Background: Previous research studies have demonstrated that medical content image retrieval can play an important role by assisting dermatologists in skin lesion diagnosis. However, current state-of-the-art approaches have not been adopted in routine consultation, partly due to the lack of interpretability limiting trust by clinical users.
Objective: This study developed a new image retrieval architecture for polarized or dermoscopic imaging guided by interpretable saliency maps. This approach provides better feature extraction, leading to better quantitative retrieval performance as well as providing interpretability for an eventual real-world implementation.
Methods: Content-based image retrieval (CBIR) algorithms rely on the comparison of image features embedded by convolutional neural network (CNN) against a labeled data set. Saliency maps are computer vision–interpretable methods that highlight the most relevant regions for the prediction made by a neural network. By introducing a fine-tuning stage that includes saliency maps to guide feature extraction, the accuracy of image retrieval is optimized. We refer to this approach as saliency-enhanced CBIR (SE-CBIR). A reader study was designed at the University Hospital Zurich Dermatology Clinic to evaluate SE-CBIR’s retrieval accuracy as well as the impact of the participant’s confidence on the diagnosis.
Results: SE-CBIR improved the retrieval accuracy by 7% (77% vs 84%) when doing single-lesion retrieval against traditional CBIR. The reader study showed an overall increase in classification accuracy of 22% (62% vs 84%) when the participant is provided with SE-CBIR retrieved images. In addition, the overall confidence in the lesion’s diagnosis increased by 24%. Finally, the use of SE-CBIR as a support tool helped the participants reduce the number of nonmelanoma lesions previously diagnosed as melanoma (overdiagnosis) by 53%.
Conclusions: SE-CBIR presents better retrieval accuracy compared to traditional CBIR CNN-based approaches. Furthermore, we have shown how these support tools can help dermatologists and residents improve diagnosis accuracy and confidence. Additionally, by introducing interpretable methods, we should expect increased acceptance and use of these tools in routine consultation.
Melanoma is one of the top-5 most common cancers in Switzerland with a standardized incidence ratio per 100,000 inhabitants of 29.8 for men and 24.7 for women . Longitudinal data acquired since 1989 show a linear 100% increase of the standardized incidence ratio for men [ ] in the last 30 years. Unfortunately, this is a worldwide trend. According to Arnold et al [ ], melanoma incidence and deaths are expected to increase 50% and 68%, respectively, by 2024. However, it is known that if diagnosed on time, skin cancer can be cured with a simple surgical procedure, dramatically increasing the survival rate. The rapid increase in cases every year has not been followed by an increase in the available number of dermatologists. This causes the system to operate inefficiently with increasing waiting times to get access to specialist consultation. To cope with such a situation, a study [ ] proposed to provide specific training in skin cancer diagnosis to general practitioners to improve their competence in such cases. The outcome of that study showed a positive impact during a limited period after the initial training, but eventually, the accuracy dropped again after one year. Thus, it is necessary to explore long-term solutions that can support dermatologists to face this pandemic in a reliable way.
The use of deep learning has raised substantial interest in dermatology. Seminal studies such as Esteva et al  showed how deep learning algorithms can outperform board-certified dermatologists in certain dermoscopic image triage tasks. More recent studies broadened the scope by developing algorithms for automatic screening and ugly-duckling characterization in wide surface images [ ]. Despite such encouraging applications and results, the transition from academic research studies to real-world application is only slowly being addressed. Different surveys showed a favorable position from dermatologists [ ] as well as patients [ ] with respect to the introduction of artificial intelligence (AI) in routine consultations. However, interpretability [ ] and the need of a specialist to supervise the outcome are issues that concern clinical users and patients. When exploring the option of implementing such support tools in real-world consultation, there are 2 main aspects to monitor: on the one hand, diagnosis accuracy and, in the other, diagnosis confidence, since both play a key role. For example, the overdiagnosis of benign lesions as malignant lesions and diagnoses with low confidence cause unnecessary surgeries. A frequent scenario in clinical practice is the surgical removal of atypical-looking benign lesions. Even in the hands of experts, there are 5 benign lesions removed for every 1 melanoma [ ], and in the hands of nonexperts, this increases exponentially. Another important fact that needs to be carefully considered is the bias that can be introduced in the user’s decision-making. However, one of the main limitations in the adoption of these tools for nonexpert users comes from the interpretability of such tools, which does not transmit confidence in the predictions even if the demonstrated accuracy is high.
Use of Image Retrieval in Dermatology
Content-based image retrieval (CBIR) is a powerful tool in medical practice that proposes similar cases to the ones under study, thus mimicking an automatized bibliography search. Dermatology is not an exception, since large data sets of images are traditionally available. Early applications of CBIR in skin lesion categorization relied on the comparison retrieved from text and annotations  and progressed to consider colors and shapes [ , ]. However, with the introduction of convolutional neural networks (CNNs), CBIR is now using them as the backbone for feature extraction [ - ]. A pilot study [ ] concluded that CBIR was perceived as easy to use and engaging; however, trust still has to be gained for routine use. Furthermore, in another high-impact study [ ] where the impact of using CBIR as a support tool was evaluated, one of the conclusions was that users tend to prefer other AI approaches such per-class probability in the long term. The authors suggested exploring other CBIR architectures that might overcome this issue. One possible reason is the lack of interpretability of the retrieval process. To overcome this limitation, recent studies [ - ] evaluated the option of introducing interpretability methods such as saliency maps to guide the CNN feature extraction. These newly proposed architectures were tested using radiography data sets, which present substantial differences with dermoscopic imaging. For this reason, in this study, we proposed an updated version of the algorithm by Silva et al [ ] for the specific needs of skin lesion diagnosis, which we refer to as saliency-enhanced CBIR (SE-CBIR).
The benefits of SE-CBIR in routine dermatologic consultations are as follows:
- Improved interpretability: Saliency maps provide a visual representation of the regions of interest considered by the neural network during feature extraction.
- Improved retrieval performance: By guiding feature extraction toward the regions of interests, we reduce the noise and nonrelevant information in the retrieval process.
- More efficient real-world implementation: By improving interpretability, trust by nonexperts should increase, supporting the wider adoption and implementation of such tools.
Data Set Description
For the training, validation, and testing of our algorithm, the HAM10000 data set [, ] was used. It consists of a total of 10,015 labeled dermoscopic images belonging to 7 different categories of pigmented lesions. The ground truth was determined in more than 50% of images by histopathology, and the other half were either confirmed by follow-up examinations, expert consensus, or in vivo confocal microscopy [ ]. Follow-up images of lesions were removed during the retrieval process to maximize the diversity. The HAM10000 data set designations for the 7 different classes were kept. The description for each category is as follows: actinic keratosis, squamous cell carcinoma, and Bowen disease (akiec); basal cell carcinoma (bcc); benign keratosis (bkl; solar lentigo, seborrheic keratosis, and lichen planus-like keratosis); dermatofibroma (df); melanoma (mel); melanocytic nevi (nv); and vascular lesions (vasc; angiomas, angiokeratomas, pyogenic granulomas, and hemorrhage). The data set is highly unbalanced, with 67% of the total number of lesions corresponding to melanocytic nevi and 11% to melanoma. To counteract such imbalance, a multiclass focal loss function was used during training.
In this study, we used a public data set, namely HAM10000 [, ]. This data set is routinely used by research studies and is available under a Creative Commons Attribution-NonCommercial 4.0 International Public License. Therefore, the use of this data set is not subject to ethics board approval.
CBIR algorithms are designed to retrieve images from a data set that are related to the image under study. Their performance is boosted by introducing CNNs for feature extraction and representation. We chose EfficientNetB4  as the backbone for our classifier. The choice was driven by the fact that ensemble models combining different versions of EfficientNet were used by the top teams in International Skin Imaging Collaboration (ISIC) competitions [ ]. We profited from transfer learning by initializing the weights using “ImageNet” default values. The EfficientNetB4 top layer was removed and replaced by average pooling, batch normalization, and 2 combinations of a dropout and a dense layer. The output layer features a softmax activation function. We extracted the deep features just after the last convolutional layer and before the classification layers.
The traditional CBIR approach, as depicted in, includes 3 steps: step 1, a CNN classifier using the HAM10000 data set is trained; step 4, extracted features from the query image are compared to those of each image in the retrieval set; and step 5, images are ranked according to similarity.
In clinical imaging, the relevant information is usually spatially constrained, so new approaches such as that in Silva et al  propose to add a fine-tuning stage, guiding the training using saliency maps to enhance the retrieval performance. With the emerging field of explainable AI, a variety of saliency methods have been developed, such as vanilla gradient [ ], SmoothGrad [ ], and integrated gradients [ ]. Saliency methods aim to create a map highlighting the pixels that are relevant for the network’s classification of a particular input image. We chose vanilla gradient for various reasons: it is model agnostic, simple to integrate, interpretable, and consistent. In Silva et al [ ], the superiority of latent representations derived from saliency maps over those directly generated from the input x-ray images was demonstrated, where latent representations of saliency maps were used for medical image retrieval purposes and in sample selection for active learning.
Due to the substantial differences between the x-ray images and our dermoscopic data set, we designed a second fine-tune stage that input the original image (3 RGB channels) plus the saliency map computed in the first step. This was done to avoid excessive loss of information included in the original colored dermoscopic images in contrast with the approach in Silva et al . As shown in , the traditional CBIR approach is comprised of 3 steps: the training of a skin lesion classifier (step 1), followed by deep feature extraction for every single image (step 4), and finally image retrieval according to similarity score (step 5), whereas SE-CBIR comprises 5 steps, adding the computation of saliency maps (step 2) and the training of a 4-channel classifier (step 3) [ ]. It is worth it noting that in step 3, the input and first convolutional layer are expanded to a 4-channel input to input the saliency maps together with the original image.
Standard data augmentation techniques from the albumentations libraries  were applied during training to avoid overfitting. These transformations include geometric augmentations and noise, distortion, brightness and contrast, and color modifications. The data augmentation is applied on the fly during each epoch. All images were resized and randomly cropped to the expected EfficientNetB4 input resolution (380 × 380). Additionally, coarse dropout to enhance regularization was applied. During the training of SE-CBIR (fine-tuning stage), no data augmentation, other than random flips and rotations, was performed.
For training, validation, and test purposes, the data set was randomly stratified in an 80:10:10 split, ensuring the same class distribution in all subsets. During training, an adaptative learning rate (α) approach was chosen to allow different α values for the pretrained EfficientnetB4 layers and the added layers, aiming to adapt the parameters of the pretrained layers just slightly or not at all compared to the new layers. A learning rate schedule was set up, with an initial ramp-up during the first epochs that aims to keep the learned features. After the ramp-up, the learning rate decays exponentially. Exponential decay is a widely used learning rate scheduling method to improve convergence. The adaptability of this approach reduces the time required to train neural networks and makes a neural model scalable, as they can adapt to structure and input data at any point in time while being trained.
For step 1, the learning rate was set to a low value (α=5 × 10–5) to profit from prelearned features, whereas for the additional layers, the α was set to .01. This approach focuses the training on the classification layers. In this second step, we added an additional channel to the original input size, creating 4D images including the 3 RGB channels plus the 1D saliency maps. An adaptive learning rate approach similar to traditional CBIR was applied. However, the adapted layers were found to perform better at being trained with a 100-times larger (α=1 × 10–3) learning rate than that in the first step. The learning rate was kept low at α=1 × 10–5 for the pretrained layers to profit again from transfer learning, and the same number of epochs and loss function were used.
The multicategorical focal loss function  was chosen for training. This function adds a term to the cross-entropy loss, improving performance in imbalanced data sets such as HAM10000. This is achieved through down weighting, which reduces the influence of easy examples on the loss function, resulting in more attention to hard ones. With a softmax activation function in the last layer, the focal loss ℒfocal for each sample can be derived by:
The parameters αi and γ define the weights on this additional term, whereas y*i represents the softmax prediction value of an input.
Image Retrieval and Ranking Metrics
In CBIR and image classification-based models, high-level image visuals are represented in the form of feature vectors that consists of numerical values. Image retrieval is based on the comparison of features extracted from the images. These vector features are then compared among the different image features to search and rank the “closest” ones. We chose cosine similarity to compare the features’ latent representation due to its adequacy of handling high-dimensional vectors (1792 × 1 in our case). The mathematical representation is as follows:
where A and B are the feature vectors of the query image and each of the images in the labeled data set. The lower the value of Sc, the closer both images are in terms of features extracted.
To evaluate the retrieval performance and compare the traditional CBIR algorithm against SE-CBIR, we used the retrieval precision for k retrieved images (P@k) or cut-off value metric:
Similarly, for multiclass retrieval, we defined the average precision for k retrieved images (AP@k):
with M being the total number of classes.
Rater Recruitment and Reader Study Design
To evaluate the impact of this type of decision support, a reader study was designed within the University Hospital Zurich Dermatology Clinic. A total of 9 participants were recruited based on their willingness and availability to participate: 1 expert on pigmented skin lesions with more than 25 years of experience and 8 residents from the clinic with 1 to 5 years of experience. However, all of them had reasonable experience with melanoma diagnosis using dermoscopic imaging.
To evaluate the impact of the diagnosis support tool, the reader study was divided into 2 tasks. In task 1, volunteers were asked to provide a diagnosis on 100 randomly chosen dermoscopic images extracted from our HAM10000 test set. In parallel, they were also asked to provide their level of confidence in their diagnosis on a 5-point Likert scale . After a break period of 1 day, the participants underwent task 2. They were asked to rediagnose the same set of images while being supported with the 6 “closest” images proposed by the retrieval algorithm that are characterized by their associated ground-truth label, as shown in . Note that the saliency maps were not presented to the participants along with the retrievals; however, in a real-world implementation, they would be available. To minimize bias, the single images were rotated by 180° with respect to the original orientation. Additionally, each participant was presented with a different, randomly selected set of images for the evaluation. As in task 1, the volunteers were asked to provide a diagnosis and their confidence level. To facilitate participation, a web survey was created ad hoc for the evaluation process, allowing the users to examine the lesions on high-quality screens and providing flexibility on when and where to perform the evaluation.
An example of a saliency map calculated using the vanilla gradients is shown in(center). Additionally, an overlay of the original image and the saliency map is shown (right) for illustration purposes. To minimize the loss of information during the fine-tuning training step, the original image (3 RBG channels) and the corresponding saliency map (1 channel) are fed as a 4-channel input to the CNN.
Image Retrieval Results
Retrieval precision comparing the traditional CBIR algorithm versus SE-CBIR is summarized in. The saliency-enhanced approach improves both the per-class prediction (P@k) and the average (AP@k) for all values k of retrieved images. The improvement in AP@k increases with k, from 7% (k=1) to a maximum of 18% (k=9). For a single class, the largest difference was found for class df where the prediction accuracy was almost doubled. The only class that did not experience a significant improvement was akiec. Both classes df and akiec were underrepresented with only 1% and 2% of the total retrieval data set. In the case of class df, SE-CBIR identified additional features leading to a significant improvement, whereas for class akiec, it could be that the limited samples available for class akiec presented similar appearances as solar lentigo or seborrheic keratosis and larger statistics are needed. For all other classes, P@k was >0.8 for k=1.
|Method and retrieval image (k)||Per-class P@ka||AP@kb|
aP@k: retrieval precision for k retrieved images.
bAP@k: average precision for k retrieved images.
cakiec: actinic keratosis, squamous cell carcinoma, and Bowen disease.
dbcc: basal cell carcinoma.
ebkl: benign keratosis.
hnv: melanocytic nevi.
ivasc: vascular lesions.
jSE-CBIR: saliency-enhanced content-based image retrieval.
kCBIR: content-based image retrieval.
Reader Study Outcome
The outcome of the clinical evaluation by the 9 volunteers is summarized in. Diagnosis accuracy for the SE-CBIR algorithm was computed by majority voting among the k=6 retrieved images, which reached 89% on average for all evaluations. In task 1, participants recorded the lowest accuracy at 38%, with an average of 62.2%. The performance was not uniform with σ=12.7 points. As expected, the board-certified dermatologists performed at the same level as the algorithm. In task 2, a significant improvement in diagnosis accuracy was observed for all participants. The average accuracy increased by 22.7 points to 84.9%, which also helped to reduce diagnosis spread, bringing it down to σ=7.1 points. Since each participant was presented with a different image data set, to evaluate the consistency of their selection, the Cohen κ coefficient for each participant for tasks 1 and 2 was calculated, as well as the average among them. The results showed that agreement between raters improved with the support of AI, going from fair agreement (k_average=0.35) in task 1 to substantial agreement (k_average=0.66) in task 2. In addition, the 2 best-performing participants in task 1 were able to outperform the SE-CBIR majority-voting prediction when provided with retrieved images. It is worth noting that the HAM10000 data set is composed of hand-picked lesions whose diagnosis can be challenging without access to the patient’s context. Regarding the average diagnosis confidence, an improvement from 3.11 to 3.86 was found (+24%). With the support of the retrieved cases, the confidence in the correct diagnosis increased from 3.35 to 4.03. Indeed, most (6/9, 67%) of the participants expressed a confidence level above 4, both in absolute confidence and confidence for correct diagnoses, whereas in task 1, only 1 participant showed that level of confidence. On the other hand, the confidence in incorrectly predicted lesions increased much less (2.72 to 2.90).
shows the performance of the retrieval algorithm with an average accuracy of 89.2%. This result is substantially higher than the largest reader study using the ISIC18 data set published by Tschandl et al [ ]; however, it used a different test set for the evaluation.
A closer look at the predictions per class is shown inwith the help of confusion matrices. Each row represents the total number of instances of a given class, whereas each column represents the total number of instances predicted for each class. (right) shows a substantial improvement of per-class accuracy with the help of SE-CBIR–retrieved images.
|Participant||SE-CBIR, accuracy (%)||Task 1||Task 2|
|Accuracy (%)||Absolute confidence||Confidence for correct diagnosis||Confidence for incorrect diagnosis||Accuracy (%)||Absolute confidence||Confidence for correct diagnosis||Confidence for incorrect diagnosis|
|Total, mean (SD)||89.2 (2.3)||62.2 (12.7)||3.11 (0.6)||3.35 (0.5)||2.72 (0.6)||84.9 (7.1)||3.86 (0.51)||4.03 (0.5)||2.9 (0.7)|
Class nv represented almost 80% of the test data set, and the SE-CBIR results show that almost 25% of the diagnoses were reconsidered into the correct class. Similarly, for class bkl, incorrect diagnoses were corrected in task 2 improving the accuracy by 17%. For class mel, the accuracy improved to 79% for correct melanoma cases. In total, just 20 correct decisions of task 1 were changed to an incorrect diagnosis in task 2, whereas 224 misclassified lesions from task 1 were correctly classified in task 2. This is an important indication that the algorithm does not lead to overconfident misclassifications but rather improves the number of correct diagnoses and their confidence—although we should note that 12 of the incorrectly overturned diagnoses agreed with the algorithm’s majority voting. Those cases probably would require additional patient-context information for a better evaluation. From a skin management point of view, melanoma overdiagnosis is a conservative approach where a negative biopsy might be justified. Regarding melanoma (mel) diagnosis, in task 1, a total of 21 cases were underdiagnosed (33% of the 64 total mel cases), whereas in task 2, this value decreased to 14 (22%) out of 64 cases. Regarding overdiagnosis in task 1, a total of 171 nonmelanoma cases (where 144 were nv cases) were misclassified into mel, whereas in task 2, this value decreased to 81 cases (with 65 nv cases).
This study presents a novel algorithm for skin lesion diagnosis support based on the use of saliency maps for feature extraction guidance in contrast to state-of-the-art image retrieval (CBIR). We refer to this architecture as SE-CBIR due to the addition of a second fine-tuning stage, combining saliency maps and dermoscopic imaging.
It was shown that SE-CBIR improved retrieval precision in dermoscopic data sets compared to traditional CBIR by 7%. Clinical relevance was assessed by a reader study where the participants improved their overall diagnosis accuracy by +22%, as well their confidence level by +24%. Considering only melanomas, the study demonstrated that SE-CBIR helped to decrease overdiagnosis by 53%.
However, the study has limitations, such as the use of a single data set and the fragility of saliency maps. It is well-known that different methods for saliency map calculation might lead to different results. Another potential limitation identified in the evaluation process is the fact that the participants did not have access to the patient’s context, which penalizes the diagnosis accuracy in certain difficult cases. Future work should include increasing the number of participants, including different target groups such as general practitioners, nurses, or technicians, to evaluate the impact and usefulness of such tools in different scenarios, as well as addressing the imbalance of the classes such as akiec, where the current data set does not seem to be representative and is difficult to generalize. In this case, data augmentation techniques will not solve the issue and additional images of such a class should be sought and included.
In conclusion, we have demonstrated the superior quantitative performance of SE-CBIR in comparison to the state-of-the-art CBIR by introducing saliency maps for feature extraction. By introducing interpretable methods, we also expect to improve acceptance by users since they have access to human-understandable information to better comprehend the algorithm’s decision process.
This work was supported by the University Hospital Zurich Clinical Research Priority Program “Artificial Intelligence in Radio Oncology” and by a grant from the Swiss National Supercomputing Centre (CSCS) under project ID “sm42.”
The origin of the image data set is reported in the data set publication of HAM10000 .
The code used in the study is available upon request to the corresponding author for academic purposes.
M Gassner, JBG, ST-L, RPB, AA, NA, M Guckenberger, and MR contributed to the conception and design of the study. M Gassner developed the web-based tool for the reader study. M Gassner and JBG performed the analysis with the supervision of ST-L, RPB, and AA. RPB, FB, FF, SH, CP, PS, DS, RS, and FV participated in the reader study. M Gassner and JBG wrote the first draft of the manuscript. All authors contributed to manuscript revision, read, and approved the submitted version. Funding acquisition and project administration was carried out by M Guckenberger, NA, RPB, and ST-L.
Conflicts of Interest
- Spezifische krebskrankheiten. Article in German. Bundesamt für Statistik. URL: https://www.bfs.admin.ch/bfs/de/home/statistiken/gesundheit/gesundheitszustand/krankheiten/krebs/spezifische.html [accessed 2023-06-27]
- Arnold M, Singh D, Laversanne M, Vignat J, Vaccarella S, Meheus F, et al. Global burden of cutaneous melanoma in 2020 and projections to 2040. JAMA Dermatol 2022 May 01;158(5):495-503 [https://europepmc.org/abstract/MED/35353115] [CrossRef] [Medline]
- Badertscher N, Tandjung R, Senn O, Kofmehl R, Held U, Rosemann T, et al. A multifaceted intervention: no increase in general practitioners' competence to diagnose skin cancer (minSKIN) - randomized controlled trial. J Eur Acad Dermatol Venereol 2015 Aug;29(8):1493-1499 [CrossRef] [Medline]
- 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 02;542(7639):115-118 [https://europepmc.org/abstract/MED/28117445] [CrossRef] [Medline]
- Soenksen LR, Kassis T, Conover ST, Marti-Fuster B, Birkenfeld JS, Tucker-Schwartz J, et al. Using deep learning for dermatologist-level detection of suspicious pigmented skin lesions from wide-field images. Sci Transl Med 2021 Feb 17;13(581):eabb3652 [CrossRef] [Medline]
- Scheetz J, Rothschild P, McGuinness M, Hadoux X, Soyer HP, Janda M, et al. A survey of clinicians on the use of artificial intelligence in ophthalmology, dermatology, radiology and radiation oncology. Sci Rep 2021 Mar 04;11(1):5193 [https://doi.org/10.1038/s41598-021-84698-5] [CrossRef] [Medline]
- Jutzi TB, Krieghoff-Henning EI, Holland-Letz T, Utikal JS, Hauschild A, Schadendorf D, et al. Artificial intelligence in skin cancer diagnostics: the patients' perspective. Front Med (Lausanne) 2020 Jun 2;7:233 [https://europepmc.org/abstract/MED/32671078] [CrossRef] [Medline]
- Hauser K, Kurz A, Haggenmüller S, Maron RC, von Kalle C, Utikal JS, et al. Explainable artificial intelligence in skin cancer recognition: a systematic review. Eur J Cancer 2022 May;167:54-69 [https://linkinghub.elsevier.com/retrieve/pii/S0959-8049(22)00123-X] [CrossRef] [Medline]
- Kovalyshyn I, Dusza SW, Siamas K, Halpern AC, Argenziano G, Marghoob AA. The impact of physician screening on melanoma detection. Arch Dermatol 2011 Nov;147(11):1269-1275 [CrossRef] [Medline]
- Müller H, Michoux N, Bandon D, Geissbuhler A. A review of content-based image retrieval systems in medical applications-clinical benefits and future directions. Int J Med Inform 2004 Feb;73(1):1-23 [http://doi.org/10.1016/j.ijmedinf.2003.11.024)] [CrossRef] [Medline]
- Kakumanu P, Makrogiannis S, Bourbakis N. A survey of skin-color modeling and detection methods. Pattern Recognition 2007 Mar;40(3):1106-1122 [http://doi.org/10.1016/j.patcog.2006.06.010)] [CrossRef]
- Cheng Y, Swamisai R, Umbaugh SE, Moss RH, Stoecker WV, Teegala S, et al. Skin lesion classification using relative color features. Skin Res Technol 2008 Feb 19;14(1):53-64 [https://europepmc.org/abstract/MED/18211602] [CrossRef] [Medline]
- Ballerini L, Li X, Fisher RB, Rees J. A query-by-example content-based image retrieval system of non-melanoma skin lesions. 2010 Presented at: MCBR-CDS 2009: Medical Content-Based Retrieval for Clinical Decision Support; September 20, 2009; London, UK p. 31-38 [CrossRef]
- Tschandl P, Argenziano G, Razmara M, Yap J. Diagnostic accuracy of content-based dermatoscopic image retrieval with deep classification features. Br J Dermatol 2019 Jul;181(1):155-165 [https://europepmc.org/abstract/MED/30207594] [CrossRef] [Medline]
- Allegretti S, Bolelli F, Pollastri F, Longhitano SS, Pellacani G, Grana C. Supporting skin lesion diagnosis with content-based image retrieval. 2021 May 5 Presented at: 2020 25th International Conference on Pattern Recognition (ICPR); January 10-15, 2021; Milan, Italy p. 8053-8060 [CrossRef]
- Sadeghi M, Chilana P, Yap J, Tschandl P, Atkins. Using content-based image retrieval of dermoscopic images for interpretation and education: a pilot study. Skin Res Technol 2020 Jul;26(4):503-512 [https://doi.org/10.1111/srt.12822] [CrossRef] [Medline]
- Tschandl P, Rinner C, Apalla Z, Argenziano G, Codella N, Halpern A, et al. Human-computer collaboration for skin cancer recognition. Nat Med 2020 Aug;26(8):1229-1234 [CrossRef] [Medline]
- Silva W, Poellinger A, Cardoso JS, Reyes M. Interpretability-guided content-based medical image retrieval. 2020 Sep 29 Presented at: MICCAI 2020: Medical Image Computing and Computer Assisted Intervention – MICCAI 2020; October 4-8, 2020; Lima, Peru p. 305-314 [CrossRef]
- Silva W, Gonçalves T, Härmä K, Schröder E, Obmann VC, Barroso MC, et al. Computer-aided diagnosis through medical image retrieval in radiology. Sci Rep 2022 Dec 01;12(1):20732 [https://boris.unibe.ch/id/eprint/175420] [CrossRef] [Medline]
- Mahapatra D, Poellinger A, Reyes M. Interpretability-guided inductive bias for deep learning based medical image. Med Image Anal 2022 Oct;81:102551-108415 [https://doi.org/10.1016/j.media.2022.102551] [CrossRef] [Medline]
- Codella NCF, Gutman D, Celebi ME, Helba B, Marchetti MA, Dusza SW, et al. Skin lesion analysis toward melanoma detection: A challenge at the 2017 International symposium on biomedical imaging (ISBI), hosted by the international skin imaging collaboration (ISIC). 2018 May 24 Presented at: 2018 IEEE 15th International Symposium on Biomedical Imaging (ISBI 2018); April 4-7, 2018; Washington, DC p. 168-172 [CrossRef]
- Tschandl P, Rosendahl C, Kittler H. The HAM10000 dataset, a large collection of multi-source dermatoscopic images of common pigmented skin lesions. Sci Data 2018 Aug 14;5(1):180161 [https://doi.org/10.1038/sdata.2018.161] [CrossRef] [Medline]
- Tan M, Le Q. EfficientNet: rethinking model scaling for convolutional neural networks. 2019 Presented at: 36th International Conference on Machine Learning; June 9-15, 2019; Long Beach, CA p. 6105-6114 URL: http://proceedings.mlr.press/v97/tan19a.html
- ISIC 2019 Challenge. ISIC Challenge. URL: https://challenge2019.isic-archive.com/leaderboard.html [accessed 2023-06-27]
- Simonyan K, Vedaldi A, Zisserman A. Deep inside convolutional networks: visualising image classification models and saliency maps. 2014 Presented at: 2nd International Conference on Learning Representations, ICLR2014; April 14-16, 2014; Banff, AB URL: https://www.robots.ox.ac.uk/~vgg/publications/2014/Simonyan14a/
- Daniel S, Thorat N, Kim B, Viégas F, Wattenberg M. Smoothgrad: removing noise by adding noise. arXiv. Preprint posted online on Jun 12, 2017 [CrossRef]
- Sundararajan M, Taly A, Yan Q. Axiomatic attribution for deep networks. 2017 Presented at: 34th International Conference on Machine Learning; August 6-11, 2017; Sydney, Australia p. 3319-3328 URL: http://proceedings.mlr.press/v70/sundararajan17a.html
- Murabito F, Spampinato C, Palazzo S, Giordano D, Pogorelov K, Riegler M. Top-down saliency detection driven by visual classification. Computer Vision and Image Understanding 2018 Jul;172:67-76 [https://doi.org/10.1016/j.cviu.2018.03.005] [CrossRef]
- Buslaev A, Iglovikov VI, Khvedchenya E, Parinov A, Druzhinin M, Kalinin AA. Albumentations: fast and flexible image augmentations. Information 2020 Feb 24;11(2):125 [CrossRef]
- Lin TY, Goyal P, Girshick R, He K, Dollár P. Focal loss for dense object detection. 2017 Presented at: 2017 IEEE International Conference on Computer Vision (ICCV); October 22-29, 2017; Venice, Italy p. 2999-3007 [CrossRef]
- Preedy VR, Watson RR, editors. 5-point Likert scale. In: Handbook of Disease Burdens and Quality of Life Measures. New York, NY: Springer; 2010:4288
|AI: artificial intelligence|
|AP@k: average precision for k retrieved images|
|CBIR: content-based image retrieval|
|CNN: convolutional neural network|
|ISIC: International Skin Imaging Collaboration|
|P@k: retrieval precision for k retrieved images|
|SE-CBIR: saliency-enhanced content-based image retrieval|
Edited by R Dellavalle; submitted 24.08.22; peer-reviewed by B Qian, HW Chiu, J Lei, J Robinson; comments to author 20.10.22; revised version received 07.04.23; accepted 16.06.23; published 24.08.23Copyright
©Mathias Gassner, Javier Barranco Garcia, Stephanie Tanadini-Lang, Fabio Bertoldo, Fabienne Fröhlich, Matthias Guckenberger, Silvia Haueis, Christin Pelzer, Mauricio Reyes, Patrick Schmithausen, Dario Simic, Ramon Staeger, Fabio Verardi, Nicolaus Andratschke, Andreas Adelmann, Ralph P Braun. Originally published in JMIR Dermatology (http://derma.jmir.org), 24.08.2023.
This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR Dermatology, is properly cited. The complete bibliographic information, a link to the original publication on http://derma.jmir.org, as well as this copyright and license information must be included.