normal and they reported
no motor or neurological disorders.
Task and Procedure
Participants completed a 10 minute training session where
they were introduced to the ECG trace and shown how to
identify the QT-interval, and then shown the visualisation
techniques used in the experiment. Each participant then
completed an assessment task to check that they understood
how to perform the measurement, where they were asked
to highlight the start and end point of the QT-intervals on
two dierent ECGs using the four visualisation techniques.
The experiment used a "two alternative forced choice"
(2AFC) psychophysical discrimination task [
67
]. Within a
trial, the participant was presented with two ECG stimuli;
a baseline showing no QT-prolongation and a comparator
showing an increased (or the same) QT-interval; the partici-
pant had to select the ECG that they perceived to have the
longer QT-interval using the left/right arrows in the Polar
condition, and up/down arrows in the Cartesian condition,
according to the stimulus’ position on the screen.
Participants completed all trials for one visualisation tech-
nique before moving to the next. The location of the ECG
with the longer QT interval (i.e. top/bottom or left/right) was
counterbalanced in the design; stimuli were then presented
at random.
To determine the eects of T-wave morphology, we split
the experiment into two separate 2AFC tasks, as follows.
Normal-T-wave:
In this condition, participants completed
a total of 20 experimental trials. In each trial, two ECGs were
presented; a baseline ECG showing a normal QT-interval
with a normal T-wave morphology and a comparator ECG.
Two trials showed exactly the same ECG for the baseline
CHI 2019 Paper
CHI 2019, May 4–9, 2019, Glasgow, Scotland, UK
Paper 123
Page 4

Figure 3: ECGs with normal and very prolonged QT intervals on Cartesian coordinates with (A) and without (B) pseudo-colour.
and the comparator stimuli, in order to test the validity of
our method (the probability of choosing each alternative
should be equal to 0.5). The other 18 trials presented the
same ECG baseline (QT = 417 msec), and an ECG with a
longer QT-interval that was selected from the following set
of QT values, where each value was presented twice: 421,
441, 447, 455, 468, 485, 537, 565, 579 milliseconds. Figure 4
and Figures 3 show examples of the ECGs used in the normal
T-wave morphology condition.
Abnormal T-wave:
This condition was used to evaluate
whether the visualisation techniques can help people to per-
ceive QT-prolongation regardless of the T-wave morphology.
To reduce potential fatigue, this condition contained only
8 trials. In each trial, two ECGs were presented: a baseline
ECG showing a borderline QT-interval with an abnormal
(at) T-wave morphology and an ECG with an increased
QT-interval. Participants had to choose the ECG with the
longer QT-interval. The borderline ECG had a QT value of
447 millisecond and the comparator ECG was selected from
the following set of QT values, which had either a normal
or abnormal T-wave as indicated: 468 (abnormal), 485 (nor-
mal), 565 (abnormal) and 579 (normal) milliseconds. Figure 5
shows examples of the ECGs used in the abnormal T-wave
morphology condition.
Apparatus
A Tobii Pro Spectrum eye-tracker and Tobii Pro lab 1.95 soft-
ware were used to record eye gaze with a sampling rate of
600 HZ. Key press events were recorded to collect partici-
pants responses. The study was performed on a 23.8 inch
(diagonal) Tobii Pro Spectrum eye-tracking monitor, with
a resolution of 1920 x 1080 pixels. Each Cartesian coordi-
nate ECG stimulus was 32.31cm x 6.14cm, and each Polar
coordinate stimulus was 15.61cm x 12.93cm.
4 RESULTS
All anonymised raw data, along with relevant R-scripts and
SPSS outputs are available in our Github repository2.
Accuracy
Psychometric function
.We used a psychometric func-
tion, which is an inferential model employed in psychophysi-
cal detection and discrimination tasks, to model the relation-
ship between the gradual increase in the QT-interval and the
correctness of participants’ responses across the four visuali-
sation techniques. The psychometric function was plotted as
the percentage of correct responses (trials where the longer
QT-interval stimulus was correctly identied) as a function
of the QT-interval increase (Figure 6).
The results show that pseudo-colour signicantly improves
perception of QT-interval increases regardless of the T-wave
morphology, with people able to detect smaller increases
with the Polar coordinates than the Cartesian coordinates.
This is important, as even these small increases are clinically
signicant.
When pseudo-colour is not used, T-wave morphology in-
teracts with coordinate system in the detection of the QT-
interval increases. When the T-wave morphology is normal,
people perform better with Cartesian coordinates (Figure 6
(A)). However, when the T-wave morphology is abnormal
(increased attening of the T-wave), people perform better
with Polar coordinates (Figure 6 (B)).
Just noticeable dierence (JND) threshold
.In psy-
chophysics, the JND threshold is dened as the minimum
amount of change in a stimulus necessary for it to be ‘just no-
ticeable’. In this study, we dened it as the minimum increase
in the QT-interval required for it to be detectable. We esti-
mated the 75% JND threshold as the value of the QT-interval
increase from the normal baseline at which the percentage
of correct responses is equal to 75%. Only the normal T-wave
morphology condition was used for estimating the JND, as
the abnormal condition contained insucient trials for it
2https://github.com/mbchxaa6/Data_Analysis.
CHI 2019 Paper
CHI 2019, May 4–9, 2019, Glasgow, Scotland, UK
Paper 123
Page 5

Figure 4: ECGs with normal and very prolonged QT intervals on Polar coordinates with (A) and without (B) pseudo-colour.
to be tted with this statistical model. The JND thresholds,
determined by tting the psychometric function using a lo-
gistic function with maximum likelihood estimation (MLE)
(Figure 7), were 29, 19, 65 and 9 milliseconds for Cartesian,
Cartesian with pseudo-colour,Polar and Polar with pseudo-
colour respectively. Pseudo-colour thus reduces the JND in
both co-ordinate systems, with the eect being strongest for
Polar co-ordinates.
Reaction Time
We measured reaction time as the period between the ap-
pearance of the stimuli on the screen and the key press event
when people made their decision. As shown in Figure 8,
pseudo-colour reduced the reaction time as the QT-interval
increased in all conditions.
A Shapiro-Wilks test showed the reaction time data was
not normally distributed (p < 0.05). We thus used a non-
parametric Friedman test to compare reaction times across
CHI 2019 Paper
CHI 2019, May 4–9, 2019, Glasgow, Scotland, UK
Paper 123
Page 6

Figure 5: ECGs with abnormal T-wave morphology on Polar coordinates with (A) and without (B) pseudo-colour.
Figure 6: The psychometric function plot shows the percentage of correct responses as a function of the QT-interval increase
from the baseline with (A) Normal T-wave morphology and (B) Abnormal T-wave morphology.
the four conditions. The test was conducted for each QT-
interval increase and under each condition of the T-wave
morphology. For all QT-interval increases, there was a sta-
tistically signicant dierence in reaction time according to
visualisation technique, under both conditions of the T-wave
morphology (p < 0.05) (Table 1).
To examine where the dierences actually occur, post hoc
pairwise comparisons were performed using a Wilcoxon
signed-rank test with Bonferroni correction (
α=
0
.
008).
This showed that when the QT-interval was clinically pro-
longed (equal to 485 msec and increased from the baseline
by 68 msec) or very prolonged (greater than 500 msec and
increased from the baseline by over 100 msec), reaction time
was signicantly faster when pseudo-colour was used, for
both types of coordinate system (p < 0.008), regardless of
T-wave morphology.
When the QT-interval was in the borderline range, the
T-wave morphology and coordinate system interacted with
pseudo-colour. In the trial that shows a borderline QT-interval
(increased by 38 msec) with a normal T-wave morphology, re-
action time when pseudo-colour was used was signicantly
faster for Polar coordinates than Cartesian coordinates (Z =
-3.806, p < 0.008, Figure 8 (A)). However, in the trial showing
a borderline QT-interval (increased by 21 msec), but with an
abnormal T-wave morphology, there was the opposite eect,
with people responding faster in the Cartesian with pseudo-
colouring condition than in the Polar with pseudo-colouring
condition (Z = -2.870, p < 0.008, Figure 8 (B)).
CHI 2019 Paper
CHI 2019, May 4–9, 2019, Glasgow, Scotland, UK
Paper 123
Page 7

Figure 7: (A) The tted psychometric function plot shows the proportion of correct responses as a function of the QT-interval
increase from the normal baseline. (B) The just noticeable dierence (JND) thresholds plot. The error bars represent bootstrap
condence intervals.
Figure 8: Mean reaction time in seconds over the QT-interval
increases (msec) from the baseline with (A) Normal T-wave
morphology (B) Abnormal T-wave morphology. Error bars
represent 95% condence intervals.
Eye-tracking Metrics
To calculate eye movement metrics, Tobii Pro lab software
was used to create two areas of interest (AOIs) for each
Table 1: Results of the Friedman test comparing re-
action times in the four visualisation conditions for
all QT-interval increases, and in each condition of the
T-wave morphology. QT represents the value of the
longer QT-interval in milliseconds. ∆QT represents
the dierence in milliseconds between the value of
the longer QT-interval and the baseline QT-interval
(i.e. the amount of QT-interval increase).
T−wave QT ∆QT Ranдeχ2(3)p−value
Normal 421 4 normal 9.543 p<0.05
morpholoдy 441 24 borderl ine 68.100 p<0.05
447 30 borderl ine 13.443 p<0.05
455 38 borderl ine 101.954 p<0.05
468 51 borderl ine 41.471 p<0.05
485 68 prolonдed 101.886 p<0.05
537 120 veryprolonдed 79.225 p<0.05
565 148 veryprolonдed 90.286 p<0.05
579 162 veryprolonдed 72.265 p<0.05
Abnormal 468 21 border line 50.671 p<0.05
morpholoдy 485 38 prolonдed 114.783 p<0.05
565 118 veryprolonдed 138.529 p<0.05
579 132 veryprolonдed 94.409 p<0.05
experimental trial: one for the baseline ECG stimulus, and
one for the comparator ECG stimulus.
Mean fixation duration
.The mean xation duration
metric, which is an indicator of cognitive load [12, 29], was
CHI 2019 Paper
CHI 2019, May 4–9, 2019, Glasgow, Scotland, UK
Paper 123
Page 8

used to understand whether the visualisation techniques
helped people to focus on the target ECG stimulus that had
the longer QT-interval. As shown in Figure 9, regardless of
T-wave morphology and coordinate system, using pseudo-
colour results in longer xations on the stimulus with the
longer QT-interval, compared with the baseline stimulus,
and this eect becomes more pronounced as the QT-interval
increases.
Satisfaction
Following the experiment, participants completed a ve
point Likert-type scale ranging from ‘bad’ (1) to ‘good’ (5)
to rate the eectiveness of each visualisation technique in
supporting the detection of increases in the QT-interval. A
Friedman test showed there to be a statistically signicant
dierence in satisfaction depending on which visualisation
technique was used (
χ2(
3
)=
90
.
860
,p<
0
.
05). A post-hoc
analysis with a Wilcoxon signed-rank test utilising a Bonfer-
roni correction (
α=
0
.
008) showed a signicant preference
for pseudo-colour (p < 0.008). However, although people
were faster and more accurate in the Polar coordinate condi-
tion when pseudo-colour was used, there was no dierence
in people’s satisfaction for either coordinate system (Z =
-0.435, p = 0.664).
5 DISCUSSION
Recognizing QT-interval prolongation on the standard ECG
is dicult. A previous study with medical professionals has
shown that accurate classication of the QT-interval as ei-
ther “prolonged” or “normal” was achieved by 96% of QT
experts and 62% of arrhythmia experts, but by less than 25%
of cardiologists and noncardiologists [
72
]. The QT-interval is
also underestimated or unreported by computerised methods
[
18
,
36
,
45
,
58
,
69
,
70
], and as such, human visual validation
is strongly recommended [
14
,
63
]. To support lay people,
who have no experience in ECG interpretation, in detecting
life-threatening changes in the ECG, we need to understand
how people perceive ECG data, and the extent to which
visualisation techniques can aid the interpretation process.
We used psychophysical methods to model lay people’s
detection of QT-interval increases when using four visualisa-
tion techniques. The results show that using pseudo-colour
to represent time signicantly improves accuracy in detect-
ing increases in the QT-interval, for both coordinate systems.
People are most accurate in detecting small, but clinically sig-
nicant increases in the QT-interval with Polar coordinates,
regardless of whether the T-wave morphology is normal or
abnormal (Figure 6).
Clinical research has shown that even a small (
∼
10 msec)
QT-interval increase from the baseline is considered a signif-
icant side eect of a QT-prolonging drug [
8
,
15
,
59
]. When
the T-wave morphology is normal, the 75% just noticeable
dierence (JND) thresholds were 29, 19, 65 and 9 millisec-
onds for Cartesian,Cartesian with pseudo-colour,Polar and
Polar with pseudo-colour respectively. This shows that using
a combination of Polar coordinates and pseudo-colour has
the potential to support lay people in detecting the smallest
clinically signicant change. It also shows that colour can
improve sensitivity to changes such that people can perceive
increases that are much smaller than a 1mm square on the
standard ECG grid (which represents 40 msec).
As well as improving accuracy, using pseudo-colour re-
duced reaction times and increased attention to the longer
QT-interval stimulus. Eye-tracking data showed that the av-
erage xation duration increased more on the comparator
stimulus, which has the longer QT-interval, than the baseline
stimulus, as the interval length increased (Figure 9). Figure
10 shows a heat map of absolute xation duration across all
participants, demonstrating that even when the interval is
borderline, rather than prolonged (QT = 455 msec, increased
by 38 msec), people still xate longer on the comparator stim-
ulus. This shows the power of the colour codes used with the
spectrum-approximation pseudo-colouring sequence, where
warmer colours including orange and red help to attract
attention to abnormal QT-interval levels.
Visualisation Design Implications
This study shows that colour as a pre-attentive attribute
can support the detection of small dierences in time-series
data represented along a continuous scale, that are otherwise
dicult to perceive. While people nd it relatively easy to
perceive quantity along a vertical scale, they are known
to be poor at judging size or quantity displayed along a
horizontal scale (see e.g. [
39
,
40
,
53
,
76
]). Time-series data are
conventionally displayed horizontally. Although this study
focused on a specic problem within ECG interpretation, the
results may have a wider application to other forms of time-
series data, for example, in supporting detection of change in
seasonality in nancial data, or follow-up months of survival
rate among cancer treatments (e.g. in Kaplan-Meier curves).
People are able to detect the smallest dierences when
the ECG is presented using Polar coordinates and pseudo-
colour. Eye-tracking research has shown that people’s initial
eye movements are more commonly located in the center of
the screen [
6
]. According to the study’s eye-tracking data,
the warmer hues of the pseudo-colour helped to focus vi-
sual attention; as Polar coordinates concentrate more colour
in the center of the screen than Cartesian coordinates, the
increased salience may be easier to perceive in foveal vision.
6 LIMITATIONS AND FUTURE WORK
Limitations of this study include: (1) we investigated detec-
tion of QT-interval prolongation, and it is not clear whether
these techniques would generalise to interpretation of other
CHI 2019 Paper
CHI 2019, May 4–9, 2019, Glasgow, Scotland, UK
Paper 123
Page 9

4 24 30 38 51 68 120 148 162 21 38 118 132
Mean Fixation Duration (milliseconds)
A
B
4 24 30 38 51 68 120 148 162 21 38 118 132
Figure 9: Mean xation duration of the baseline and the comparator stimuli over the QT-interval increases (msec) with (A)
Normal T-wave morphology and (B) Abnormal T-wave morphology. The error bars represent 95% condence intervals.
Figure 10: Heatmap of absolute xation duration for all par-
ticipants. Fixation is longer on the borderline QT-interval.
ECG abnormalities, such as changes in ST-segment eleva-
tion, or to other signal/time-series data; (2) the data used to
design the stimuli were from a high quality signal with little
noise; they were acquired from a 12-lead ECG, not a mobile
monitoring device, where the signal is much more likely to
be aected by noise; (3) we assessed an irregularity from
the perspective of abnormal T-wave morphology, but this, of
course, is one of many; fast or slow heart rates, abnormalities
in ST-T changes and the presence of some common types
of arrhythmia such as atrial brillation (AF) can all aect
QT-interval calculation; (4) in a self-monitoring situation
people may be using tablets or phones. We hypothesise that
the visualisation techniques will still be benecial, but this
will need to be conrmed in a further study examining the
eects of screen size and lighting setting on the visualisation
techniques; and (5) our participants were highly educated,
and we do not know whether the results would generalise
to other demographics. Future work will include evaluat-
ing the visualisation techniques with more diverse clinical
populations, particularly with low-literacy and low-income
minority populations, who are taking medication that can
lead to LQTS, and are using a mobile device with a wearable
ECG monitor.
7 CONCLUSION
This study shows that using simple visualisation techniques
signicantly improves lay people’s ability to accurately mea-
sure the QT-interval. This may help with self-monitoring
drug-induced LQTS and enable treatment to be altered to
prevent the development of life threatening complications.
Whilst using a pseudo-colour sequence signicantly im-
proves people’s ability to detect increases in the QT-interval
when the ECG is displayed on a standard Cartesian coordi-
nate system, the greatest accuracy is achieved when pseudo-
colour is combined with Polar coordinates.
8 ACKNOWLEDGEMENT
This research is sponsored by Taibah University, Kingdom of
Saudi Arabia, College of Computer Science and Engineering,
Yanbu. We would like to thank the Digital Experimental
Cancer Medicine Team (dgitalECMT), Manchester, UK, for
their feedback and support.
REFERENCES
[1]
Muhammad Adnan, Mike Just, and Lynne Baillie. 2016. Investigating
time series visualisations to improve the user experience. In Proceedings
CHI 2019 Paper
CHI 2019, May 4–9, 2019, Glasgow, Scotland, UK
Paper 123
Page 10

of the 2016 CHI Conference on Human Factors in Computing Systems.
ACM, 5444–5455.
[2]
Alaa Alahmadi, Alan Davies, Markel Vigo, and Caroline Jay. 2018.
Can lay people identify a drug-induced QT-interval prolongation? A
psychophysical and eye-tracking experiment examining the ability
of non-experts to interpret an ECG. Journal of the American Medical
Informatics Association (2018).
[3]
Mark E Anderson, Sana M Al-Khatib, Dan M Roden, Robert M Cali,
Duke Clinical Research Institute, et al
.
2002. Cardiac repolarization:
current knowledge, critical gaps, and new approaches to drug develop-
ment and patient management. American heart journal 144, 5 (2002),
769–781.
[4]
Ashish Atreja, Emamuzo Otobo, Karthik Ramireddy, and Allyssa De-
orocki. 2018. Remote patient monitoring in IBD: current state and
future directions. Current gastroenterology reports 20, 2 (2018), 6.
[5]
Arlie R Barnes, Harold EB Pardee, Paul D White, Frank N Wilson,
Charles C Wolferth, et al
.
1938. Standardization of precordial leads:
supplementary report. American Heart Journal 15, 2 (1938), 235–239.
[6]
Markus Bindemann. 2010. Scene and screen center bias early eye
movements in scene viewing. Vision research 50, 23 (2010), 2577–2587.
[7]
Raymond R Bond, Dewar D Finlay, Chris D Nugent, George Moore,
and Daniel Guldenring. 2013. Methods for presenting and visualising
electrocardiographic data: from temporal signals to spatial imaging.
Journal of electrocardiology 46, 3 (2013), 182–196.
[8]
Joanna M Brell. 2010. Prolonged QTc interval in cancer therapeutic
drug development: dening arrhythmic risk in malignancy. Progress
in cardiovascular diseases 53, 2 (2010), 164–172.
[9]
A John Camm, Marek Malik, and Yee Guan Yap. 2008. Acquired long
QT syndrome. John Wiley & Sons.
[10]
Huihua Kenny Chiang, Chao-Wei Chu, Gau-Yang Chen, and Cheng-
Deng Kuo. 2001. A new 3-D display method for 12-lead ECG. IEEE
transactions on biomedical engineering 48, 10 (2001), 1195–1202.
[11]
Ehud Chorin, Ofer Havakuk, Arnon Adler, Arie Steinvil, Uri Rozovski,
Christian van der Werf, Pieter G Postema, Guy Topaz, Arthur AM
Wilde, Sami Viskin, et al
.
2015. Diagnostic value of T-wave morphology
changes during ’QT stretching’ in patients with long QT syndrome.
Heart Rhythm 12, 11 (2015), 2263–2271.
[12]
Nicolas Debue and Cécile Van De Leemput. 2014. What does germane
load mean? An empirical contribution to the cognitive load theory.
Frontiers in psychology 5 (2014), 1099.
[13]
Andrew T Duchowski. 2007. Eye tracking methodology. Theory and
practice 328 (2007).
[14]
NA Mark Estes III. 2013. Computerized interpretation of ECGs: sup-
plement not a substitute.
[15]
Food, HHS Drug Administration, et al
.
2005. International Con-
ference on Harmonisation; guidance on E14 Clinical Evaluation of
QT/QTc Interval Prolongation and Proarrhythmic Potential for Non-
Antiarrhythmic Drugs; availability. Notice. Federal register 70, 202
(2005), 61134.
[16]
Ernest Frank. 1956. An accurate, clinically practical system for spatial
vectorcardiography. circulation 13, 5 (1956), 737–749.
[17]
Pieter G Postema and Arthur AM Wilde. 2014. The measurement of
the QT interval. Current cardiology reviews 10, 3 (2014), 287–294.
[18]
Anubhav Garg and Michael H Lehmann. 2013. Prolonged QT interval
diagnosis suppression by a widely used computerized ECG analysis
system. Circulation: Arrhythmia and Electrophysiology 6, 1 (2013),
76–83.
[19]
Ary L Goldberger, Luis AN Amaral, Leon Glass, Jerey M Hausdor,
Plamen Ch Ivanov, Roger G Mark, Joseph E Mietus, George B Moody,
Chung-Kang Peng, and H Eugene Stanley. 2000. PhysioBank, Phys-
ioToolkit, and PhysioNet: components of a new research resource for
complex physiologic signals. Circulation 101, 23 (2000), e215–e220.
[20]
Ilan Goldenberg and Arthur J Moss. 2008. Long QT syndrome. Journal
of the American College of Cardiology 51, 24 (2008), 2291–2300.
[21]
ILAN Goldenberg, Arthur J Moss, and Wojciech Zareba. 2006. QT inter-
val: how to measure it and what is “normal”. Journal of cardiovascular
electrophysiology 17, 3 (2006), 333–336.
[22]
Carl Gutwin, Andy Cockburn, and Ashley Coveney. 2017. Periph-
eral Popout: The Inuence of Visual Angle and Stimulus Intensity on
Popout Eects. In Proceedings of the 2017 CHI Conference on Human
Factors in Computing Systems. ACM, 208–219.
[23]
Christopher G Healey. 1996. Choosing eective colours for data visu-
alization. In Visualization’96. Proceedings. IEEE, 263–270.
[24]
Christopher G Healey, Kellogg S Booth, and James T Enns. 1995. Vi-
sualizing real-time multivariate data using preattentive processing.
ACM Transactions on Modeling and Computer Simulation (TOMACS) 5,
3 (1995), 190–221.
[25]
Christopher G Healey, Kellogg S Booth, and James T Enns. 1996. High-
speed visual estimation using preattentive processing. ACM Transac-
tions on Computer-Human Interaction (TOCHI) 3, 2 (1996), 107–135.
[26]
PD Higham and RW Campbell. 1994. QT dispersion. British heart
journal 71, 6 (1994), 508.
[27]
Katerina Hnatkova, Peter Smetana, Ondrej Toman, Georg Schmidt,
and Marek Malik. 2016. Sex and race dierences in QRS duration. Ep
Europace 18, 12 (2016), 1842–1849.
[28]
Patrick Homan, Georges Grinstein, Kenneth Marx, Ivo Grosse, and
Eugene Stanley. 1997. DNA visual and analytic data mining. In Visual-
ization’97., Proceedings. IEEE, 437–441.
[29]
Kenneth Holmqvist, Marcus Nyström, Richard Andersson, Richard De-
whurst, Halszka Jarodzka, and Joost Van de Weijer. 2011. Eye tracking:
A comprehensive guide to methods and measures. OUP Oxford.
[30]
Richard H Hongo and Nora Goldschlager. 2006. Status of computerized
electrocardiography. Cardiology clinics 24, 3 (2006), 491–504.
[31]
Ross Ihaka and Robert Gentleman. 1996. R: a language for data analysis
and graphics. Journal of computational and graphical statistics 5, 3
(1996), 299–314.
[32]
Lars Johannesen, Jose Vicente, JW Mason, Carlos Sanabria, Kristin
Waite-Labott, Mira Hong, Ping Guo, John Lin, Jens Stampe Sørensen,
Loriano Galeotti, et al
.
2014. Dierentiating drug-induced multichan-
nel block on the electrocardiogram: randomized study of dofetilide,
quinidine, ranolazine, and verapamil. Clinical Pharmacology & Thera-
peutics 96, 5 (2014), 549–558.
[33]
Jonathan N Johnson and Michael J Ackerman. 2009. QTc: how long is
too long? British journal of sports medicine 43, 9 (2009), 657–662.
[34]
Ijaz A Khan. 2002. Clinical and therapeutic aspects of congenital and
acquired long QT syndrome. The American journal of medicine 112, 1
(2002), 58–66.
[35]
Paulus Kirchhof, Michael R Franz, Abdennasser Bardai, and Arthur M
Wilde. 2009. Giant T–U Waves Precede Torsades de Pointes in Long
QT Syndrome: A Systematic Electrocardiographic Analysis in Patients
With Acquired and Congenital QT Prolongation. Journal of the Ameri-
can College of Cardiology 54, 2 (2009), 143–149.
[36]
Paul Kligeld, Fabio Badilini, Isabelle Denjoy, Saeed Babaeizadeh,
Elaine Clark, Johan De Bie, Brian Devine, Fabrice Extramiana, Gi-
anluca Generali, Richard Gregg, et al
.
2018. Comparison of automated
interval measurements by widely used algorithms in digital electro-
cardiographs. American heart journal 200 (2018), 1–10.
[37]
Jan A Kors and Gerard van Herpen. 2008. Mirror image electrocar-
diograms and additional electrocardiographic leads: new wine in old
wineskins? Journal of electrocardiology 41, 3 (2008), 245–250.
[38]
Teri M Kozik and Shu-Fen Wung. 2012. Acquired long QT syndrome:
frequency, onset, and risk factors in intensive care patients. Critical
care nurse 32, 5 (2012), 32–41.
CHI 2019 Paper
CHI 2019, May 4–9, 2019, Glasgow, Scotland, UK
Paper 123
Page 11

[39]
Gaëlle Leroux, Jeanne Spiess, Laure Zago, Sandrine Rossi, Amélie
Lubin, Marie-Renée Turbelin, Bernard Mazoyer, Nathalie Tzourio-
Mazoyer, Olivier Houdé, and Marc Joliot. 2009. Adult brains don’t fully
overcome biases that lead to incorrect performance during cognitive
development: an fMRI study in young adults completing a Piaget-like
task. Developmental Science 12, 2 (2009), 326–338.
[40]
Lynn S Liben. 1991. Adults’ performance on horizontality tasks: Con-
icting frames of reference. Developmental Psychology 27, 2 (1991),
285.
[41] Emanuela T Locati. 2017. New directions for ambulatory monitoring
following 2017 HRS-ISHNE expert consensus. Journal of electrocardi-
ology 50, 6 (2017), 828–832.
[42]
Robert L Lux, Creig R Smith, Roland F Wyatt, and JA Abildskov. 1978.
Limited lead selection for estimation of body surface potential maps
in electrocardiography. IEEE Transactions on Biomedical Engineering 3
(1978), 270–276.
[43]
PW Macfarlane, SC McLaughlin, B Devine, and TF Yang. 1994. Ef-
fects of age, sex, and race on ECG interval measurements. Journal of
Electrocardiology 27 (1994), 14–19.
[44]
John E Madias. 2004. The 13th multiuse ECG lead: ShouldnâĂŹt we
use it more often, and on the same hard copy or computer screen, as
the other 12 leads? Journal of electrocardiology 37, 4 (2004), 285–287.
[45]
Misha D Miller, J Porter Co-burn, and Michael J Ackerman. 2001. Diag-
nostic accuracy of screening electrocardiograms in long QT syndrome
I. Pediatrics 108, 1 (2001), 8–12.
[46]
J Morganroth. 2001. Focus on issues in measuring and interpreting
changes in the QTc interval duration. European Heart Journal Supple-
ments 3, suppl_K (2001), K105–K111.
[47]
Arthur J Moss, Peter J Schwartz, Richard S Crampton, Dan Tzivoni,
Emanuela H Locati, Jean MacCluer, W Jackson Hall, Lowell Weitkamp,
G Michael Vincent, and A Garson Jr. 1991. The long QT syndrome.
Prospective longitudinal study of 328 families. circulation 84, 3 (1991),
1136–1144.
[48]
Stefan P Nelwan, Jan A Kors, Simon H Meij, Jan H van Bemmel, and
Maarten L Simoons. 2004. Reconstruction of the 12-lead electrocardio-
gram from reduced lead sets. Journal of electrocardiology 37, 1 (2004),
11–18.
[49]
Magnus O Nimmermark, John J Wang, Charles Maynard, Mauricio
Cohen, Ian Gilcrist, John Heitner, Michael Hudson, Sebastian Palmeri,
Galen S Wagner, and Olle Pahlm. 2011. The impact of numeric and
graphic displays of ST-segment deviation levels on cardiologists’ deci-
sions of reperfusion therapy for patients with acute coronary occlusion.
Journal of electrocardiology 44, 5 (2011), 502–508.
[50]
Hans-Christoph Nothdurft. 1993. The role of features in preattentive
vision: Comparison of orientation, motion and color cues. Vision
research 33, 14 (1993), 1937–1958.
[51]
Alex Page, Mehmet K Aktas, Tolga Soyata, Wojciech Zareba, and Jean-
Philippe Couderc. 2016. âĂIJQT clockâĂİ to improve detection of QT
prolongation in long QT syndrome patients. Heart Rhythm 13, 1 (2016),
190–198.
[52]
Alex Page, Tolga Soyata, Jean-Philippe Couderc, and Mehmet Aktas.
2015. An open source ECG clock generator for visualization of long-
term cardiac monitoring data. IEEE Access 3 (2015), 2704–2714.
[53]
Konstantinos Papadopoulos and Eleni Koustriava. 2011. Piaget’s water-
level task: The impact of vision on performance. Research in develop-
mental disabilities 32, 6 (2011), 2889–2893.
[54]
Ronald M Pickett, Georges Grinstein, Haim Levkowitz, and Stuart
Smith. 1995. Harnessing preattentive perceptual processes in visual-
ization. In Perceptual issues in visualization. Springer, 33–45.
[55]
Philip J Podrid. 2017. Ambulatory ECG monitoring. UpToDate,
Waltham, MA. Accessed 6 (2017), 05–18.
[56]
Dallas Price. 2010. How to read an electrocardiogram (ECG). Part 1:
Basic principles of the ECG. The normal ECG. South Sudan Medical
Journal 3, 2 (2010), 26–31.
[57]
Pentti M Rautaharju. 2007. The birth of computerized electrocardiog-
raphy: Hubert V. Pipberger (1920-1993). Cardiology journal 14, 4 (2007),
420–421.
[58]
Pentti M Rautaharju, Borys Surawicz, and Leonard S Gettes. 2009.
AHA/ACCF/HRS recommendations for the standardization and inter-
pretation of the electrocardiogram: part IV: the ST segment, T and U
waves, and the QT interval a scientic statement from the American
Heart Association Electrocardiography and Arrhythmias Committee,
Council on Clinical Cardiology; the American College of Cardiology
Foundation; and the Heart Rhythm Society Endorsed by the Inter-
national Society for Computerized Electrocardiology. Journal of the
American College of Cardiology 53, 11 (2009), 982–991.
[59]
Edwin P Rock, John Finkle, Howard J Fingert, Brian P Booth, Chris-
tine E Garnett, Stephen Grant, Robert L Justice, Richard J Kovacs,
Peter R Kowey, Ignacio Rodriguez, et al
.
2009. Assessing proarrhyth-
mic potential of drugs when optimal studies are infeasible. American
heart journal 157, 5 (2009), 827–836.
[60]
Dan M Roden and Prakash C Viswanathan. 2005. Genetics of acquired
long QT syndrome. The Journal of clinical investigation 115, 8 (2005),
2025–2032.
[61]
Takafumi Saito, Hiroko Nakamura Miyamura, Mitsuyoshi Yamamoto,
Hiroki Saito, Yuka Hoshiya, and Takumi Kaseda. 2005. Two-tone
pseudo coloring: Compact visualization for one-dimensional data.
(2005).
[62]
Leo Schamroth, Colin Schamroth, and Leo Schamroth. 1966. An intro-
duction to electrocardiography. (1966).
[63]
Jürg Schläpfer and Hein J Wellens. 2017. Computer-interpreted electro-
cardiograms: benets and limitations. Journal of the American College
of Cardiology 70, 9 (2017), 1183–1192.
[64]
Maully Shah and Christopher Carter. 2008. Long QT syndrome: A
therapeutic challenge. Annals of pediatric cardiology 1, 1 (2008), 18.
[65]
Rashmi R Shah. 2002. The signicance of QT interval in drug develop-
ment. British journal of clinical pharmacology 54, 2 (2002), 188–202.
[66]
Luke Solomons, Adrian Treloar, and Ryan Noronha. 2008. Competence
of psychiatric clinicians in interpreting electrocardiograms and QT
intervals: can they do this? Does it matter? Psychiatric Bulletin 32, 8
(2008), 291–294.
[67]
Stanley Smith Stevens. 2017. Psychophysics: Introduction to its percep-
tual, neural and social prospects. Routledge.
[68]
Bruno Taccardi. 1963. Distribution of heart potentials on the thoracic
surface of normal human subjects. Circulation Research 12, 4 (1963),
341–352.
[69]
Soheila Talebi, Azhir Alaleh, Zuber Sam, Soman Sandeep, Ferdi-
nand Visco, Holly Totouom-Tangho, Hossein Kalantari, and Getaw
Worku Hassen. 2015. Underestimated and unreported prolonged QTc
by automated ECG analysis in patients on methadone: can we rely on
computer reading? Acta cardiologica 70, 2 (2015), 211–216.
[70]
Benoît Tyl, Sara Azzam, Nathalie Blanco, and William Wheeler. 2011.
Improvement and limitation of the reliability of automated QT mea-
surement by recent algorithms. Journal of electrocardiology 44, 3 (2011),
320–325.
[71]
Jose Vicente, Lars Johannesen, Jay W Mason, William J Crumb, Esther
Pueyo, Norman Stockbridge, and David G Strauss. 2015. Comprehen-
sive T wave morphology assessment in a randomized clinical study
of dofetilide, quinidine, ranolazine, and verapamil. Journal of the
American Heart Association 4, 4 (2015), e001615.
[72]
Sami Viskin, Uri Rosovski, Andrew J Sands, Edmond Chen, Peter M
Kistler, Jonathan M Kalman, Laura Rodriguez Chavez, Pedro Iturralde
Torres, Osmar A Centurión, Akira Fujiki, et al
.
2005. Inaccurate elec-
trocardiographic interpretation of long QT: the majority of physicians
CHI 2019 Paper
CHI 2019, May 4–9, 2019, Glasgow, Scotland, UK
Paper 123
Page 12

cannot recognize a long QT when they see one. Heart Rhythm 2, 6
(2005), 569–574.
[73]
Colin Ware. 1988. Color sequences for univariate maps: Theory, ex-
periments and principles. IEEE Computer Graphics and Applications 8,
5 (1988), 41–49.
[74]
Colin Ware. 2012. Information visualization: perception for design.
Elsevier.
[75]
Eric Welch and S Kobourov. 2017. Measuring symmetry in drawings
of graphs. In Computer Graphics Forum, Vol. 36. Wiley Online Library,
341–351.
[76]
Michele Andrisin Wittig and Mary J Allen. 1984. Measurement of
adult performance on Piaget’s water horizontality task. Intelligence 8,
4 (1984), 305–313.
[77]
Ke Xu, Shunan Guo, Nan Cao, David Gotz, Aiwen Xu, Huamin Qu,
Zhenjie Yao, and Yixin Chen. 2018. ECGLens: Interactive Visual Ex-
ploration of Large Scale ECG Data for Arrhythmia Detection. In Pro-
ceedings of the 2018 CHI Conference on Human Factors in Computing
Systems. ACM, 663.
[78]
Yee Guan Yap and A John Camm. 2003. Drug induced QT prolongation
and torsades de pointes. Heart 89, 11 (2003), 1363–1372.
CHI 2019 Paper
CHI 2019, May 4–9, 2019, Glasgow, Scotland, UK
Paper 123
Page 13