Assessment Of Diagnostic Accuracy Using A Digital Camera For Teledermatology

1
Assessment Of Diagnostic Accuracy Using A Digital
Camera For Teledermatology
2
Elizabeth Krupinski, PhDBen LeSueur,
BSLansing Ellsworth, MDNorman Levine, MDRonald
Hansen, MDNancy Silvis, MDPeter Sarantopoulos,
BSPamela Hite, MD, MBAJames Wurzel, BSRonald
Weinstein, MDAna Marie Lopez, MD, MPH
3
Patients Opinions

• 24 satisfied by care from non-dermatologist
• 89 satisfied by care from a dermatologist
• 6 believe a generalist can treat their skin
  disease
• 87 say access to dermatologist very important to
  their health care
• Owen SA, Maeyens E, Weary PE. Patients opinions
  regarding direct access to dermatologic specialty
  care. JAAD 199736250-256.

4
Goal

• The goal of this study was to compare
  diagnostic accuracy of a dermatologic diagnosis
  based on in-person examination compared to
  diagnosis with
• still photo images acquired
• with a digital camera and
• displayed on a CRT monitor.

5
Rationale

• Real-time video conferencing technologies may
  not be available or may be too costly for some
  rural sites. Store-forward technologies may be
  more appropriate, and should be tested before
  they are used clinically.

6
Subjects Readers

• 308 consecutive patients referred for specialty
  consultation by PCP or general dermatologist to
  the Dermatology Clinic at the University of
  Arizona Health Sciences Center
• 104 of these cases were ultimately biopsied
• 3 board-certified specialty dermatologists

7
Exams

• Each dermatologist examined approximately 1/3 of
  the patients in person at the AHSC clinic
• Rendered either a single diagnosis (75 of the
  cases, n 230) or 2 or 3 differential diagnoses
  (25 of the cases, n 78)
• Up to 5 photos of the lesion ROIs were taken with
  a digital camera by 4 medical students trained in
  the use of the camera
• Global and close-up shots

8
The Digital Camera

• Canon PowerShot600
• CCD image sensor
• 832 x 608 pixels
• 24-bit color resolution
• f/2.5 lens
• Built-in flash
• 150 kB file size

9
The Display

• Gateway 2000 computer
• Gateway CrystalScan color monitor 1024 x 768
• PhotoImpact Album v 3.0 display software
• Brief patient history in each case folder
• Randomized case presentation

10
(No Transcript)
11
(No Transcript)
12
(No Transcript)
13
Procedure

• Approximately 2 months after in-person exams,
  digital images were examined
• Same 3 dermatologists examined all 308 cases
• Single most likely diagnosis rendered
• Decision confidence very definite, definite,
  probable, possible
• Rate image sharpness color excellent, good,
  fair, poor
• Viewing time recorded

14
Analyses

• In-person diagnosis truth
• Correct match digital diagnosis matches one of
  the differential possibilities listed during
  in-person
• Incorrect mismatch digital diagnosis does not
  match any of the in-person differentials
• For biopsy analyses, biopsy truth

15
Types of Cases

• Clinical Diagnosis Number of Cases
• Malignant or Premalignant 91
• Benign Proliferations 74
• Eczema/Dermatitis 36
• Pigmented Lesions 32
• Infections/Infestations 20
• Papulosquamous Disorders 12
• Urticarial Allergic 5
• Collagen/Vascular 1
• Miscellaneous 37

16
Biopsied Lesions

• Lesion Type Number
  Percent
• Infection/Infestation 3 3
• Pigmented Lesions 26 25
• Malignant/Premalignant 49 47
• Dermatitis/Eczema 4 4
• Benign Proliferations 12 11
• Miscellaneous 10 10

17
Diagnostic Accuracy

• F 0.011, df 2, p 0.989
• no differences in accuracy between
  dermatologists

Reader
18
Decision Confidence
Reader
19
Observer Variation

• Intra-Observer Variation
• Reader 1 90 agreement (n 104 cases)
• Reader 2 85 agreement (n 102 cases)
• Reader 3 76 agreement (n 102 cases)
• Inter-Observer Variation
• Reader 1 vs 2 kappa 0.82
• Reader 1 vs 3 kappa 0.81
• Reader 2 vs 3 kappa 0.80

20
Correct Decisions Biopsy Cases

• In-Person Photo vs Photo vs
• Reader vs Biopsy Biopsy In-Person
• 1 80 78 87
• 2 97 76 79
• 3 90 73 85
• Mean 89 76 84

21
Biopsy vs In-Person Mismatches
22
Image Quality
Reader
23
Correlations

• Color sharpness r 0.73
• Color Decision Confidence r 0.48
• Sharpness Decision Confidence r 0.47
• Decision confidence was not affected
  significantly by overall quality of the images

24
Viewing Time
Reader
25
Correlations

• View time vs color rating r 0.35
• View time vs sharpness rating r 0.24
• View time vs accuracy r 0.21
• View time vs confidence r 0.54
• View time vs single/multiple diagnoses r 0.15

26
Conclusions

• Digital photography can be used to acquire
  dermatology images for telediagnosis
• Image quality is overall excellent to good
• Diagnostic accuracy and confidence are high
• Digital image diagnoses compare favorably with
  in-person diagnoses and with biopsy results