Download PDF
Elizaveta (Lizzie) Rakhmanova
School of Medicine
Faculty of Medical and Health Sciences
University of Auckland
Introduction
With recent advances in artificial intelligence (AI), not a week goes by
without an article with a catchy headline stating that a certain med-
ical specialty will soon be replaced by “robots”. But are such claims
substantiated? In this essay, I hope to explore this fascinating topic by
firstly, reviewing recent literature on the role of AI technologies such
as deep learning convolutional neural networks (CNNs) in the diag-
nosis of dermatological disease. Secondly, I will outline some of the
existing technologies that aim to complement contemporary derma-
tologic practices. Some examples of this include teledermatology, mo-
bile dermoscopy/dermatoscopy, and smartphone apps. Finally, I will
briefly discuss patient-centred care as relevant to AI in dermatology.
Throughout the essay, I will draw on some relevant personal experi-
ences both as a student doctor and as a patient, to hopefully, provide
the reader with additional context from my perspective. Due to the
concerningly high rates, and thus, the public health importance of mel-
anoma in New Zealand, 1 as well as the breadth of the topic, for the
purposes of this entry I am going to focus on the use of AI and other
technological tools in detection of melanocytic cancer specifically. At
the same time, I will also acknowledge that AI and technology may be
successfully utilised to diagnose other types of skin disease, too.
Recent advances of AI and machine learning in dermatology
As alluded to in the introduction, of particular interest to derma-
tology is the concept of deep learning CNNs. CNNs are artificial,
feed-forward neural networks capable of analysing and learning from
visual imagery. 2,3 CNNs are able to improve their future performance
according to their previous experiences in image recognition and
classification – this process is referred to as machine learning. 2,3 The
concept of CNNs has become especially topical after the results of
a study by Esteva et al were published in Nature last year. 4 In this
landmark study (the largest of its kind), over 100,000 biopsy-backed
clinical photographs were used to teach a deep learning CNN-based
algorithm to discern malignant skin lesions from their benign mimick-
ers. 4 When asked to differentiate between, firstly, melanomas and
benign naevi, and, secondly, keratinocyte (i.e. non-melanocytic) carci-
nomas and seborrheic keratoses, the CNN system performed com-
parably to a cohort of 21 board-certified dermatologists. 4 For the first
time, successful utilisation of a computer algorithm capable of expert
level thinking was demonstrated for a relatively subjective task, which
is of increasing importance in everyday dermatological practice.
Less than a year later, Haenssle et al reported that a deep learning
CNN that was trained specifically to distinguish dermatoscopic imag-
es of benign and malignant melanocytic lesions has shown to be, on
average, superior in both sensitivity and specificity when compared to
an international panel of 51 dermatologists. 6 More than half of those
physicians were considered experts with five plus years of dermato-
scopic experience. 6 When additional clinical information was provid-
ed to the dermatologists (to simulate the real life setting more closely),
their overall sensitivity was improved, yet the algorithm still outper-
formed clinicians in terms of specificity. 5 Therefore, it was suggested
that a competently trained CNN may be a helpful addition to any der-
matologist’s diagnostic toolbox, regardless of their level of expertise. 5,6
Earlier this year, I had the privilege of attending and presenting at the
New Zealand Dermatological Society Incorporated annual confer-
ence. Two of the scheduled sessions addressed the topic of machine
learning and AI in melanoma diagnosis. These talks, which heavily
featured data from the two studies described above, stimulated heat-
ed discussion among dermatologists. It soon became obvious that, at
present, even local experts may not necessarily be able to reach an
agreement; some were sceptical about the technical abilities of AI
or expressed concerns regarding patient satisfaction, whereas others
warmly welcomed the idea of incorporating CNN systems into their
practice, provided it reliably results in fewer missed cancerous lesions
and misdiagnosed benign ones.
While, understandably, there is a considerable amount of excitement
surrounding CNN, the tangible benefits of the demonstrated accura-
cy and efficiency of this technology may still be distant. 4–6 This is be-
cause initial CNN training requires a substantial amount of resources
and time, and actual implementation into routine clinical practice is
only possible once local medico-legal boundaries are better defined,
and security risks are addressed. 4–6
Other technology for the most visual specialty
It has been postulated that because of the highly visual nature of
diagnosis and management of skin conditions, modern technology
constitutes an especially valuable addition to dermatology – perhaps,
even more so than any other medical specialty. 7,8 The number of now
routine dermatological practices that heavily rely on machines of var-
ious kinds (not necessarily AI based) for visual assessment of skin
disease in one form or another is vast; among them are whole-body
photography, dermatoscopy, and teledermatology.
Given the recent surge of interest in healthcare-related technology, it
is not surprising that personal electronic devices are being increasing-
ly utilised by healthcare professionals. 7,8 Indeed, I have personally wit-
nessed numerous dermatologists regularly utilising their mobile phones
in their everyday practice, whether to quickly access reputable refer-
ence sources (such as DermNet NZ), or to use convenient smart-
phone dermoscope attachments, which are becoming increasingly
popular. Not to mention conventional dermoscopy, which can be con-
sidered the gold standard of clinical dermatologic assessment today. 7,8
With the rise of telemedicine, mobile devices and computers are now
becoming increasingly important for patients with skin problems, too,
especially those who may struggle to access in-person dermatology
advice (for example, individuals from rural/remote or low socioeco-
nomic status communities). 9,10 While both the store-and-forward and
live interactive forms of teledermatology have limitations (such as se-
curity issues or inability to incorporate palpation, a core component
of skin examination), research suggests that, overall, teledermatology
is a promising way of efficiently delivering quality dermatological care
at a lower cost compared to face-to-face visits. 8–11
Smartphone applications
Over the last couple of years, countless smartphone applications and
internet websites that aim to educate, diagnose, or even help manage
various health conditions have become available to both the general
public and the physician community. 12,13 Among the more popular are
apps designed specifically to help consumers detect malignant skin le-
sions, especially melanomas, at home. 14–18 Some of these are designed
to be more of a triage tool, whereas others virtually aim to replace
a dermatologist’s consult; most have ambiguous legal/regulatory sta-
tus. 14–18 Because of heterogeneity in the software employed in such
apps and in their purpose, the diagnostic accuracy, and thus, practical
utility of this class of apps as a whole is difficult to evaluate. 14–18 Accord-
ing to a large 2018 systematic review conducted by Rat et al, automated
smartphone medical apps aimed at melanoma diagnosis are currently
considered to be unreliable from accuracy and safety standpoints. 18 Is-
sues commonly reported in the literature include unacceptable rates of
false positive results, which could result in unwarranted patient anxiety
and increase in demand for unnecessary specialist care, as well as high
false negative rates and thus missed opportunities for timely identifi-
cation and treatment of potentially dangerous skin lesions due to false
reassurance. 14–18 The latter especially raises the complex issue of med-
ico-legal liability. 18 Regardless, these tools remain a popular conversa-
tion topic among patients: during my time as a student attached to
Dermatology and General Practice clinics, discussions around “self-as-
sessment” skin-check apps were a near everyday occurrence.
Way forward
Clearly, considerable efforts to improve melanoma-detecting apps
are required before they can become appropriate and widely ac-
cepted alternatives for proper clinical skin specialist consultations. 14–18
However, with the impressive results achieved by Esteva and
Haennsle using deep neural networks in mind, it is not unreason-
able to infer that if similar CNN technology could be competently
trained and incorporated into a user-friendly phone application, it
would represent a major step forward for skin cancer-detecting apps
from a diagnostic accuracy standpoint. 4–6,18 It would also be inter-
esting to observe the future interplay of the fields of whole-body
photography, mobile dermoscopy, teledermatology, and modern AI.
A successful fusion of these technologies could facilitate the diagnostic
process even further and benefit everyone involved in the detection
and treatment of skin cancer, from patients to experts. 5,6,8,14 Uncer-
tainty regarding dermatological diagnoses is prevalent among primary
care practitioners: despite dedicating large amounts of clinical time to
patients with skin complaints, many general practitioners lack formal
dermatological training and/or expertise. 19 Thus, the advent of such
CNN-based tools for the purposes of decision-making support could
be very helpful in the community setting as it could improve system
efficiency and reduce the burden of unnecessary referrals to spe-
cialists. 5,19 Dermatologists that currently look after high-risk patients
would also benefit from AI-based apps due to a streamlined, tar-
geted surveillance process, while patients themselves may enjoy the
enhanced convenience and reliability of self-skin checks. 5,12,13
Touch and empathy versus technology
Finally, I wanted to touch on some of the more philosophical aspects
of the interplay between technology and doctor-patient relationships
by reflecting on my own recent experience. I was a patient evaluat-
ed and treated for a pigmented skin lesion suspicious for malignancy.
Without going into too much detail, it was a drawn-out, stressful affair
comprised of long periods of waiting and uncertainty, multiple refer-
rals, appointments, and, finally, surgery. This process could probably
be vastly simplified, had the timely utilisation of technology such as
CNN been possible. However, despite being inefficient and frustrat-
ing at times, the overall experience ended up being memorable in a
good way because of the wonderful advice, respect, empathy, and
reassurance offered by the doctors I encountered on my journey as a
patient. At the time of writing this, I still do not know the result of the
biopsy, but I do know that, no matter the histological outcome, per-
sonally I would not have traded the excellent in-person care I received
for a quicker, definitive diagnosis made by a computer algorithm.
Upon reviewing relevant literature, I discovered that similar senti-
ments (i.e. valuing treatment with compassion, respect, and dignity
over efficiency or technical skills in the healthcare setting) are not un-
common among patients. 20–22 Indeed, the positive influence of warm,
patient-centric communication and of the act of physical examination
on the doctor-patient relationship is a well-documented theme in
medical and social sciences literature. 23 Despite claims that tradition-
ally-taught “doctoring” and interpersonal skills are losing importance
in the age of modern medicine characterised by staggering techno-
logical advance, or that the imperfect art of clinical examination is
slowly become obsolete, evidence suggests that patient-centred care
(which relies heavily on thoughtful utilisation of these long-established
modalities) still appears to be the key to patient satisfaction. 21–24 It
has even been postulated that biomedical developments may actually
widen both the psychological and physical distance between doctors
and patients, although further research is needed to explore this ef-
fect. 24 While computer-aided diagnostic systems are abundant and
have unique, undisputable advantages, 25 they obviously cannot (yet)
incorporate empathy and physical touch as powerful ways of con-
necting with and healing the patients with skin conditions. 5
Conclusion
In 2018, both clinicians and patients are equipped with a variety of
technological tools that may aid them in the diagnosis of skin condi-
tions. These range from the popular self-assessment mobile applica-
tions, to the more formal use of personal electronic devices for the
purposes of communicating with a specialist (as in teledermatology).
While CNN-trained AI has recently shown some truly impressive
abilities in detection of skin cancers, by no means does this represent
a replacement for all aspects of traditional physician consultations
such as thorough history taking, physical examination, human touch,
and empathy. After all, good medical practice is about much more
than solely diagnostic accuracy. Furthermore, heavy reliance of med-
ical practices on any technology inevitably brings with it a unique set
of concerns (including legality and cybersecurity issues) that must be
adequately addressed before widespread implementation is possible.
To conclude, despite significant technological advances of diagnostic
techniques in recent years, I do not believe that machines will replace
dermatologists in the diagnosis of skin disease (including, but not limit-
ed to malignant melanoma) any time soon. In my opinion, the empha-
sis should be on using the ever-evolving technology to complement
and augment the conventional skill set of physicians, rather than to re-
place doctors altogether. Such symbiosis would ideally help us achieve
enhanced rates of access to high-quality, appropriate dermatological
care and improved outcomes for patients with melanoma and other
skin disease in the most efficient and economical way possible.
References
1. Whiteman DC, Green AC, Olsen CM. Growing burden of invasive
melanoma: projections of incidence rates and numbers of new
cases in six susceptible populations through 2031. J Invest Dermatol.
2016;136:1161–71.
2. Anwar SM, Majid M, Qayyum A, Awais M, Alnowami M, Khan MK.
Medical image analysis using convolutional neural networks: a review. J
Med Syst.2018 Nov 1;42(11):226.
3. Rawat W, Wang Z. Deep convolutional neural networks for
image classification: a comprehensive review. Neural Comput. 2017
Sep;29(9):2352–449.
4. Esteva A, Kuprel B, Novoa RA, Ko J, Swetter SM, Blau HM, et al.
Dermatologist-level classification of skin cancer with deep neural
networks. Nature. 2017 Feb;542(7639):115.
5. Haenssle HA, Fink C, Schneiderbauer R, Toberer F, Buhl T, Blum
A, et al. Man against machine: diagnostic performance of a deep
learning convolutional neural network for dermoscopic melanoma
recognition in comparison to 58 dermatologists. Ann Oncol. 2018
May 28;29(8):1836–42.
6. Mar VJ, Soyer HP. Artificial intelligence for melanoma diagnosis:
how can we deliver on the promise? Ann Oncol. 2018
Aug;29(8):1625–1628.
7. Anyanwu CO, Lipoff JB. Smartphones, photography, and security in
dermatology. J Am Acad Dermatol. 2015 Jan 1;72(1):193–5.
8. Börve A, Terstappen K, Sandberg C, Paoli J. Mobile
teledermoscopy—there’s an app for that! Dermatol Pract Concept.
2013 Apr;3(2):41.
9. Bashshur RL, Shannon GW, Tejasvi T, Kvedar JC, Gates M. The
empirical foundations of teledermatology: a review of the research
evidence. Telemed J E Health. 2015 Dec 1;21(12):953–79.
10. Tensen E, Van Der Heijden JP, Jaspers MW, Witkamp L. Two
decades of teledermatology: current status and integration in national
healthcare systems. Curr Dermatol Rep. 2016 Jun 1;5(2):96–104.
11. Cox NH. Palpation of the skin—an important issue. J Roy Soc Med.
2006 Dec;99(12):598–600.
12. Brewer AC, Endly DC, Henley J, Amir M, Sampson BP, Moreau JF,
et al. Mobile applications in dermatology. JAMA Dermatol. 2013 Nov
1;149(11):1300–4.
13. Wolf JA, Moreau JF, Patton TJ, Winger DG, Ferris LK. Prevalence
and impact of health-related internet and smartphone use among
dermatology patients. Cutis. 2015 Jun;95(6):323–8.
14. Wang JV. Chapman LW, Keller, MS. Challenges to smartphone
applications for melanoma detection. Dermatol Online J. 2017
Feb 15;23(2).
15. Wolf JA, Moreau JF, Akilov O, Patton T, English JC, Ho J, et al.
Diagnostic inaccuracy of smartphone applications for melanoma
detection. JAMA Dermatol. 2013 Apr 1;149(4):422–6.
16. Kassianos AP, Emery JD, Murchie P, Walter FM. Smartphone
applications for melanoma detection by community, patient
and generalist clinician users: a review. Br J Dermatol. 2015
Jun;172(6):1507–18.
17. Chao E, Meenan CK, Ferris LK. Smartphone-based applications for
skin monitoring and melanoma detection. Dermatol Clin. 2017 Oct
1;35(4):551–7.
18. Rat C, Hild S, Sérandour JR, Gaultier A, Quereux G, Dreno B, et
al. Use of smartphones for early detection of melanoma: systematic
review. J Med Internet Res. 2018 Apr;20(4).
19. Kownacki S. Skin diseases in primary care: what should GPs be
doing? Br J Gen Pract. 2014 Aug;64(625):380–1.
20. Joffe S, Manocchia M, Weeks JC, Cleary PD. What do patients
value in their hospital care? An empirical perspective on autonomy
centred bioethics. J Med Ethics. 2003 Apr 1;29(2):103–8.
21. Epstein, RM, Street RL. The values and value of patient-centered
care. Ann Fam Med. 2011;9(2):100–3.
22. Miles A, Mezzich J. The care of the patient and the soul of the
clinic: person-centered medicine as an emergent model of modern
clinical practice. Int J Pers Cent Med. 2011 Jun 30;1(2):207–22.
23. Kugler J, Verghese A. The physical exam and other forms of fiction.
J Gen Intern Med. 2010 Aug 1;25(8):756–7.
24. Iida J, Nishigori H. Physical examination and the physician-patient
relationship: a literature review. MedEdPublish. 2016 Oct 12;5.
25. Masood A, Al-Jumaily AA. Computer aided diagnostic support
system for skin cancer: a review of techniques and algorithms. Int J
Biomed Imaging [Internet]. 2013 [cited 2018 Oct 28]. Available from:
https://www.hindawi.com/journals/ijbi/2013/323268/abs/.
Funding
Wilson-Allison Memorial Prize
(New Zealand Dermatological Society Incorporated)
Correspondence
Elizaveta (Lizzie) Rakhmanova: erak359@aucklanduni.ac.nz
Elizaveta (Lizzie) Rakhmanova
School of Medicine
Faculty of Medical and Health Sciences
University of Auckland
Introduction
With recent advances in artificial intelligence (AI), not a week goes by
without an article with a catchy headline stating that a certain med-
ical specialty will soon be replaced by “robots”. But are such claims
substantiated? In this essay, I hope to explore this fascinating topic by
firstly, reviewing recent literature on the role of AI technologies such
as deep learning convolutional neural networks (CNNs) in the diag-
nosis of dermatological disease. Secondly, I will outline some of the
existing technologies that aim to complement contemporary derma-
tologic practices. Some examples of this include teledermatology, mo-
bile dermoscopy/dermatoscopy, and smartphone apps. Finally, I will
briefly discuss patient-centred care as relevant to AI in dermatology.
Throughout the essay, I will draw on some relevant personal experi-
ences both as a student doctor and as a patient, to hopefully, provide
the reader with additional context from my perspective. Due to the
concerningly high rates, and thus, the public health importance of mel-
anoma in New Zealand, 1 as well as the breadth of the topic, for the
purposes of this entry I am going to focus on the use of AI and other
technological tools in detection of melanocytic cancer specifically. At
the same time, I will also acknowledge that AI and technology may be
successfully utilised to diagnose other types of skin disease, too.
Recent advances of AI and machine learning in dermatology
As alluded to in the introduction, of particular interest to derma-
tology is the concept of deep learning CNNs. CNNs are artificial,
feed-forward neural networks capable of analysing and learning from
visual imagery. 2,3 CNNs are able to improve their future performance
according to their previous experiences in image recognition and
classification – this process is referred to as machine learning. 2,3 The
concept of CNNs has become especially topical after the results of
a study by Esteva et al were published in Nature last year. 4 In this
landmark study (the largest of its kind), over 100,000 biopsy-backed
clinical photographs were used to teach a deep learning CNN-based
algorithm to discern malignant skin lesions from their benign mimick-
ers. 4 When asked to differentiate between, firstly, melanomas and
benign naevi, and, secondly, keratinocyte (i.e. non-melanocytic) carci-
nomas and seborrheic keratoses, the CNN system performed com-
parably to a cohort of 21 board-certified dermatologists. 4 For the first
time, successful utilisation of a computer algorithm capable of expert
level thinking was demonstrated for a relatively subjective task, which
is of increasing importance in everyday dermatological practice.
Less than a year later, Haenssle et al reported that a deep learning
CNN that was trained specifically to distinguish dermatoscopic imag-
es of benign and malignant melanocytic lesions has shown to be, on
average, superior in both sensitivity and specificity when compared to
an international panel of 51 dermatologists. 6 More than half of those
physicians were considered experts with five plus years of dermato-
scopic experience. 6 When additional clinical information was provid-
ed to the dermatologists (to simulate the real life setting more closely),
their overall sensitivity was improved, yet the algorithm still outper-
formed clinicians in terms of specificity. 5 Therefore, it was suggested
that a competently trained CNN may be a helpful addition to any der-
matologist’s diagnostic toolbox, regardless of their level of expertise. 5,6
Earlier this year, I had the privilege of attending and presenting at the
New Zealand Dermatological Society Incorporated annual confer-
ence. Two of the scheduled sessions addressed the topic of machine
learning and AI in melanoma diagnosis. These talks, which heavily
featured data from the two studies described above, stimulated heat-
ed discussion among dermatologists. It soon became obvious that, at
present, even local experts may not necessarily be able to reach an
agreement; some were sceptical about the technical abilities of AI
or expressed concerns regarding patient satisfaction, whereas others
warmly welcomed the idea of incorporating CNN systems into their
practice, provided it reliably results in fewer missed cancerous lesions
and misdiagnosed benign ones.
While, understandably, there is a considerable amount of excitement
surrounding CNN, the tangible benefits of the demonstrated accura-
cy and efficiency of this technology may still be distant. 4–6 This is be-
cause initial CNN training requires a substantial amount of resources
and time, and actual implementation into routine clinical practice is
only possible once local medico-legal boundaries are better defined,
and security risks are addressed. 4–6
Other technology for the most visual specialty
It has been postulated that because of the highly visual nature of
diagnosis and management of skin conditions, modern technology
constitutes an especially valuable addition to dermatology – perhaps,
even more so than any other medical specialty. 7,8 The number of now
routine dermatological practices that heavily rely on machines of var-
ious kinds (not necessarily AI based) for visual assessment of skin
disease in one form or another is vast; among them are whole-body
photography, dermatoscopy, and teledermatology.
Given the recent surge of interest in healthcare-related technology, it
is not surprising that personal electronic devices are being increasing-
ly utilised by healthcare professionals. 7,8 Indeed, I have personally wit-
nessed numerous dermatologists regularly utilising their mobile phones
in their everyday practice, whether to quickly access reputable refer-
ence sources (such as DermNet NZ), or to use convenient smart-
phone dermoscope attachments, which are becoming increasingly
popular. Not to mention conventional dermoscopy, which can be con-
sidered the gold standard of clinical dermatologic assessment today. 7,8
With the rise of telemedicine, mobile devices and computers are now
becoming increasingly important for patients with skin problems, too,
especially those who may struggle to access in-person dermatology
advice (for example, individuals from rural/remote or low socioeco-
nomic status communities). 9,10 While both the store-and-forward and
live interactive forms of teledermatology have limitations (such as se-
curity issues or inability to incorporate palpation, a core component
of skin examination), research suggests that, overall, teledermatology
is a promising way of efficiently delivering quality dermatological care
at a lower cost compared to face-to-face visits. 8–11
Smartphone applications
Over the last couple of years, countless smartphone applications and
internet websites that aim to educate, diagnose, or even help manage
various health conditions have become available to both the general
public and the physician community. 12,13 Among the more popular are
apps designed specifically to help consumers detect malignant skin le-
sions, especially melanomas, at home. 14–18 Some of these are designed
to be more of a triage tool, whereas others virtually aim to replace
a dermatologist’s consult; most have ambiguous legal/regulatory sta-
tus. 14–18 Because of heterogeneity in the software employed in such
apps and in their purpose, the diagnostic accuracy, and thus, practical
utility of this class of apps as a whole is difficult to evaluate. 14–18 Accord-
ing to a large 2018 systematic review conducted by Rat et al, automated
smartphone medical apps aimed at melanoma diagnosis are currently
considered to be unreliable from accuracy and safety standpoints. 18 Is-
sues commonly reported in the literature include unacceptable rates of
false positive results, which could result in unwarranted patient anxiety
and increase in demand for unnecessary specialist care, as well as high
false negative rates and thus missed opportunities for timely identifi-
cation and treatment of potentially dangerous skin lesions due to false
reassurance. 14–18 The latter especially raises the complex issue of med-
ico-legal liability. 18 Regardless, these tools remain a popular conversa-
tion topic among patients: during my time as a student attached to
Dermatology and General Practice clinics, discussions around “self-as-
sessment” skin-check apps were a near everyday occurrence.
Way forward
Clearly, considerable efforts to improve melanoma-detecting apps
are required before they can become appropriate and widely ac-
cepted alternatives for proper clinical skin specialist consultations. 14–18
However, with the impressive results achieved by Esteva and
Haennsle using deep neural networks in mind, it is not unreason-
able to infer that if similar CNN technology could be competently
trained and incorporated into a user-friendly phone application, it
would represent a major step forward for skin cancer-detecting apps
from a diagnostic accuracy standpoint. 4–6,18 It would also be inter-
esting to observe the future interplay of the fields of whole-body
photography, mobile dermoscopy, teledermatology, and modern AI.
A successful fusion of these technologies could facilitate the diagnostic
process even further and benefit everyone involved in the detection
and treatment of skin cancer, from patients to experts. 5,6,8,14 Uncer-
tainty regarding dermatological diagnoses is prevalent among primary
care practitioners: despite dedicating large amounts of clinical time to
patients with skin complaints, many general practitioners lack formal
dermatological training and/or expertise. 19 Thus, the advent of such
CNN-based tools for the purposes of decision-making support could
be very helpful in the community setting as it could improve system
efficiency and reduce the burden of unnecessary referrals to spe-
cialists. 5,19 Dermatologists that currently look after high-risk patients
would also benefit from AI-based apps due to a streamlined, tar-
geted surveillance process, while patients themselves may enjoy the
enhanced convenience and reliability of self-skin checks. 5,12,13
Touch and empathy versus technology
Finally, I wanted to touch on some of the more philosophical aspects
of the interplay between technology and doctor-patient relationships
by reflecting on my own recent experience. I was a patient evaluat-
ed and treated for a pigmented skin lesion suspicious for malignancy.
Without going into too much detail, it was a drawn-out, stressful affair
comprised of long periods of waiting and uncertainty, multiple refer-
rals, appointments, and, finally, surgery. This process could probably
be vastly simplified, had the timely utilisation of technology such as
CNN been possible. However, despite being inefficient and frustrat-
ing at times, the overall experience ended up being memorable in a
good way because of the wonderful advice, respect, empathy, and
reassurance offered by the doctors I encountered on my journey as a
patient. At the time of writing this, I still do not know the result of the
biopsy, but I do know that, no matter the histological outcome, per-
sonally I would not have traded the excellent in-person care I received
for a quicker, definitive diagnosis made by a computer algorithm.
Upon reviewing relevant literature, I discovered that similar senti-
ments (i.e. valuing treatment with compassion, respect, and dignity
over efficiency or technical skills in the healthcare setting) are not un-
common among patients. 20–22 Indeed, the positive influence of warm,
patient-centric communication and of the act of physical examination
on the doctor-patient relationship is a well-documented theme in
medical and social sciences literature. 23 Despite claims that tradition-
ally-taught “doctoring” and interpersonal skills are losing importance
in the age of modern medicine characterised by staggering techno-
logical advance, or that the imperfect art of clinical examination is
slowly become obsolete, evidence suggests that patient-centred care
(which relies heavily on thoughtful utilisation of these long-established
modalities) still appears to be the key to patient satisfaction. 21–24 It
has even been postulated that biomedical developments may actually
widen both the psychological and physical distance between doctors
and patients, although further research is needed to explore this ef-
fect. 24 While computer-aided diagnostic systems are abundant and
have unique, undisputable advantages, 25 they obviously cannot (yet)
incorporate empathy and physical touch as powerful ways of con-
necting with and healing the patients with skin conditions. 5
Conclusion
In 2018, both clinicians and patients are equipped with a variety of
technological tools that may aid them in the diagnosis of skin condi-
tions. These range from the popular self-assessment mobile applica-
tions, to the more formal use of personal electronic devices for the
purposes of communicating with a specialist (as in teledermatology).
While CNN-trained AI has recently shown some truly impressive
abilities in detection of skin cancers, by no means does this represent
a replacement for all aspects of traditional physician consultations
such as thorough history taking, physical examination, human touch,
and empathy. After all, good medical practice is about much more
than solely diagnostic accuracy. Furthermore, heavy reliance of med-
ical practices on any technology inevitably brings with it a unique set
of concerns (including legality and cybersecurity issues) that must be
adequately addressed before widespread implementation is possible.
To conclude, despite significant technological advances of diagnostic
techniques in recent years, I do not believe that machines will replace
dermatologists in the diagnosis of skin disease (including, but not limit-
ed to malignant melanoma) any time soon. In my opinion, the empha-
sis should be on using the ever-evolving technology to complement
and augment the conventional skill set of physicians, rather than to re-
place doctors altogether. Such symbiosis would ideally help us achieve
enhanced rates of access to high-quality, appropriate dermatological
care and improved outcomes for patients with melanoma and other
skin disease in the most efficient and economical way possible.
References
1. Whiteman DC, Green AC, Olsen CM. Growing burden of invasive
melanoma: projections of incidence rates and numbers of new
cases in six susceptible populations through 2031. J Invest Dermatol.
2016;136:1161–71.
2. Anwar SM, Majid M, Qayyum A, Awais M, Alnowami M, Khan MK.
Medical image analysis using convolutional neural networks: a review. J
Med Syst.2018 Nov 1;42(11):226.
3. Rawat W, Wang Z. Deep convolutional neural networks for
image classification: a comprehensive review. Neural Comput. 2017
Sep;29(9):2352–449.
4. Esteva A, Kuprel B, Novoa RA, Ko J, Swetter SM, Blau HM, et al.
Dermatologist-level classification of skin cancer with deep neural
networks. Nature. 2017 Feb;542(7639):115.
5. Haenssle HA, Fink C, Schneiderbauer R, Toberer F, Buhl T, Blum
A, et al. Man against machine: diagnostic performance of a deep
learning convolutional neural network for dermoscopic melanoma
recognition in comparison to 58 dermatologists. Ann Oncol. 2018
May 28;29(8):1836–42.
6. Mar VJ, Soyer HP. Artificial intelligence for melanoma diagnosis:
how can we deliver on the promise? Ann Oncol. 2018
Aug;29(8):1625–1628.
7. Anyanwu CO, Lipoff JB. Smartphones, photography, and security in
dermatology. J Am Acad Dermatol. 2015 Jan 1;72(1):193–5.
8. Börve A, Terstappen K, Sandberg C, Paoli J. Mobile
teledermoscopy—there’s an app for that! Dermatol Pract Concept.
2013 Apr;3(2):41.
9. Bashshur RL, Shannon GW, Tejasvi T, Kvedar JC, Gates M. The
empirical foundations of teledermatology: a review of the research
evidence. Telemed J E Health. 2015 Dec 1;21(12):953–79.
10. Tensen E, Van Der Heijden JP, Jaspers MW, Witkamp L. Two
decades of teledermatology: current status and integration in national
healthcare systems. Curr Dermatol Rep. 2016 Jun 1;5(2):96–104.
11. Cox NH. Palpation of the skin—an important issue. J Roy Soc Med.
2006 Dec;99(12):598–600.
12. Brewer AC, Endly DC, Henley J, Amir M, Sampson BP, Moreau JF,
et al. Mobile applications in dermatology. JAMA Dermatol. 2013 Nov
1;149(11):1300–4.
13. Wolf JA, Moreau JF, Patton TJ, Winger DG, Ferris LK. Prevalence
and impact of health-related internet and smartphone use among
dermatology patients. Cutis. 2015 Jun;95(6):323–8.
14. Wang JV. Chapman LW, Keller, MS. Challenges to smartphone
applications for melanoma detection. Dermatol Online J. 2017
Feb 15;23(2).
15. Wolf JA, Moreau JF, Akilov O, Patton T, English JC, Ho J, et al.
Diagnostic inaccuracy of smartphone applications for melanoma
detection. JAMA Dermatol. 2013 Apr 1;149(4):422–6.
16. Kassianos AP, Emery JD, Murchie P, Walter FM. Smartphone
applications for melanoma detection by community, patient
and generalist clinician users: a review. Br J Dermatol. 2015
Jun;172(6):1507–18.
17. Chao E, Meenan CK, Ferris LK. Smartphone-based applications for
skin monitoring and melanoma detection. Dermatol Clin. 2017 Oct
1;35(4):551–7.
18. Rat C, Hild S, Sérandour JR, Gaultier A, Quereux G, Dreno B, et
al. Use of smartphones for early detection of melanoma: systematic
review. J Med Internet Res. 2018 Apr;20(4).
19. Kownacki S. Skin diseases in primary care: what should GPs be
doing? Br J Gen Pract. 2014 Aug;64(625):380–1.
20. Joffe S, Manocchia M, Weeks JC, Cleary PD. What do patients
value in their hospital care? An empirical perspective on autonomy
centred bioethics. J Med Ethics. 2003 Apr 1;29(2):103–8.
21. Epstein, RM, Street RL. The values and value of patient-centered
care. Ann Fam Med. 2011;9(2):100–3.
22. Miles A, Mezzich J. The care of the patient and the soul of the
clinic: person-centered medicine as an emergent model of modern
clinical practice. Int J Pers Cent Med. 2011 Jun 30;1(2):207–22.
23. Kugler J, Verghese A. The physical exam and other forms of fiction.
J Gen Intern Med. 2010 Aug 1;25(8):756–7.
24. Iida J, Nishigori H. Physical examination and the physician-patient
relationship: a literature review. MedEdPublish. 2016 Oct 12;5.
25. Masood A, Al-Jumaily AA. Computer aided diagnostic support
system for skin cancer: a review of techniques and algorithms. Int J
Biomed Imaging [Internet]. 2013 [cited 2018 Oct 28]. Available from:
https://www.hindawi.com/journals/ijbi/2013/323268/abs/.
Funding
Wilson-Allison Memorial Prize
(New Zealand Dermatological Society Incorporated)
Correspondence
Elizaveta (Lizzie) Rakhmanova: erak359@aucklanduni.ac.nz
