Comparison of veterinarians and a deep learning tool in the diagnosis of equine ophthalmic diseases

BACKGROUND/OBJECTIVES

The aim was to compare ophthalmic diagnoses made by veterinarians to a deep learning (artificial intelligence) software tool which was developed to aid in the diagnosis of equine ophthalmic diseases. As equine ophthalmology is a very specialised field in equine medicine, the tool may be able to help in diagnosing equine ophthalmic emergencies such as uveitis.

STUDY DESIGN

In silico tool development and assessment of diagnostic performance.

METHODS

A deep learning tool which was developed and trained for classification of equine ophthalmic diseases was tested with 40 photographs displaying various equine ophthalmic diseases. The same data set was shown to different groups of veterinarians (equine, small animal, mixed practice, other) using an opinion poll to compare the results and evaluate the performance of the programme. Convolutional Neural Networks (CNN) were trained on 2346 photographs of equine eyes, which were augmented to 9384 images. Two hundred and sixty-one separate unmodified images were used to evaluate the trained network. The trained deep learning tool was used on 40 photographs of equine eyes (10 healthy, 12 uveitis, 18 other diseases). An opinion poll was used to evaluate the diagnostic performance of 148 veterinarians in comparison to the software tool.

RESULTS

The probability for the correct answer was 93% for the AI programme. Equine veterinarians answered correctly in 76%, whereas other veterinarians reached 67% probability for the correct diagnosis.

MAIN LIMITATIONS

Diagnosis was solely based on images of equine eyes without the possibility to evaluate the inner eye.

CONCLUSIONS

The deep learning tool proved to be at least equivalent to veterinarians in assessing ophthalmic diseases in photographs. We therefore conclude that the software tool may be useful in detecting potential emergency cases. In this context, blindness in horses may be prevented as the horse can receive accurate treatment or can be sent to an equine hospital. Furthermore, the tool gives less experienced veterinarians the opportunity to differentiate between uveitis and other ocular anterior segment disease and to support them in their decision-making regarding treatment.

Evaluation of substance P as a biomarker for pain in equine colic

Equine colic is an important condition associated with acute abdominal pain and one of the leading causes of death in horses. As such, objectively evaluating pain is of interest for attending veterinarians. Pain scales for assessment are present, but no single pain-specific biomarker has been reported. The aim of this study was to determine if substance P (SP) could be a reliable biomarker to reflect pain and serve as a parameter to predict outcome in equine colic. The hypothesis was that horses displaying severe colic signs present with higher values of SP in contrast to those with mild colic signs. Thirty warmblood horses, aged between 3 and 20 years were recruited; evenly distributed (10 horses each) in three colic groups (mild, moderate, severe). To classify the colic signs, the horses were graded by the Equine Acute Abdominal Pain Scale (EAAPS). Clinical examination and EAAPS were performed at arrival in the hospital. Blood samples were collected four times in hourly intervals commencing from arrival. For comparison, already established parameters for prognosticating equine colic (heart rate, serum cortisol, and blood lactate concentration) were also measured. The assumption of increasing SP concentrations along with pain could not be confirmed. SP did not show any association with heart rate, cortisol, lactate, or EAAPS. Whereas the established parameters increased according to the EAAPS, SP remained stable in individual horses regardless of clinical signs, treatment, and disease progression. Consequently, SP was not a reliable parameter to reflect painful conditions or to predict outcome in equine colic.

Evaluation of Substance P as a New Stress Parameter in Horses in a Stress Model Involving Four Different Stress Levels

Stress has a significant impact on equine welfare. There are some studies on the stress response in horses ridden with tight nosebands, but little is known about other stress parameters than cortisol, which potentially could address an emotional component. In this study, blood samples of a total of 74 warmblood horses were used to establish reference values for plasma substance P (SP) concentrations. Moreover, 16 of these warmblood horses were included in a stress model. Four different stress levels (level 1: horses ridden with loose noseband, level 2: tight noseband, level 3: loose noseband and overground endoscope, level 4: tight noseband and overground endoscope) were applied to evaluate SP as a potential stress parameter in horses. Blood samples were taken at rest (t0) and directly after inducing stress (noseband tightening, insertion of endoscope; t1), as well as after 20 min of riding at all gaits (t2). A ridden horse ethogram was applied and showed that horses in the tight noseband group resorted to other stress-related behavioral issues than horses with loose nosebands. Serum cortisol showed a linear increase concurrent with the increase in stress levels with a significant difference between level 1 and level 4 (p = 0.043), proving that stress factors were adequate to evaluate the stress response, whereas SP did not show a correlation with the stress levels. Furthermore, concentrations of SP differed widely between horses but stayed within more narrow limits in the individual horse. As a conclusion, SP might not be a reliable stress parameter in horses in the applied minor stress model.

Tailoring fatty food simulants made from solvent mixtures (2): determining the equivalent migration behaviour of olive oil and of solvents in the case of polyolefins

It is suggested that solvent mixtures consisting of an ester and of an inert solvent can be used as fatty food simulants capable of having the same migration behaviour as olive oil with plastics. Migration tests carried out with low-density polyethylene for 20 and 48 h in an 8 and 5% mixture of tert-butyl acetate in ethanol respectively gave results equivalent to those obtained with olive oil after 10 days at 40 degrees C. The use of solvent mixtures facilitated the analysis and improved detection limits, giving good repeatability. Furthermore, the more rapid migration in solvent mixtures can be particularly useful for industrial controls as alternative test methods.

Safety and quality of food contact materials. Part 1: evaluation of analytical strategies to introduce migration testing into good manufacturing practice

The results of a research project (EU AIR Research Programme CT94-1025) aimed to introduce control of migration into good manufacturing practice and into enforcement work are reported. Representative polymer classes were defined on the basis of chemical structure, technological function, migration behaviour and market share. These classes were characterized by analytical methods. Analytical techniques were investigated for identification of potential migrants. High-temperature gas chromatography was shown to be a powerful method and 1H-magnetic resonance provided a convenient fingerprint of plastic materials. Volatile compounds were characterized by headspace techniques, where it was shown to be essential to differentiate volatile compounds desorbed from those generated during the thermal desorption itself. For metal trace analysis, microwave mineralization followed by atomic absorption was employed. These different techniques were introduced into a systematic testing scheme that is envisaged as being suitable both for industrial control and for enforcement laboratories. Guidelines will be proposed in the second part of this paper.

Alternative test methods to control the compliance of polyolefin food packaging materials with the European Union regulation: the case of aromatic antioxidants and of bis(ethanolamine) antistatics based on 1H-NMR and UV-visible spectrophotometry

In order to decide whether a plastic food packaging material complies with the European Communities (EC) regulation on migration, a quick analysis of two functional classes of plastics additives (aromatic antioxidants and antistatic agents) from polyolefin materials by 1H-NMR and UV-visible spectrophotometry is presented. The scope of spectroscopic methods for alternative and migration tests is presented. 1H-NMR can be used in several ways, from a simple fingerprint of the potential migrants to an identification procedure. Extraction is optimized using UV spectrophotometry. Optimization relies on extraction kinetics, which include the demonstration that extraction is more severe than migration. Only a few hours are required to conclude whether a material complies with the regulation. The specific migration limits are expressed as specific absorbance limit values, alpha. These data can be annexed by food industries to specifications of a plastic packaging material.

Quick methods to control compliance of plastic materials with food packaging regulations

A general strategy is presented, aiming to provide plastics manufacturers, food industries and enforcement laboratories with quick methods to check whether migration from materials for food contact will be acceptable during the time of use. The strategy involves several steps, with increasing time demand and cost. Monitoring extraction kinetics allows both the optimization of the extraction time, and the selection of conditions where extraction is more severe than migration. The influence of the extracting solvent is discussed. It may give rise to specific non-extraction of some migrants, which may change the conclusions when the solvent is used in replacement of a fatty food simulant. Factors ruling this effect at a given temperature are identified: the affinity to the solvent with the migrant (selectivity), its ability to reach molecules entangled in the polymeric network (accessibility) and its interaction with the polymer (penetration). The kinetic parameters of the penetration of olive oil into polypropylene have been determined by the determination of profiles of concentration.

Guidelines of the Commission of the European Communities: a challenge for the control of packaging

Questions arising from the Commission of the European Communities Directives and guidelines regulating packaging materials are discussed in relation to whether compliance ensures safety in use and the consequent analytical problems. Difficulties may arise from interactions between food contact materials and food involving mass transfer (migration, off-odours, 'scalping', loss of aroma) or mass transfer and chemical interactions and the implications for safety assurance and regulation are addressed. The criteria for suitable low molecular weight fatty food simulants and conditions for migration testing are presented. In food surveillance, the usefulness of various methods of analysis differs for monomers and for additives. For monomers, IR spectroscopy can identify the polymer type and which specific monomers need controlling; for unknown mixtures of additives, preliminary functional group identification by techniques such as 1H-NMR is useful.

1H-NMR for the safety control of food packaging materials: analysis of extracts from polyolefin samples

A new approach for the identification of the additives in extracts of packaging plastics, using proton nuclear magnetic resonance (1H-NMR), is presented. The technique can be used in a preliminary step for the determination of contaminants potentially released by the food packaging materials; it may greatly reduce the time required to identify the constituents of the materials, either individually or as functional classes. A classification of the EEC additives is proposed on the basis of chemical shift. 1H-NMR can also be used as a fingerprint technique in the quality control of the food packaging materials.

The increase in skin 7-dehydrocholesterol induced by an hypocholesterolemic agent is associated with elevated 25-hydroxyvitamin D3 plasma level

Vitamin D3 is generated in skin by UV irradiation of 7-dehydrocholesterol (7-DEHC). Whether the 7-DEHC amount in skin affects vitamin D3 formation, and thereby the plasma level of 25-hydroxyvitamin D3 (25[OH]D3) is not known. In the present work we report on the influence on vitamin D and Ca metabolism of a new hypocholesterolemic agent, HCG-917 (0-2-[hydroxy-3-]N'-(2-chlorophenyl)-N-piperazinyl-1- [propyl]-4-chloro-benz-aldoxim-hydrochloride) which inhibits 7-DEHC reductase and thereby increases skin 7-DEHC. Rats were treated with HCG 917 (0.3 and 5.0 mg/kg, orally) for 13 days. HCG 917 caused a dose-dependent decrease in cholesterol and concomitant accumulation of 7-DEHC in plasma and skin. In skin, 7-DEHC was: control: 1.05 +/- 0.20; HCG 917, 0.3 mg/kg: 1.41 +/- 0.22; HCG 917, 5.0 mg/kg: 2.35 +/- 0.35 mg/g. At a dose of 0.3 mg/kg, HCG 917 had no significant influence on the plasma level of neither 25(OH)D3 nor 1,25(OH)2D3. However, at a dose of 5.0 mg/kg, HCG 917 induced a significant increase in plasma 25(OH)D3 (control: 36.2 +/- 2.2; HCG 917 5.0 mg/kg: 57.6 +/- 6.5 nmol/l) and a slight but not significant rise in 1,25(OH)2D3. Calcium balance studies indicated that HCG 917 did not influence intestinal Ca absorption nor urinary Ca excretion. At a dose of 5.0 mg/kg HCG 917 slightly induced a decrease in total plasma Ca. In conclusion, HCG 917 treatment can induce a significant rise in skin 7-DEHC with an increase in plasma 25(OH)D3. These results suggest that variation in the skin level of 7-DEHC can directly influence the cutaneous production of vitamin D3 and thereby the vitamin D status of the organism.