Automated melanoma detection: multispectral imaging and neural network approach for classification

Non-invasive diagnostic techniques in pigmentary skin disorders and skin cancer

BACKGROUND

Diagnosis of pigmentary skin disorders, pre-cancerous and cancerous skin diseases is traditionally relied on visual assessment. The most widely applied invasive diagnostic technique is the skin biopsy. There have been significant technological advances in non-invasive diagnostic methods for skin disorders.

OBJECTIVE

The objective of this article is to discuss different non-invasive diagnostic modalities, used in the diagnosis of pigmentary skin disorders and cutaneous cancers.

METHODS

Comprehensive literature search was performed to screen articles related to non-invasive diagnostic techniques in pigmentary skin disorders and cutaneous cancers. Articles published in journals indexed in PubMed were searched along with those in Google Scholar. Clinical trials, review articles, case series, case reports and other relevant articles were considered for review. References of relevant articles were also considered for review.

RESULTS

Dermoscopy and ultrasonography were the only non-invasive diagnostic and imaging techniques available to dermatologists for many years. The advent of computed tomography (CT) and magnetic resonance imaging (MRI) augmented the visualization of deeper structures. Confocal laser microscopy (CLM) and reflectance spectrophotometers have showed promising results in the non-invasive detection of pigmented lesions. Optical coherence tomography (OCT), electrical impedance spectroscopy (EIS), multispectral imaging, high frequency ultrasonography (HFUS) and adhesive patch biopsy aid in the accurate diagnosis of benign, as well as neoplastic skin diseases.

CONCLUSION

There have been significant advancements in non-invasive methods for diagnosis of dermatological diseases. These techniques can be repeatedly used in a comfort manner for the patient, and may offer an objective way to follow the course of a disease.

Optical Technologies for the Improvement of Skin Cancer Diagnosis: A Review

The worldwide incidence of skin cancer has risen rapidly in the last decades, becoming one in three cancers nowadays. Currently, a person has a 4% chance of developing melanoma, the most aggressive form of skin cancer, which causes the greatest number of deaths. In the context of increasing incidence and mortality, skin cancer bears a heavy health and economic burden. Nevertheless, the 5-year survival rate for people with skin cancer significantly improves if the disease is detected and treated early. Accordingly, large research efforts have been devoted to achieve early detection and better understanding of the disease, with the aim of reversing the progressive trend of rising incidence and mortality, especially regarding melanoma. This paper reviews a variety of the optical modalities that have been used in the last years in order to improve non-invasive diagnosis of skin cancer, including confocal microscopy, multispectral imaging, three-dimensional topography, optical coherence tomography, polarimetry, self-mixing interferometry, and machine learning algorithms. The basics of each of these technologies together with the most relevant achievements obtained are described, as well as some of the obstacles still to be resolved and milestones to be met.

Artificial Intelligence Estimates the Importance of Baseline Factors in Predicting Response to Anti-PD1 in Metastatic Melanoma

OBJECTIVE

Prognosis of patients with metastatic melanoma has dramatically improved over recent years because of the advent of antibodies targeting programmed cell death protein-1 (PD1). However, the response rate is ~40% and baseline biomarkers for the outcome are yet to be identified. Here, we aimed to determine whether artificial intelligence might be useful in weighting the importance of baseline variables in predicting response to anti-PD1.

METHODS

This is a retrospective study evaluating 173 patients receiving anti-PD1 for melanoma. Using an artificial neuronal network analysis, the importance of different variables was estimated and used in predicting response rate and overall survival.

RESULTS

After a mean follow-up of 12.8 (±11.9) months, disease control rate was 51%. Using artificial neuronal network, we observed that 3 factors predicted response to anti-PD1: neutrophil-to-lymphocyte ratio (NLR) (importance: 0.195), presence of ≥3 metastatic sites (importance: 0.156), and baseline lactate dehydrogenase (LDH) > upper limit of normal (importance: 0.154). Looking at connections between different covariates and overall survival, the most important variables influencing survival were: presence of ≥3 metastatic sites (importance: 0.202), age (importance: 0.189), NLR (importance: 0.164), site of primary melanoma (cutaneous vs. noncutaneous) (importance: 0.112), and LDH > upper limit of normal (importance: 0.108).

CONCLUSIONS

NLR, presence of ≥3 metastatic sites, LDH levels, age, and site of primary melanoma are important baseline factors influencing response and survival. Further studies are warranted to estimate a model to drive the choice to administered anti-PD1 treatments in patients with melanoma.

Photoacoustic/Ultrasound/Optical Coherence Tomography Evaluation of Melanoma Lesion and Healthy Skin in a Swine Model

The marked increase in the incidence of melanoma coupled with the rapid drop in the survival rate after metastasis has promoted the investigation into improved diagnostic methods for melanoma. High-frequency ultrasound (US), optical coherence tomography (OCT), and photoacoustic imaging (PAI) are three potential modalities that can assist a dermatologist by providing extra information beyond dermoscopic features. In this study, we imaged a swine model with spontaneous melanoma using these modalities and compared the images with images of nearby healthy skin. Histology images were used for validation.

Computer-assisted diagnosis techniques (dermoscopy and spectroscopy-based) for diagnosing skin cancer in adults

BACKGROUND

Early accurate detection of all skin cancer types is essential to guide appropriate management and to improve morbidity and survival. Melanoma and cutaneous squamous cell carcinoma (cSCC) are high-risk skin cancers which have the potential to metastasise and ultimately lead to death, whereas basal cell carcinoma (BCC) is usually localised with potential to infiltrate and damage surrounding tissue. Anxiety around missing early curable cases needs to be balanced against inappropriate referral and unnecessary excision of benign lesions. Computer-assisted diagnosis (CAD) systems use artificial intelligence to analyse lesion data and arrive at a diagnosis of skin cancer. When used in unreferred settings ('primary care'), CAD may assist general practitioners (GPs) or other clinicians to more appropriately triage high-risk lesions to secondary care. Used alongside clinical and dermoscopic suspicion of malignancy, CAD may reduce unnecessary excisions without missing melanoma cases.

OBJECTIVES

To determine the accuracy of CAD systems for diagnosing cutaneous invasive melanoma and atypical intraepidermal melanocytic variants, BCC or cSCC in adults, and to compare its accuracy with that of dermoscopy.

SEARCH METHODS

We undertook a comprehensive search of the following databases from inception up to August 2016: Cochrane Central Register of Controlled Trials (CENTRAL); MEDLINE; Embase; CINAHL; CPCI; Zetoc; Science Citation Index; US National Institutes of Health Ongoing Trials Register; NIHR Clinical Research Network Portfolio Database; and the World Health Organization International Clinical Trials Registry Platform. We studied reference lists and published systematic review articles.

SELECTION CRITERIA

Studies of any design that evaluated CAD alone, or in comparison with dermoscopy, in adults with lesions suspicious for melanoma or BCC or cSCC, and compared with a reference standard of either histological confirmation or clinical follow-up.

DATA COLLECTION AND ANALYSIS

Two review authors independently extracted all data using a standardised data extraction and quality assessment form (based on QUADAS-2). We contacted authors of included studies where information related to the target condition or diagnostic threshold were missing. We estimated summary sensitivities and specificities separately by type of CAD system, using the bivariate hierarchical model. We compared CAD with dermoscopy using (a) all available CAD data (indirect comparisons), and (b) studies providing paired data for both tests (direct comparisons). We tested the contribution of human decision-making to the accuracy of CAD diagnoses in a sensitivity analysis by removing studies that gave CAD results to clinicians to guide diagnostic decision-making.

MAIN RESULTS

We included 42 studies, 24 evaluating digital dermoscopy-based CAD systems (Derm-CAD) in 23 study cohorts with 9602 lesions (1220 melanomas, at least 83 BCCs, 9 cSCCs), providing 32 datasets for Derm-CAD and seven for dermoscopy. Eighteen studies evaluated spectroscopy-based CAD (Spectro-CAD) in 16 study cohorts with 6336 lesions (934 melanomas, 163 BCC, 49 cSCCs), providing 32 datasets for Spectro-CAD and six for dermoscopy. These consisted of 15 studies using multispectral imaging (MSI), two studies using electrical impedance spectroscopy (EIS) and one study using diffuse-reflectance spectroscopy. Studies were incompletely reported and at unclear to high risk of bias across all domains. Included studies inadequately address the review question, due to an abundance of low-quality studies, poor reporting, and recruitment of highly selected groups of participants.Across all CAD systems, we found considerable variation in the hardware and software technologies used, the types of classification algorithm employed, methods used to train the algorithms, and which lesion morphological features were extracted and analysed across all CAD systems, and even between studies evaluating CAD systems. Meta-analysis found CAD systems had high sensitivity for correct identification of cutaneous invasive melanoma and atypical intraepidermal melanocytic variants in highly selected populations, but with low and very variable specificity, particularly for Spectro-CAD systems. Pooled data from 22 studies estimated the sensitivity of Derm-CAD for the detection of melanoma as 90.1% (95% confidence interval (CI) 84.0% to 94.0%) and specificity as 74.3% (95% CI 63.6% to 82.7%). Pooled data from eight studies estimated the sensitivity of multispectral imaging CAD (MSI-CAD) as 92.9% (95% CI 83.7% to 97.1%) and specificity as 43.6% (95% CI 24.8% to 64.5%). When applied to a hypothetical population of 1000 lesions at the mean observed melanoma prevalence of 20%, Derm-CAD would miss 20 melanomas and would lead to 206 false-positive results for melanoma. MSI-CAD would miss 14 melanomas and would lead to 451 false diagnoses for melanoma. Preliminary findings suggest CAD systems are at least as sensitive as assessment of dermoscopic images for the diagnosis of invasive melanoma and atypical intraepidermal melanocytic variants. We are unable to make summary statements about the use of CAD in unreferred populations, or its accuracy in detecting keratinocyte cancers, or its use in any setting as a diagnostic aid, because of the paucity of studies.

AUTHORS' CONCLUSIONS

In highly selected patient populations all CAD types demonstrate high sensitivity, and could prove useful as a back-up for specialist diagnosis to assist in minimising the risk of missing melanomas. However, the evidence base is currently too poor to understand whether CAD system outputs translate to different clinical decision-making in practice. Insufficient data are available on the use of CAD in community settings, or for the detection of keratinocyte cancers. The evidence base for individual systems is too limited to draw conclusions on which might be preferred for practice. Prospective comparative studies are required that evaluate the use of already evaluated CAD systems as diagnostic aids, by comparison to face-to-face dermoscopy, and in participant populations that are representative of those in which the test would be used in practice.

In Vivo Evaluation of Melanoma Thickness by Multispectral Imaging and An Artificial Neural Network. A Retrospective Study on 250 Cases of Cutaneous Melanoma

Aims and Background

Noninvasive diagnostic methods such as dermoscopy, sonography, palpation or combined approaches have been developed in an attempt to preoperatively estimate melanoma thickness. However, the clinical presentation is often complex and the evaluation subjective. Multispectral image analysis of melanomas allows selection of features related to the content and distribution of absorbers, mainly melanin and hemoglobin, present within the lesion. Hence, it is reasonable to assume that the same features might be useful to predict melanoma thickness.

Methods

A multispectral image system was used to analyze in vivo 1939 pigmented skin lesions. The lesion selection was based on clinical and/or dermoscopic features that supported a suspicion for melanoma. All the lesions were then subjected to surgery for the histopathological diagnosis, and 250 were melanomas. From the multispectral images of the melanomas, we selected 12 features, seven of which were used to train and test an artificial neural network on 155 and 95 melanomas, respectively.

Results

Sensitivity (i.e., melanoma ≥0.75 mm thick correctly classified) and specificity (i.e., melanoma <0.75 mm thick correctly classified) evaluated from the receiving operating characteristic curves ranged from 76 to 90% and from 91 to 74%, respectively.

Conclusions

Our approach provides results similar to those obtained with other methods and has the advantage that it is not related to the expertise of the clinician. In addition, the physical interpretation of the selected features suggests a possible role of spectrophotometry as an objective method to study the natural history of the early phases of the disease.

Comparison of instruments for investigation of microcirculatory blood flow and red blood cell concentration

The use of laser Doppler perfusion imaging (LDPI) and laser speckle perfusion imaging (LSPI) is well known in the noninvasive investigation of microcirculatory blood flow. This work compares the two techniques with the recently developed tissue viability (TiVi) imaging system, which is proposed as a useful tool to quantify red blood cell concentration in microcirculation. Three systems are evaluated with common skin tests such as the use of vasodilating and vasoconstricting drugs (methlynicotinate and clobetasol, respectively) and a reactive hyperaemia maneuver (using a sphygmomanometer). The devices investigated are the laser Doppler line scanner (LDLS), the laser speckle perfusion imager (FLPI)-both from Moor Instruments (Axminster, United Kingdom)-and the TiVi imaging system (WheelsBridge AB, Linkoping, Sweden). Both imaging and point scanning by the devices are used to quantify the provoked reactions. Perfusion images of vasodilatation and vasoconstriction are acquired with both LDLS and FLPI, while TiVi images are acquired with the TiVi imager. Time acquisitions of an averaged region of interest are acquired for temporal studies such as the reactive hyperaemia. In contrast to the change in perfusion over time with pressure, the TiVi imager shows a different response due its measurement of blood concentration rather than perfusion. The responses can be explained by physiological understanding. Although the three devices sample different compartments of tissue, and output essentially different variables, comparisons can be seen between the three systems. The LDLS system proves to be suited to measurement of perfusion in deeper vessels, while FLPI and TiVi showed sensitivity to more superficial nutritional supply. LDLS and FLPI are insensitive to the action of the vasoconstrictor, while TiVi shows the clear boundaries of the reaction. Assessment of the resolution, penetration depth, and acquisition rate of each instrument show complimentary features that should be taken into account when choosing a system for a particular clinical measurement.

Systematic review of dermoscopy and digital dermoscopy/ artificial intelligence for the diagnosis of melanoma

BACKGROUND

Dermoscopy improves diagnostic accuracy of the unaided eye for melanoma, and digital dermoscopy with artificial intelligence or computer diagnosis has also been shown useful for the diagnosis of melanoma. At present there is no clear evidence regarding the diagnostic accuracy of dermoscopy compared with artificial intelligence.

OBJECTIVES

To evaluate the diagnostic accuracy of dermoscopy and digital dermoscopy/artificial intelligence for melanoma diagnosis and to compare the diagnostic accuracy of the different dermoscopic algorithms with each other and with digital dermoscopy/artificial intelligence for the detection of melanoma.

METHODS

A literature search on dermoscopy and digital dermoscopy/artificial intelligence for melanoma diagnosis was performed using several databases. Titles and abstracts of the retrieved articles were screened using a literature evaluation form. A quality assessment form was developed to assess the quality of the included studies. Heterogeneity among the studies was assessed. Pooled data were analysed using meta-analytical methods and comparisons between different algorithms were performed.

RESULTS

Of 765 articles retrieved, 30 studies were eligible for meta-analysis. Pooled sensitivity for artificial intelligence was slightly higher than for dermoscopy (91% vs. 88%; P = 0.076). Pooled specificity for dermoscopy was significantly better than artificial intelligence (86% vs. 79%; P < 0.001). Pooled diagnostic odds ratio was 51.5 for dermoscopy and 57.8 for artificial intelligence, which were not significantly different (P = 0.783). There were no significance differences in diagnostic odds ratio among the different dermoscopic diagnostic algorithms.

CONCLUSIONS

Dermoscopy and artificial intelligence performed equally well for diagnosis of melanocytic skin lesions. There was no significant difference in the diagnostic performance of various dermoscopy algorithms. The three-point checklist, the seven-point checklist and Menzies score had better diagnostic odds ratios than the others; however, these results need to be confirmed by a large-scale high-quality population-based study.

Artificial neural network: border detection in echocardiography

Being non-invasive and low cost, the echocardiography has become a diagnostic technique largely applied for the determination of the left ventricle systolic and diastolic volumes, which are used indirectly to calculate the left ventricle ejection volume, the cardiac cavities muscular contraction, the regional ejection fraction, the myocardial thickness, and the ventricular mass, etc. However, the image is very noisy, which renders the delineation of the borders of the left ventricle very difficult. While there are many techniques image segmentation, this work chooses the artificial neural network (ANN) since it is not very sensitive to noise. In order to reduce the processing time, the operator selects the region of interest where the neural network will identify the borders. Neighborhood and gradient search techniques are then employed to link the points and the left ventricle contour is traced. The present method has been efficient in detecting the left ventricle borders echocardiography images compared to those whose borders were delineated by the specialists. For good results, it is important to choose properly the areas to be analyzed and the central points of these areas.

Skin and cutaneous melanocytic lesion simulation in biomedical optics with multilayered phantoms

The complex inner layered structure of skin influences the photon diffusion inside the cutaneous tissues and determines the reflectance spectra formation. Phantoms are very useful tools to understand the biophysical meaning of parameters involved in light propagation through the skin. To simulate the skin reflectance spectrum, we realized a multilayered skin-like phantom and a multilayered skin phantom with a melanoma-like phantom embedded inside. Materials used were Al(2)O(3) particles, melanin of sepia officinalis and a calibrator for haematology systems dispersed in transparent silicon. Components were optically characterized with indirect techniques. Reflectance phantom spectra were compared with average values of in vivo spectra acquired on a sample of 573 voluntary subjects and 132 pigmented lesions. The phantoms' reflectance spectra agreed with those measured in vivo, mimicking the optical behaviour of the human skin. Further, the phantoms were optically stable and easily manageable, and represented a valid resource in spectra formation comprehension, in diagnostic laser applications and simulation model implementation, such as the Monte Carlo code for non-homogeneous media.

Multispectral imaging and artificial neural network: mimicking the management decision of the clinician facing pigmented skin lesions

Various instruments based on acquisition and elaboration of images of pigmented skin lesions have been developed in an attempt to in vivo establish whether a lesion is a melanoma or not. Although encouraging, the response of these instruments, e.g. epiluminescence microscopy, reflectance spectrophotometry and fluorescence imaging, cannot currently replace the well-established diagnostic procedures. However, in place of the approach to instrumentally assess the diagnosis of the lesion, recent studies suggest that instruments should rather reproduce the assessment by an expert clinician of whether a lesion has to be excised or not. The aim of this study was to evaluate the performance of a spectrophotometric system to mimic such a decision. The study involved 1794 consecutively recruited patients with 1966 doubtful cutaneous pigmented lesions excised for histopathological diagnosis and 348 patients with 1940 non-excised lesions because clinically reassuring. Images of all these lesions were acquired in vivo with a multispectral imaging system. The data set was randomly divided into a train (802 reassuring and 1003 excision-needing lesions, including 139 melanomas), a verify (464 reassuring and 439 excision-needing lesions, including 72 melanomas) and a test set (674 reassuring and 524 excision-needing lesions, including 76 melanomas). An artificial neural network (ANN(1)) was set up to perform the classification of the lesions as excision-needing or reassuring, according to the expert clinicians' decision on how to manage each examined lesion. In the independent test set, the system was able to emulate the clinicians with a sensitivity of 88% and a specificity of 80%. Of the 462 correctly classified as excision-needing lesions, 72 (95%) were melanomas. No major variations in receiver operating characteristic curves were found between the test and the train/verify sets. On the same data set, a further artificial neural network (ANN(2)) was then architected to perform classification of the lesions as melanoma or non-melanoma, according to the histological diagnosis. Having set the sensitivity in recognizing melanoma to 95%, ANN(1) resulted to be significantly better in the classification of reassuring lesions than ANN(2). This study suggests that multispectral image analysis and artificial neural networks could be used to support primary care physicians or general practitioners in identifying pigmented skin lesions that require further investigations.

Optical devices used for image analysis of pigmented skin lesions: a proposal for quality assurance protocol using tissue-like phantoms

Different technological tools have been developed to aid in the diagnosis of pigmented skin lesions, including cameras working with conventional RGB colour systems, epiluminescence microscopy and spectrophotometric methods using visible and near infrared wavelengths. All the different procedures should provide in an objective and reproducible fashion quantitative measurements of the colour and shape features of a given skin mole. At present, many devices have been introduced in experimental stages for clinical diagnosis, mainly used to provide to the clinicians an objective, computer-assisted second opinion. As for any diagnostic instruments, optical devices should also be subjected to a dedicated quality assurance protocol in order to evaluate the response repeatability of each device (intra-instrument agreement) and to check the accordance among the responses of different devices (inter-instrument agreement). The aim of this study was to design a quality assurance protocol for optical devices dedicated to image analysis of pigmented skin lesions and, in case, to detect cutaneous melanoma by using suitable tissue-like phantoms as standard references that enable testing of both hardware and software components. As an example, we report the results of intra-instrument and inter-instrument agreement when the protocol was applied on a series of 30 SpectroShade instruments, a novel optical device based on multi-spectral image analysis of colour and shape features of pigmented skin lesion.

The use of artificial neural networks in decision support in cancer: a systematic review

Artificial neural networks have featured in a wide range of medical journals, often with promising results. This paper reports on a systematic review that was conducted to assess the benefit of artificial neural networks (ANNs) as decision making tools in the field of cancer. The number of clinical trials (CTs) and randomised controlled trials (RCTs) involving the use of ANNs in diagnosis and prognosis increased from 1 to 38 in the last decade. However, out of 396 studies involving the use of ANNs in cancer, only 27 were either CTs or RCTs. Out of these trials, 21 showed an increase in benefit to healthcare provision and 6 did not. None of these studies however showed a decrease in benefit. This paper reviews the clinical fields where neural network methods figure most prominently, the main algorithms featured, methodologies for model selection and the need for rigorous evaluation of results.

Automated segmentation of pigmented skin lesions in multispectral imaging

The aim of this study was to develop an algorithm for the automatic segmentation of multispectral images of pigmented skin lesions. The study involved 1700 patients with 1856 cutaneous pigmented lesions, which were analysed in vivo by a novel spectrophotometric system, before excision. The system is able to acquire a set of 15 different multispectral images at equally spaced wavelengths between 483 and 951 nm. An original segmentation algorithm was developed and applied to the whole set of lesions and was able to automatically contour them all. The obtained lesion boundaries were shown to two expert clinicians, who, independently, rejected 54 of them. The 97.1% contour accuracy indicates that the developed algorithm could be a helpful and effective instrument for the automatic segmentation of skin pigmented lesions.

Multi-spectral image analysis and classification of melanoma using fuzzy membership based partitions

The sensitivity and specificity of melanoma diagnosis can be improved by adding the lesion depth and structure information obtained from the multi-spectral, trans-illumination images to the surface characteristic information obtained from the epi-illumination images. Wavelet transform based bi-modal channel energy features obtained from the images are used in the analysis. Methods using both crisp and fuzzy membership based partitioning of the feature space are evaluated. For this purpose, the ADWAT classification method that uses crisp partitioning is extended to handle multi-spectral image data. Also, multi-dimensional fuzzy membership functions with Gaussian and Bell profiles are proposed for classification. Results show that the fuzzy membership functions with Bell profile are more effective than the extended ADWAT method in discriminating melanoma from dysplastic nevus.

Automated melanoma detection with a novel multispectral imaging system: results of a prospective study

The aim of this research was to evaluate the performance of a new spectroscopic system in the diagnosis of melanoma. This study involves a consecutive series of 1278 patients with 1391 cutaneous pigmented lesions including 184 melanomas. In an attempt to approach the 'real world' of lesion population, a further set of 1022 not excised clinically reassuring lesions was also considered for analysis. Each lesion was imaged in vivo by a multispectral imaging system. The system operates at wavelengths between 483 and 950 nm by acquiring 15 images at equally spaced wavelength intervals. From the images, different lesion descriptors were extracted related to the colour distribution and morphology of the lesions. Data reduction techniques were applied before setting up a neural network classifier designed to perform automated diagnosis. The data set was randomly divided into three sets: train (696 lesions, including 90 melanomas) and verify (348 lesions, including 53 melanomas) for the instruction of a proper neural network, and an independent test set (347 lesions, including 41 melanomas). The neural network was able to discriminate between melanomas and non-melanoma lesions with a sensitivity of 80.4% and a specificity of 75.6% in the 1391 histologized cases data set. No major variations were found in classification scores when train, verify and test subsets were separately evaluated. Following receiver operating characteristic (ROC) analysis, the resulting area under the curve was 0.85. No significant differences were found among areas under train, verify and test set curves, supporting the good network ability to generalize for new cases. In addition, specificity and area under ROC curve increased up to 90% and 0.90, respectively, when the additional set of 1022 lesions without histology was added to the test set. Our data show that performance of an automated system is greatly population dependent, suggesting caution in the comparison with results reported in the literature. In our opinion, scientific reports should provide, at least, the median values of thickness and dimension of melanomas, as well as the number of small (6 mm) melanomas.

Multivariate image analysis in biomedicine

In recent years, multivariate imaging techniques are developed and applied in biomedical research in an increasing degree. In research projects and in clinical studies as well m-dimensional multivariate images (MVI) are recorded and stored to databases for a subsequent analysis. The complexity of the m-dimensional data and the growing number of high throughput applications call for new strategies for the application of image processing and data mining to support the direct interactive analysis by human experts. This article provides an overview of proposed approaches for MVI analysis in biomedicine. After summarizing the biomedical MVI techniques the two level framework for MVI analysis is illustrated. Following this framework, the state-of-the-art solutions from the fields of image processing and data mining are reviewed and discussed. Motivations for MVI data mining in biology and medicine are characterized, followed by an overview of graphical and auditory approaches for interactive data exploration. The paper concludes with summarizing open problems in MVI analysis and remarks upon the future development of biomedical MVI analysis.