Abstract
I have read with interest the article entitled “Can We Trust the Use of Smartphone Cameras in Clinical Practice? Laypeople Assessment of Their Image Quality,” by Boissin et al. 1
The title of this article poses a crucial question: Can we trust the use of smartphone cameras in clinical practice? The question is crucial because nowadays the smartphone is the most easily reached image acquisition device, and there are indeed reports on their usage for telemedicine needs, as mentioned by the authors in the introduction of their article.
However, several points regarding the Material and Methods sections are unclear. The camera mentioned as a gold standard is not fully specified. The Mark II is not a Canon model, but the postfix for version 2 at present. Currently three different models (1D, 5D, 7D), with sensor resolution ranging from 8.2 to 21.1 megapixels and also different sensor size, are available. Although smartphones have resolution described, the camera does not. One sentence indicates that “The final images were used in their raw state, without digital processing or retouching.” In photographic terms, “raw” has a very specific meaning, which could be applied on the Canon camera only, because smartphones normally do not provide raw images but JPEG only (which are internally processed). I suppose they were all JPEG, although this is not written in the article. In this case, a mention on the JPEG parameters used in the camera is needed, since its generation depends on that. Also, the aperture used for taking the pictures should be declared, because it is a subjective parameter that influences depth of field and thus focus perception.
In my opinion, the rest of the title (“Laypeople Assessment of Their Image Quality”) does not provide the right answer to the question: although perceived image quality is important for pleasing the eye, visual satisfaction is not the aim of telemedicine. In fact, the two articles cited as justification for the experiment design are not demonstrating that assessment by laypeople is the most reliable tool to assess image quality: in one article, subjective assessment was used to compare different high dynamic range (HDR) algorithms 2 (thus, an experiment similar to this one), the other compares four subjective assessment methods. 3
Subjective assessment is a practical and adequate way for measuring user preferences, but in clinical practice the most important aim is to preserve clinical content, that is, most of times, diagnostic content of images. How the content is preserved is a manifold issue: resolution is needed if the clinical detail is very small, color precision is needed if clinical meaning is related to color, etc. The real resolution of a system can be measured: it is not matter of preference.
In fact, it depends very much on the image content: if a detail to be examined has a specific minimum size, there are constraints on sensor resolution and lens needed to capture it. This is the reason why, for example, telepathology guidelines recommend that “all laboratories implementing a telepathology service for clinical diagnostic purposes shall perform their own validation studies.” 4
However, reasoning on why perceived quality is better on lesser quality systems can be useful even from a telemedicine point of view. Since this point is not dealt in the discussion, I wish to provide some key points: 1. Focus: the very small size of smartphone sensors, and thus the short focal length of their objectives, makes the depth of field larger than that of the reference camera, thus almost everything in a picture is in focus (and in turn its perceived quality can be greater). Reflex cameras, with suitable objectives, allow for shallow depth of field (and this is a highly regarded feature). 2. Color: smartphone image processing engines are optimized for obtaining pleasant pictures (e.g., by saturating colors) and minimize some of the generic user errors. This is made easier by the fact that a smartphone has only a fixed, nonzooming lens, so the software can be optimized for that. On the other side, a professional camera is aimed at professionals, who normally do the postprocessing work by themselves, to obtain the most from the image instead of letting the camera (or smartphone) process it. 3. Resolution: in no way the resolution provided by a smartphone is more than the gold standard camera. Likely, the missing need for a deep examination of images (like when you look at a clinical image) flattened the perceived differences. 4. Compression: JPEG compression always produces artifacts, which could be more or less pronounced depending on the settings. At low compression, they can be irrelevant even for clinical purposes, depending on field and kind of images.
The former is the only technical limitation of a smartphone that can actually provide an advantage for clinical pictures, because it minimizes focusing errors. For a doctor taking a picture of a patient, not worrying about focus is clearly positive, provided that the obtained image is of good quality. On the other side, color rendition optimization might be an issue if it modifies the meaning of parts of the image, and no control on compression might make it difficult to abide to guidelines, when they ask for low or medium compression to preserve quality (e.g., in teledermatology). 5
This is not to exclude the usage of smartphones for clinical practice: it is only unlikely that results obtained on generic images can be applied to clinical images with no specific studies on them.
Editor's Note
The authors of the original article by Boissin C, Fleming J, et al. were invited to comment on these remarks. They declined to provide any follow-up.