Automatic Digital Image Analysis Research Articles

Tissue microarrey: a potential cost-effective approach for mismatch repair testing in colorectal cancer

Deficiencies in Mismatch Repair (MMR) proteins are one of the major pathways in the development of colorectal cancer (CRC). MMR status evaluation is recommended in every new CRC patient. However, this is not fully implemented due to high costs. Tissue microarray (TMA) enables allocating tissue cores from few specimens to a single paraffin block. The primary objective of this study was to evaluate the accuracy of TMA MMR immunohistochemistry (IHC) compared to whole slide. The secondary objective was to evaluate and validate automatic digital image analysis software in differentiating pathological and normal TMA cores. Pathological cores were defined if at least one MMR protein was unstained. Tumoral and normal tissue of 11 CRC patients with known MMR status was used to obtain 623 TMA cores. The MMR staining of each core was evaluated by a pathologist and compared to the whole slide result. Digital analysis software by 3DHistech Ltd. was used to identify cell nucleus and quantify nuclear staining in 323 tissue cores. To identifying pathological tissue, cores the cohort was divided into a test (N = 146 cores) and validation sets (N = 177 cores). A staining intensity score (SIS) was developed, and its performance compared to the pathologist review of each core and to the whole slide result. Compared to the whole slide, the pathologist's assessment had 100% sensitivity (n/N = 112/112) and 100% specificity (n/N = 278/278) with 95% lower limit of 97 and 99% respectively. The area under the receiver operating characteristic (ROC) curve of SIS was 77%. A cutoff of 55 was obtained from the ROC curve. By implementing the cutoff in the validation dataset, the SIS had sensitivity and specificity of 98.2% [90.1-100%] and 58.5% [49.3-67.4%] respectively. The MMR status of CRC can be evaluated in TMA tissue cores thus potentially reducing MMR testing costs. The SIS can be used as triage indicator during pathologic review. Institutional ethical approval was granted for the performance of this study (Emek Medical Center Ethics ID: EMC-19-0179).

Comparison of Morphological and Digital-Assisted Analysis for BCL6 Endometrial Expression in Women with Endometriosis.

BCL6 (B-cell lymphoma 6) is a proto-oncogene and transcriptional repressor initially described as being involved in B-cell lymphoma. Recently, this factor has been identified as a promising tissue biomarker which could be used to diagnose women affected by endometriosis. Previous studies used HSCORE for BCL6 staining quantification in the endometrium. However, this semi-quantitative technique of analysis has some limitations, including a lack of objectivity, robustness, and reproducibility that may lead to intra- and inter-observer variability. Our main goal was to develop an original computer-assisted method to quantify BCL6 staining from whole-slide images reliably. In order to test the efficiency of our new digital method of quantification, we compared endometrial BCL6 expression between fertile and infertile women without or with different stages of endometriosis by using the widely used HSCORE analysis and our new automatic digital image analysis. We find a higher expression of BCL6 in the endometrium of infertile women with endometriosis and women with stage IV endometriosis. Furthermore, we demonstrate a significant correlation between the two types of independent measurements, indicating the robustness of results and also the reliability of our computer-assisted method for BCL6 quantification. In conclusion, our work, by using this original computer-assisted method, enables BCL6 quantification more objectively, reliably, robustly, and promptly compared to HSCORE analysis.

From Individual Graphite Assignment to an Improved Digital Image Analysis of Ductile Iron

Since graphite classification by visual analysis exhibits large variations, a more integrative concept of graphite shape classification is required to evaluate the correlations of process, microstructure and properties, and to fulfill customers’ requirements. The automatic digital image analysis is partly based on visual analysis, but it is not thoroughly defined for graphite shape classification. For example, nodules and thereby nodularity are only defined by the shape parameter roundness, although several studies suggest more sophisticated approaches. Within the first of three successive round robin tests, visual assignment for a variety of graphite particles was performed to obtain a universal digital data set of classified graphite particles. For this, the classification approach from standard EN ISO 945-1 was used and extended with degenerated graphite. The assigned particles were evaluated concerning different shape parameters showing that roundness and the assigned minimum limit value of 0.6 are not sufficient to distinguish nodules from less ideal graphite particle shapes. Furthermore, the current classification approach does not represent the full spectrum of graphite morphologies and needs to be extended. The development of a universal hierarchical classification method for nodules and other graphite shapes has been initiated, and the results will contribute to an improved image analysis standard for ductile iron, particularly ISO 945-4.

Deep Learning From Limited Training Data: Novel Segmentation and Ensemble Algorithms Applied to Automatic Melanoma Diagnosis

Deep learning algorithms often require thousands of training instances to generalize well. The presented research demonstrates a novel algorithm, Predict-Evaluate-Correct K-fold (PECK), that trains ensembles to learn well from limited data. The PECK algorithm is used to train a deep ensemble on 153 non-dermoscopic lesion images, significantly outperforming prior publications and state-of-the-art methods trained and evaluated on the same dataset. The PECK algorithm merges deep convolutional neural networks with support vector machine and random forest classifiers to achieve an introspective learning method. Where the ensemble is organized hierarchically, deeper layers are provided not only more training folds, but also the predictions of previous layers. Subsequent classifiers then learn and correct the previous layer errors by training on the original data with injected predictions for new data folds. In addition to the PECK algorithm, a novel segmentation algorithm, Synthesis and Convergence of Intermediate Decaying Omnigradients (SCIDOG), is developed to accurately detect lesion contours in non-dermoscopic images, even in the presence of significant noise, hair, and fuzzy lesion boundaries. As SCIDOG is a non-learning algorithm, it is unhindered by data quantity limitations. The automatic and precise segmentations that SCIDOG produces allows for the extraction of 1,812 lesion features that quantify shape, color and texture. These morphological features are used in conjunction with convolutional neural network predictions for training the PECK ensemble. The combination of SCIDOG and PECK algorithms are used to diagnose melanomas and benign nevi through automatic digital image analysis on the MED-NODE dataset. Evaluated using 10-fold cross validation, the proposed methods achieve significantly increased diagnostic capability over the best prior methods.

Programmed death-ligand 1 expression by digital image analysis advances thyroid cancer diagnosis among encapsulated follicular lesions

Recognition of noninvasive follicular thyroid neoplasms with papillary-like nuclear features (NIFTP) that distinguishes them from invasive malignant encapsulated follicular variant of papillary thyroid carcinoma (EFVPTC) can prevent overtreatment of NIFTP patients. We and others have previously reported that programmed death-ligand 1 (PD-L1) is a useful biomarker in thyroid tumors; however, all reports to date have relied on manual scoring that is time consuming as well as subject to individual bias. Consequently, we developed a digital image analysis (DIA) protocol for cytoplasmic and membranous stain quantitation (ThyApp) and evaluated three tumor sampling methods [Systemic Uniform Random Sampling, hotspot nucleus, and hotspot nucleus/3,3′-Diaminobenzidine (DAB)]. A patient cohort of 153 cases consisting of 48 NIFTP, 44 EFVPTC, 26 benign nodules and 35 encapsulated follicular lesions/neoplasms with lymphocytic thyroiditis (LT) was studied. ThyApp quantitation of PD-L1 expression revealed a significant difference between invasive EFVPTC and NIFTP; but none between NIFTP and benign nodules. ThyApp integrated with hotspot nucleus tumor sampling method demonstrated to be most clinically relevant, consumed least processing time, and eliminated interobserver variance. In conclusion, the fully automatic DIA algorithm developed using a histomorphological approach objectively quantitated PD-L1 expression in encapsulated thyroid neoplasms and outperformed manual scoring in reproducibility and higher efficiency.

Accuracy of Reconstruction of the Tree Stem Surface Using Terrestrial Close-Range Photogrammetry

Airborne laser scanning (ALS) allows for extensive coverage, but the accuracy of tree detection and form can be limited. Although terrestrial laser scanning (TLS) can improve on ALS accuracy, it is rather expensive and area coverage is limited. Multi-view stereopsis (MVS) techniques combining computer vision and photogrammetry may offer some of the coverage benefits of ALS and the improved accuracy of TLS; MVS combines computer vision research and automatic analysis of digital images from common commercial digital cameras with various algorithms to reconstruct three-dimensional (3D) objects with realistic shape and appearance. Despite the relative accuracy (relative geometrical distortion) of the reconstructions available in the processing software, the absolute accuracy is uncertain and difficult to evaluate. We evaluated the data collected by a common digital camera through the processing software (Agisoft PhotoScan ©) for photogrammetry by comparing those by direct measurement of the 3D magnetic motion tracker. Our analyses indicated that the error is mostly concentrated in the portions of the tree where visibility is lower, i.e., the bottom and upper parts of the stem. For each reference point from the digitizer we determined how many cameras could view this point. With a greater number of cameras we found increasing accuracy of the measured object space point positions (as expected), with a significant positive change in the trend beyond five cameras; when more than five cameras could view this point, the accuracy began to increase more abruptly, but eight cameras or more provided no increases in accuracy. This method allows for the retrieval of larger datasets from the measurements, which could improve the accuracy of estimates of 3D structure of trees at potentially reduced costs.

Detection Of Automatic Digital Image Analysis Problems For The Evaluation Of Immune Markers In Breast Cancer Biopsies

Introduction/ Background Automatic digital image analysis has increased in recent years. There are several applications for image analysis in pathology, among them, immunohistochemistry biomarkers quantification. This is a simple, economic and fast method to quantify stained biomarkers in digitalized biopsies enhancing sensitivity and objectivity. However, automatic procedures do not always work satisfactorily in all digitalized samples. Aims To quantify the number of incorrectly analyzed images as a result of errors on the automatic procedure developed for quantifying immune markers in breast cancer biopsies, to evaluate the amount of time spent in this reanalysis and to define which kind of biopsy produces more errors. Methods 10,770 cores of breast tumor (intra-tumoral and peri-tumoral areas) and negative and positive axillary lymph nodes areas from ductal invasive breast cancer were included in different tissue microarrays (TMAs). Slides of each TMA were immunohistochemically stained for CD4, CD8, CD57, CD68, S100, LAMP3, CD83, CD1A, CD123 and CD21 immune markers and were digitalized at 40X with the Apperio Scanscope XT scanner. Each core was extracted as an individual digital image in TIFF format. The stained area was automatically quantified using procedures developed with Fiji (Image J) software in a HP Intel Inside Core i.7 computer with 16GB of RAM memory. Firstly, the whole area of each core was evaluated in pixels by using the luminance channel, applying the median filter and gray-scale segmentation. The second step evaluated the positive pixel number stained in brown for obtaining a brown color channel and then, a gray scale and size segmentation for positive objects, including holes inside the segmented area. Finally, this selected brown area was automatically surrounded by an overlay. Results 2,751 had to be reanalyzed (25.5%). Specifically, intra-tumoral and peri-tumoral cores were those with higher reanalysis levels (35.6% and 34.9%, respectively) while axillary lymph nodes cores present lower levels (Negative Nodes 13.5% and Positive Nodes 16.4%). Regarding the immune biomarker, CD21, LAMP3 and CD123 were those with higher reanalysis levels (38.4%, 35.2% and 34.2%, respectively) in contrast with CD8, CD68 and CD83 that were those with lower levels (16.4%, 17.7%, 18.4%, respectively). The reanalysis levels of the remaining biomarkers were 22% in CD4, 23.9% in S100, 26.% in CD1A and 27.4% in CD57. Each core took a mean of 1.56 minutes to be analyzed so the total time spent in the reanalysis of the images was 71.5 hours. The principal reasons for reanalysis were problems in the TMAs assembly accuracy and presence of adipose tissue, hemosiderin, artifacts, unspecific staining and background noise. Several TMAs presented glue bubbles and different types of dirt, as hairs or dust that were quantified as positive area. Intratumoral and peritumoral cores are those with higher levels of adipose tissue. These adipocytes cause alterations in the automatic quantification due to their different cellular structure. Specifically, in cores stained for S100 marker in which membranes of adipocytes were also stained in brown and, thus, quantified as immune marker. C o nclusions: It could be possible to reduce the time of analysis and to obtain more exact values of immune quantification in digital images improving the accuracy of TMAs assembly, overcoming unspecific staining and background and adapting the parameters of the procedures.

Measurement of the degraded depth in cementitious materials by automatic digital image processing

The combination of a staining method and an automatic digital image-processing algorithm is presented here, to measure degradation depths in cementitious materials. The measurement of those degraded depths is usually made by direct visual measurements, resulting in many errors and low reproducibility. The automatic digital image analysis (ADIA) method proposed here is mainly based on the differentiation of the degraded zone and the sound zone on the basis of the image histogram. The method comprises several steps, such as sample alignment, image calibration, background subtraction, image filtering, automatic segmentation and final measurement of the degraded depth. The algorithm developed has been used to measure the degraded depths of a set of decalcified cement mortars, made from different cement types and with varying w/b ratios. Relative to previous methods, this automatic procedure improves the precision (about 0.03 mm) and the statistical representation of the measurements. The results obtained by ADIA were compared with direct visual measurements with a very good correlation (R2 = 0.96) and a mean error of 6%.

Use of the TrichoScan to Assess Female Pattern Hair Loss

TrichoScan is a method of hair growth measurement that combines epiluminescence microscopy with automatic digital image analysis. Miniaturization is considered a cornerstone of androgenetic alopecia (AGA) and is requisite for establishing the diagnosis. To examine miniaturization in women with AGA. Twenty-eight women with Ludwig pattern (aged 21-68, mean 46) were included. Miniaturization was estimated using the TrichoScan method. Almost one-quarter (21.4%) of the women had miniaturized hair only on the top of the head (the crown), 28.6% had miniaturized hair throughout the scalp, and 50.0% showed no miniaturization but a preponderance of thinning in the crown area. In all patients, a comparative study of hair diameter revealed significant differences between the crown and the occipital region. Our results imply that, in female hair loss, miniaturization might be a relative and not an absolute measurement, and AGA in women could be diagnosed also in the presence of a difference in hair shaft diameter.

How to measure image quality in tissue-based diagnosis (diagnostic surgical pathology)

BackgroundAutomated image analysis, measurements of virtual slides, and open access electronic measurement user systems require standardized image quality assessment in tissue-based diagnosis.AimsTo describe the theoretical background and the practical experiences in automated image quality estimation of colour images acquired from histological slides.Theory, material and measurementsDigital images acquired from histological slides should present with textures and objects that permit automated image information analysis. The quality of digitized images can be estimated by spatial independent and local filter operations that investigate in homogenous brightness, low peak to noise ratio (full range of available grey values), maximum gradients, equalized grey value distribution, and existence of grey value thresholds. Transformation of the red-green-blue (RGB) space into the hue-saturation-intensity (HSI) space permits the detection of colour and intensity maxima/minima. The feature distance of the original image to its standardized counterpart is an appropriate measure to quantify the actual image quality. These measures have been applied to a series of H&E stained, fluorescent (DAPI, Texas Red, FITC), and immunohistochemically stained (PAP, DAB) slides. More than 5,000 slides have been measured and partly analyzed in a time series.ResultsAnalysis of H&E stained slides revealed low shading corrections (10%) and moderate grey value standardization (10 – 20%) in the majority of cases. Immunohistochemically stained slides displayed greater shading and grey value correction. Fluorescent stained slides are often revealed to high brightness. Images requiring only low standardization corrections possess at least 5 different statistically significant thresholds, which are useful for object segmentation. Fluorescent images of good quality only posses one singular intensity maximum in contrast to good images obtained from H&E stained slides that present with 2 – 3 intensity maxima.ConclusionEvaluation of image quality and creation of formally standardized images should be performed prior to automatic analysis of digital images acquired from histological slides. Spatial dependent and local filter operations as well as analysis of the RGB and HSI spaces are appropriate methods to reproduce evaluated formal image quality.

MIB1 expression in basal cell layer: a diagnostic tool to identify premalignancies of the vulva

Lichen sclerosus, high-grade usual vulvar intraepithelial neoplasia (VIN) and differentiated VIN have a different malignant potential. The objective of this study was to quantify the proliferative activity in the basal region of the epithelium of vulvar premalignancies. Furthermore, we investigated whether MIB1 expression in the basal region of vulvar epithelium can be helpful in diagnosing differentiated VIN, which may be hard to discern from normal epithelium. MIB1 was used to immunohistochemically visualise proliferating cells within formalin-fixed, paraffin-embedded, archival tissue sections of different vulvar premalignancies (N=48) and normal vulvar epithelium (N=16). Automatic digital image analysis software was developed to quantify the proliferating fraction in different parts of the epithelium (MIB1 positivity index). MIB1 expression differed among the various vulvar premalignancies; a MIB1-negative basal cell layer was a distinct feature of normal vulvar epithelium. No MIB1-negative basal cell layer was noted in differentiated VIN or other vulvar premalignancies. Owing to this negative cell layer, the MIB1 proliferation index in normal vulvar epithelium was significantly lower than in vulvar premalignancies. In conclusion, MIB1 expression can be a helpful tool in diagnosing a premalignancy and has additional value especially to distinguish differentiated VIN neoplasia from normal vulvar epithelium, but cannot explain the differences in malignant potential.

Recent Findings with Computerized Methods for Scalp Hair Growth Measurements

Sensitive tools have been developed in order to monitor hair loss and treatment responses. Recently the Tricho-Scan was presented (by RH) as such a method which combines epiluminescence microscopy (ELM) with automatic digital image analysis. Herewith new TrichoScan data obtained from 10 women and 21 men with androgenetic hair loss after 6 mo of treatment with 5%-minoxidil are presented. Even in this small cohort of patients, we noticed a significant increase of hair density, cumulative hair thickness and terminal hair counts. Alternative methods were developed during a human alopecia investigation and research technology (HAIR Technology) programme at Skinterface. This involves contrast-enhancement, image acquisition, and processing by qualified technicians followed by computer-assisted image analysis. The specific identification of exogen hair, further adds to this very refined non-invasive investigative method for hair follicle function investigation. Regional variations of hair growth dynamics do exist in the human scalp such as in female patients complaining of hair loss, scalp hair density and growth on top of the head differs significantly from the occipital site. Finally, from transversal studies and from detailed monitoring of subsequent hair cycles during longitudinal studies, data were obtained that support the fact that shortening of hair cycle, slowing down of growth rates and thinning of hair shafts are heralding hair miniaturisation. In the workshop the TrichoScan, the method of Canfield and Skinterface have been shown.

Modelling and analysis of a deep-drawing operation: key factors for successful comparisons between experimental and simulated results

The objective is to determine the finest finite element simulations of deep-drawing operations, including spring-back effect, until the occurrence of strain localisation.This paper insists on two aspects that have a major influence on the results of the simulations: precise determination of the equivalent stress–equivalent strain relationship and description of boundary conditions.The stress–strain curve has been determined using duplicate analysis of tensile test. The classic extensometric strain measurement is compared with the measurements of strain with an automatic analysis of digital images caught at various deformation stages. The measurements by image analysis allow the determination of local strains and give them a more precise determination for the high strains.As to boundary limit problems managed with explicit algorithm, success mainly depends on the control of the process that has to remain in quasi-static range.In order to analyse the accordance of the results of the simulation and then obtained by the experiment, the precise method of residual stress calculation is elaborated for the case of the original tensile test: simultaneous loading and subsequent unloading to the equilibrium state of assembly of three samples with different lengths.The comparisons of the experimental and numerical results are very satisfactory.

Trichoscan: results using a 5% minoxidil topical solution in men with androgenetic alopecia

Hair loss or hair thinning is a common complaint and the effects of treatment attempts are hard to measure. Recently, we have developed the TrichoScan as a method, which combines epiluminescence (skin) microscopy (ELM) with automatic digital image analysis for the measurement of human, and potentially, animal hair, in situ. The TrichoScan is able to analyze all biological parameters of hair growth with a so-called intraclass correlation of approx. 91% within the same TrichoScan operator and of approx. 97% for different TrichoScan operators. Herewith, the power and accuracy of the TrichoScan is demonstrated by comparing the hair parameters of 10 men with androgenetic alopecia (AGA) who were treated for 6 months with a 5% minoxidil topical solution. Compared to baseline, we were able to detect a statistically significant increase in total hair counts (189, 6/cm2 vs. 163, 7/cm2; p 0.067) and cumulative hair thickness (18, 04 mm/cm2 vs 14, 82 mm/cm2 p 0.016) at month 6. Furthermore, we were able to detect an increase in terminal hairs (hairs thicker than 40 m) after treatment (175, 1/cm2 vs. 148, 8/cm2; p 0.082). These results show that the TrichoScan is able to measure significant changes in a variety of hair parameters, as shown in a rather small group of men treated with a 5% minoxidil topical solution for 6 months. Therefore, the TrichoScan is well suited for clinical studies as an evaluation tool for all hair colors to compare placebo to active drugs or to compare the capacities of different hair-growth promoting substances. It is well suited for the study of AGA or other forms of diffuse hair loss and can be adopted to study the effect of drugs or laser treatment on hypertrichosis or hirsutism.

TrichoScan. A new instrument for digital hair analysis

Hair loss or hair thinning is a common complaint in clinical dermatology. Patients seeking advice for hair loss are not necessarily bald. In addition, the effects of therapy are hard to measure. Consequently, there is a need for a sensitive tool to monitor hair loss and treatment response. Such a method must be able to analyze the biological parameters of hair growth, which are: 1: hair density (n/cm(2)), 2: hair diameter (micrometer), 3: hair growth rate (mm/day) and 4: anagen/telogen ratio. We present the TrichoScan as a method which combines epiluminescence microscopy (ELM) with automatic digital image analysis for the measurement of human, and potentially animal hair, in situ. The TrichoScan is able to analyze all biological parameters of hair growth with a so called intraclass correlation of approximately 91% within the same operator and an intraclass correlation of approximately 97% for different operators. The application of the technique is demonstrated by comparison of the hair parameters in individuals without apparent hair loss with men with untreated AGA and men after treatment with finasteride (1 mg/day), and women who were treated with minoxidil. We were able to detect a significant increase in hair counts and cumulative hair thickness 3 and 6 months after treatment. The advantage of the TrichoScan is that it can be used for clinical studies to compare placebo versus treatment or to compare different hair growth promoting substances, it can be used for studying AGA or other forms of diffuse hair loss, and it can be adopted to study the effect of drugs or laser treatment on hypertrichosis or hirsutism.

Screening for diabetic retinopathy using computer based image analysis and statistical classification

Diabetic retinopathy is one of the most common causes of blindness in Europe. However, efficient therapies do exist. An accurate and early diagnosis and correct application of treatment can prevent blindness in more than 50% of all cases. Digital imaging is becoming available as a means of screening for diabetic retinopathy. As well as providing a high quality permanent record of the retinal appearance, which can be used for monitoring of progression or response to treatment, and which can be reviewed by an ophthalmologist, digital images have the potential to be processed by automatic analysis systems. We have described the preliminary development of a tool to provide automatic analysis of digital images taken as part of routine monitoring of diabetic retinopathy in our clinic. Various statistical classifiers, a Bayesian, a Mahalanobis, and a KNN classifier were tested. The system was tested on 134 retinal images. The Mahalanobis classifier had the best results: microaneurysms, haemorrhages, exudates, and cotton wool spots were detected with a sensitivity of 69, 83, 99, and 80%, respectively.

The Art and Applications of Fluorescence In Situ Hybridization in Endocrine Pathology.

Fluorescence in situ hybridization (FISH) or molecular cytogenetics is currently recognized as a reliable, sensitive, and reproducible technique for identifying the copy number and structure of chromosomes. FISH combines molecular genetics with classic cytogenetics and allows simultaneous morphologic evaluation on a single slide. Centromeric DNA probes are used to detect specific chromosomes and telomeric probes to demonstrate all chromosomes. Sequence-specific probes can localize in situ a single gene copy on a specific chromosome locus. FISH allows cytogenetic investigation of metaphase spreads and interphase nuclei. Several protocols have been proposed to analyze preparations from fresh samples or archival material. Comparative genomic hybridization (CGH) is a novel cytogenetic technique, which combines FISH with automatic digital image analysis. Comparative analysis of the hybridization products of tumor DNA and reference DNA with normal metaphase chromosomes, each labeled with color different fluorochrome, can retrieve chromosomal imbalances of the entire genome in a single experiment. FISH and CGH are powerful morphologic tools in understanding physiologic mechanisms and in resolving problems of the pathogenesis of several diseases. These techniques shed light on the cytogenetic background in many endocrinological disorders, providing a better understanding of the activities and alterations of endocrine cell function.

Automatic assessment of sheep carcasses by image analysis

AbstractThe commercial value of animal carcasses depends not only on their weight but also on their composition and shape (termed conformation). This is usually assessed subjectively by a skilled inspector. In this paper an attempt is described to assess the saleable meat yield of sheep carcasses by automatic digital image analysis. A low-cost system based on a still video camera and a personal computer was used. The results indicate that better prediction of saleable meat yield can be obtained using objective measures of carcass shape than from subjective conformation scores. Information from the intensities of colour components was not found to be useful, possibly due to difficulties with lighting and image quality. Recommendations are made for implementing a practical system.