Clifford Omer Rosendahl

Prof. Clifford Omer Rosendahl

orcid

Research Focus & Metrics

Academic Career Length

18Years (2009-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
27 July 2023 in Nature Medicine
Verified
Journal Article
Open access
PDF file
Catarina BarataCatarina Barata
Catarina Barata
VRVeronica Rotemberg
Veronica Rotemberg
NCNoel C. F. Codella
Noel C. F. Codella
PTPhilipp Tschandl
Philipp Tschandl
CRChristoph Rinner
Christoph Rinner
HKHARALD KITTLER
HARALD KITTLER
We investigated whether human preferences hold the potential to improve diagnostic artificial intelligence (AI)-based decision support using skin cancer diagnosis as a use case. We utilized nonuniform rewards and penalties based on expert-generated tables, balancing the benefits and harms of various diagnostic errors, which were applied using reinforcement learning. Compared with supervised learning, the reinforcement learning model improved the sensitivity for melanoma from 61.4% to 79.5% (95% confidence interval (CI): 73.5–85.6%) and for basal cell carcinoma from 79.4% to 87.1% (95% CI: 80.3–93.9%). AI overconfidence was also reduced while simultaneously maintaining accuracy. Reinforcement learning increased the rate of correct diagnoses made by dermatologists by 12.0% (95% CI: 8.8–15.1%) and improved the rate of optimal management decisions from 57.4% to 65.3% (95% CI: 61.7–68.9%). We further demonstrated that the reward-adjusted reinforcement learning model and a threshold-based model outperformed naïve supervised learning in various clinical scenarios. Our findings suggest the potential for incorporating human preferences into image-based diagnostic algorithms.

24 March 2014 in Journal of the European Academy of Dermatology and Venereology
Verified
Journal Article
Closed access
PDF file
PTPhilipp Tschandl
Philipp Tschandl
CRClifford Omer Rosendahl
Clifford Omer Rosendahl
HKHARALD KITTLER
HARALD KITTLER

1 November 2012 in Journal of the American Academy of Dermatology
Verified
Journal Article
Closed access
CRClifford Omer Rosendahl
Clifford Omer Rosendahl
GWGail Williams
Gail Williams
DEDiann Eley
Diann Eley
TWTobias Wilson
Tobias Wilson
GCGreg Canning
Greg Canning
DWDavid Wilkinson
David Wilkinson

31 January 2014 in Dermatology Practical & Conceptual
Verified
Journal Article
Open access
CRClifford Omer Rosendahl
Clifford Omer Rosendahl
ACAlan Cameron
Alan Cameron
PTPhilipp Tschandl
Philipp Tschandl
ABAgata Bulinska
Agata Bulinska
HKHARALD KITTLER
HARALD KITTLER
IZIris Zalaudek
Iris Zalaudek
While there are several published comprehensive stepwise algorithmic methods for diagnosing pigmented skin malignancy, only limited material has been published for the stepwise assessment of non-pigmented lesions. We present a method based on pattern analysis, with a stepwise assessment, first, for ulceration, second, for white clues (defined as white lines, or in the case of a raised lesion any of the keratin clues: dermatoscopic white circles, dermatoscopic white structureless areas or surface keratin), and third, if no ulceration or white clues are present, proceed to vessel pattern analysis. This is a novel method, and apart from the assessment of white clues in raised lesions, it has not been formally tested. The priority of keratin clues in raised lesions over vessel pattern analysis has, however, been verified. It is conceded that this method is less specific than methods which have clues of pigmented structures, and accepting these limitations, Prediction without Pigment is a decision algorithm intended to guide the clinician in the decision as to whether to perform a biopsy rather than consistently leading to a specific diagnosis. Reaching a more specific diagnosis at the end of our flowchart can be achieved by weighing of clues both clinical and dermatoscopic, and that ability can be expected to improve with both knowledge and experience, but no diagnostic method, including this one, can be 100% sensitive in diagnosing malignancy, in particular, melanoma. Taking these limitations into account, any non-pigmented lesion, regardless of pattern analysis, which is raised and firm (nodular) and for which a confident, specific benign diagnosis cannot be made, should be excised to exclude the nodular variant of amelanotic melanoma.

1 July 2012
Verified
Journal Article
Closed access
CRClifford Omer Rosendahl
Clifford Omer Rosendahl
ACAlan Cameron
Alan Cameron
IMIan McColl
Ian McColl
DWDavid Wilkinson
David Wilkinson

1 March 2024 in APL Bioengineering
Verified
Journal Article
Open access
PDF file
Tina QiTina Qi
Tina Qi
KBKarl Bertling
Karl Bertling
JTJ. Torniainen
J. Torniainen
FKF. Kong
F. Kong
TGT. Gillespie
T. Gillespie
ARAleksandar D. Rakić
Aleksandar D. Rakić
Terahertz (THz) imaging has long held promise for skin cancer detection but has been hampered by the lack of practical technological implementation. In this article, we introduce a technique for discriminating several skin pathologies using a coherent THz confocal system based on a THz quantum cascade laser. High resolution in vivo THz images (with diffraction limited to the order of 100 μm) of several different lesion types were acquired and compared against one another using the amplitude and phase values. Our system successfully separated pathologies using a combination of phase and amplitude information and their respective surface textures. The large scan field (50 × 40 mm) of the system allows macroscopic visualization of several skin lesions in a single frame. Utilizing THz imaging for dermatological assessment of skin lesions offers substantial additional diagnostic value for clinicians. THz images contain information complementary to the information contained in the conventional digital images.