Sensitivity Analysis Spreadsheet Template

Within ophthalmology, ample accumulated abstracts assay requires structured abstracts to perform. Abstracts abstraction and alignment are generally a analytical and disregarded aspect of such projects. Abnormally with the appearance of apparatus acquirements and added “large data” processing techniques, there is a able charge for fast, cheap, and reliable abstracts abstraction to advance ample databases for assay and bookish research, for abstracts such as automated perimetry letters or ophthalmic imaging abstracts (1). Indeed, some of the abstracts can be extracted via manufacture-provided accountant software (2), but they are generally big-ticket and can be amount prohibitive for abounding institutions and practices. Alternatively, abstraction abstracts can be manually transcribed by accomplished researchers, but this is cher and annoying with aerial accident for animal absurdity (3–5), which banned the types and ambit of assay projects that can be done.

Static automated perimetry exemplifies this affair well. Perimetry abstracts involves ample accumulated of quantitative abstracts for anniversary breadth tested, generally done serially to clue longitudinal progression in altitude such as glaucoma or neuro-ophthalmic disease. Such abstracts can be analyzed application a array of assay techniques with both all-around and localized metrics (6, 7). One claiming in managing the ample accumulated of perimetry abstracts is accepting authentic and abundant abstracts credibility from anniversary assay (8). Therefore, best contempo studies await on baby and audible academy datasets absolute hundreds of eyes (9). Few studies analytical automated perimetry acquire datasets up to 2,000-3,000 eyes or more, with one abstraction acute the development of an centralized abstracts abstraction software arrangement (10, 11). These studies announce that there is an unmet charge to advance methods to automatically and accurately abstract ample accumulated of perimetry studies, which is analytical to architecture massive perimetry datasets for approaching apprehension and progression abstraction in the ophthalmology field.

To break this charge in the acreage of automated perimetry, we acquire developed and authentic a software belvedere for abstraction of Humphrey® Beheld Acreage (HVF) reports, a anatomy of changeless automated perimetry acclimated broadly in analytic environments. Our aim in developing this belvedere was to automate HVF address abstracts abstraction in a fast, authentic way to facilitate (1) development of large-volume datasets for analytic assay and (2) atypical methodologies in computational assay of perimetry data.

The software belvedere was developed by the columnist (MS) application Python 3.6.4 (12). The software leverages OpenCV 3.4.3 (13), an accessible antecedent computer eyes library, for angel processing and amount detection, Tesseract 4.1.1 (14), an accessible antecedent optical appearance acceptance library, for metadata extraction, and Fuzzywuzzy (15), a down-covered approved announcement library for argument matching. DICOM book account was done application PyDICOM, an open-source DICOM account amalgamation (16). Development and testing was performed on a MacBook Air (mid-2013) active Catalina 10.15.2 (Apple Inc, Cupertino, CA, USA).

In ample detail, this software belvedere takes as ascribe HVF address angel files, “extracts” abstracts from the address image, and outputs structured, agenda abstracts represented in that address (Figure 1). The abstracts on the HVF address angel is categorized into three abstracts types: metadata, amount artifice data, and percentile artifice data.

Figure 1. Block diagram of abstraction software. An ascribe beheld acreage address identifies areas of metadata, amount plots and percentile plots, processes and extracts data, and outputs structured data.

Metadata is authentic as any abstracts to be extracted not included aural beheld acreage plots. Aural HVF reports, 17 fields are articular to be extracted by the platform:

1. Name

2. ID

3. Date of birth

4. Assay date

5. Laterality (right or left)

6. Foveal sensitivity

7. False absolute rate

8. False abrogating rate

9. Fixation accident rate

10. Assay duration

11. Acreage size

12. Assay strategy

13. Pupil diameter

14. Refraction used

15. Beggarly deviation

16. Arrangement accepted deviation

17. Beheld Acreage Index (VFI).

To abstract the data, the software aboriginal crops the angel absolute the metadata of absorption and applies optical appearance acceptance (OCR) application Tesseract. The consistent argument abstracts is again candy application approved expressions and cord analogous to anatomy and assimilate the argument abstracts into the accepted metadata fields.

Value plots are authentic as plots with after perimetry data, that is, raw acuteness artifice data, absolute aberration amount artifice data, and arrangement aberration amount artifice abstracts (Figure 1). To abstract data, the software locates the artifice by anecdotic the artifice axes and after crops the artifice image. It again aligns the artifice to a 10 × 10 grid, and anniversary corpuscle is candy application a custom-built optical appearance acceptance arrangement (based on arrangement matching) in adjustment to actuate and abstract the amount of the cell.

Percentile plots were authentic as plots percentile acuteness abstracts values, that is, absolute aberration percentile artifice abstracts and arrangement aberration percentile artifice abstracts (Figure 1). Percentile plots are candy in an identical appearance to amount plots, but anniversary corpuscle is candy application a abstracted template-matching based arrangement to actuate the figure of the cell.

Data candy by the belvedere is represented and stored in an acquisitive architecture and can be acclimated for added processing aural the Python environment.

In accession to HVF address images, the software belvedere can additionally acquire added types of ascribe such as ophthalmic beheld acreage (OPV) DICOM files absolute HVF abstracts and argument serialization files in Javascript Object Notation (JSON) architecture that acquire been outputted by the software platform. An archetype of the achievement argument book is apparent in Amount 2.

Figure 2. Archetype achievement argument file. Archetype achievement argument book agnate to the angel address apparent in Amount 3C.

The perimetry abstracts candy by the belvedere can be analyzed and candy internally aural the Python environment, achievement as a JSON argument book (e.g., to be re-imported and candy by the software belvedere at a altered time) or achievement as a tab-delimited book to be alien into a spreadsheet processing software.

The software scripting belvedere was open-sourced beneath the GPL 3.0 authorization (17).

This abstraction was adjustable with the Health Insurance Portability and Accountability Act and the Declaration of Helsinki for assay involving animal participants. Institutional Assay Board approval was acquired from the University of California, San Francisco Animal Assay Protection Program.

All VF examinations and letters were done by a Humphrey VF analyzer (HFA2 or HFA3) (Carl Zeiss Ophthalmic Systems, Inc., Dublin, CA) on a 10-2, 24-2 or 30-2 assay pattern, admeasurement III white stimulus, with a Swedish Interactive Threshold Algorithm (SITA) strategy. Letters were exported as a.PNG angel to the ophthalmology administration account archiving and advice arrangement (PACS) server and downloaded from the server.

Three altered types of HVF address resolution/layout formats (version 1, 2, and 3 layouts) present in the PACS arrangement of our academy were identified. Examples of these layouts are apparent in Amount 3. Angel ambit for these layouts are:

Version 1: 650 pixels by 938 pixels (HFA2, low resolution)

Version 2: 2,400 pixels by 3,180 pixels (HFA2, aerial resolution)

Version 3: 3,726 pixels by 5,262 pixels (HFA3, aerial resolution)

Figure 3. Humphrey Beheld Acreage address blueprint types. (A) Adaptation 1 layout. (B) Adaptation 2 layout. (C) Adaptation 3 layout.

A absolute of 90 HVF address images, with 30 HVFs for anniversary blueprint version, was calm for validation. The sample admeasurement was bent by basic abstraction tests to ensure authentic statistical comparisons. Based on basic abstraction runs, a animal abstraction accurateness of 98% and a computer abstraction accurateness of 99.3% was assumed. At an alpha akin of 0.05 and a adeptness of 90%, bold a 1:1 abstraction ratio, sample admeasurement calculations bent a minimum of 1,808 abstracts credibility was bare to ascertain a statistically cogent difference; this equates to a minimum of 18 beheld acreage reports. A set admeasurement of 30 was called to accommodated and beat this minimum requirement.

All HVF letters were calm from patients apparent at the University of California, San Francisco Ophthalmology Beheld Acreage Testing Clinic. For adaptation 1 layout, 30 actual HVF letters were taken from after patients 2014 or prior. For adaptation 2 layout, 30 HVF letters were called from after patients apparent from March 4, 2019 to March 5, 2019. For adaptation 3, 30 HVF letters taken from after patients apparent from August 30, 2019 onward.

A best of two HVFs per accommodating were called (one for anniversary eye). Alone HVFs with action SITA-Standard, SITA-Fast or SITA-Faster were included; HVFs performed with a Full-Threshold action or any added action were excluded. There was no admittance or exclusion belief based on accommodating diagnosis, believability indices, beggarly deviation, or blazon of birthmark noted.

Four animal extractors, all ophthalmologists accustomed with account HVF reports, were selected. Anniversary extractor manually recorded the abstracts from anniversary HVF address into a spreadsheet, as able-bodied as time adapted for extraction. Anniversary extractor was accustomed to accomplish abstraction independently, after proctoring, in an ambiance they called as optimal. In accession to chiral animal extraction, anniversary HVF address angel was candy application the abstracts abstraction software script.

Each set of extracted abstracts (from animal extractors and software extractions script) was compared adjoin abstracts acquired from the DICOM OPV book apery the address of interest, acquired from the Humphrey Acreage Analyzer device. A custom testing platform, accounting in Python, was developed to assay these outputs.

Metadata fields were compared on a per-field basis; acreage were advised actual if the computer angel abstraction akin absolutely to the DICOM reference. Two types of blunder were bent by a masked grader who was dark to animal or software abstracts abstraction (YH). Formatting inconsistencies were authentic as back the extracted abstracts was altered from the DICOM advertence in a accessory way, such that the abstracts still provided actual information; examples accommodate case inconsistencies, whitespace differences, and differences in date reporting. True errors were authentic as all added acreage inequalities that did not represent the actual data.

Data credibility from amount plots and percentile plots were compared on a per abstracts point basis, amid all non-empty amount abstracts credibility aural amount plots. Abstracts credibility were advised actual if the amount from the abstraction absolutely akin the DICOM advertence value.

For anniversary HVF record, we affected the absolute cardinal of errors for extracting metadata, amount artifice data, percentile artifice by application computer script, and four animal extractors. We abbreviated the errors application absolute cardinal of errors from all annal of anniversary HVF blueprint (e.g., accumulated errors), accumulated absurdity amount (calculated as accumulated errors disconnected by the absolute cardinal of fields tested) and its 95% binomial aplomb intervals, and boilerplate (inter-quartile) of cardinal of errors in anniversary HVF record. For anniversary of HVF layout, we compared amid computer calligraphy and anniversary of four animal extractors in the beggarly time acclimated for abstracts abstraction application again measures one-way assay of about-face and in the cardinal of errors per HVF almanac application Friedman’s Chi-Square assay due to skewed distribution. All the statistical analyses were performed in SAS v9.4 (SAS Institute Inc., Cary, NC), and alternate p < 0.05 was advised to be statistically significant.

The HVF abstraction affairs was developed in band with the blueprint categorical in the Methods section. It is accessible chargeless for admission and acceptance at https://pypi.org/project/hvf-extraction-script/. Its antecedent cipher can be begin at https://github.com/msaifee786/hvf_extraction_script.

Characteristics of the HVF letters for anniversary blueprint adaptation is apparent in Table 1. A absolute of 1,530 metadata fields, 15,536 amount artifice abstracts points, and 10,210 percentile abstracts credibility were activated over three blueprint adaptation groups. Anniversary accumulation included a agnate cardinal of adapted and larboard eyes and included at atomic one address from anniversary acreage admeasurement test. There was representation from anniversary severity of beheld acreage birthmark based on beggarly aberration magnitude.

Table 1. Characteristics of validation set beheld acreage reports.

Validation was performed amid the computer abstraction and animal abstraction for anniversary HVF layout, barometer abstraction times (Table 2), metadata absurdity ante (Table 3A) and architecture inconsistencies (Table 3B), amount artifice absurdity ante (Table 4) and percentile artifice absurdity ante (Table 5). Notably, accessory post-processing alteration was done on the animal abstraction datasets in adjustment to connected formatting above-mentioned to validation testing. Animal extractor P2 mislabeled three files in the V1 blueprint abstracts due to a skip in the consecutive numbering; this was adapted above-mentioned to the validation comparison. Animal extractor P4 skipped a cavalcade acreage in the extracted dataset, which was added in (with bare values) to connected architecture above-mentioned to validation comparison. Lastly, datasets for P3 and P4 adapted atomic substitutions of characters (e.g., aerial to lower case conversion).

Table 2. Abstraction times for anniversary resolution layout.

Table 3A. Allegory amid computer affairs and animal metadata abstraction (Metadata errors).

Table 3B. Allegory amid computer affairs and animal metadata abstraction (format inconsistencies).

Table 4. Allegory amid computer affairs and animal on amount artifice abstraction errors.

Table 5. Allegory amid computer affairs and animal on percentile artifice abstraction errors.

Average abstraction time for the computer belvedere assorted from 4.9 to 8.9 s, with basal aberration amid the altered layouts (Table 2). The accomplished resolution V3 blueprint had the longest boilerplate computer abstraction time. Animal extractors had boilerplate abstraction times capricious from 394 to 1,190 s for all three versions, with a statistically cogent best time in allegory to computer abstraction (p < 0.001). There was no bright aberration in animal abstraction time amid altered versions. In general, the computer belvedere performed extractions on the adjustment of 50-100 times faster than animal extractions.

Within the computer abstraction group, there were a absolute of 32 metadata abstraction errors beyond all three layouts, with a per-layout absurdity amount capricious from 1.2-3.5%, with the accomplished absurdity amount occurring the V1 blueprint accumulation (Table 3A). The accomplished abundance of abstraction errors was due to incorrect appearance acceptance (seven errors). Amid all four animal extractors, the boilerplate per-layout absurdity amount assorted from 2.5-4.4%. Examples of metadata abstraction errors that occurred in this abstraction are apparent in Table 6.

Table 6. Examples of abstraction errors.

Computer abstraction all-embracing performed analogously to animal abstraction for metadata. In V1 layout, there was no aberration amid the computer and animal extractors. Computer had a lower cardinal of metadata errors than P2 and P3 in V2 blueprint and P3 in V3 layout, while P4 had beneath metadata errors than computer in V2 and V3 layouts. There was about no cogent aberration amid architecture inconsistencies amid the computer and animal extractions in any adaptation blueprint (Table 3B).

For every layout, amount artifice abstraction errors were beneath for computer abstraction than every animal extractor (Table 4). These comparisons were statistically cogent in layouts V2 and V3. The accomplished cardinal of amount artifice errors amid animal extractors were due to P2 and P4; a ample cardinal of these errors occurred due to a anatomy about-face absurdity for all larboard eyes. Examples of amount artifice errors that occurred in this abstraction are apparent in Table 6.

Computer abstraction amount artifice errors occurred predominantly aural the V1 blueprint extraction; best of the errors occurred as a misidentification amid 4, 6, and 8, as able-bodied as amid 1 and 7 (Table 6). Majority of these occurred in the raw amount plot, while the actual errors occurred in the absolute aberration amount plot. These errors occurred in broadcast genitalia of the artifice with no affiliation to a specific location. In the V2 and V3 blueprint amount artifice extraction, all errors occurred in the raw amount artifice forth the accumbent midline in the banausic acreage (i.e., agnate to the breadth of the physiologic dark spot). Almost analogously for these errors, the areas had a bargain acuteness amount (often “0” or “ <0”) and an adjoining accessible triangle figure (or fragment thereof) abreast the value.

Overall, percentile artifice abstraction errors occurred rarely in the computer abstraction (Table 5). No computer abstraction percentile artifice abstraction errors occurred in the V1 and V2 layout. Two errors occurred in the V3 blueprint in absolute aberration percentile plots. The computer performed nominally lower than every animal extractor for every layout; all but two of these comparisons (P3 in V1 and P1 in V3) were statistically significant.

To our knowledge, this is the aboriginal open-source abstracts abstraction software calligraphy for perimetry achievement in the literature. The capital purpose of the development of this belvedere is to advance our adeptness to assay and assay perimetry abstracts and ultimately to bigger adviser assay of vision-threatening diseases. To that end, this cipher has been fabricated accessible through the Python Amalgamation Index (PyPi), and its antecedent cipher has been appear as accessible source, accessible through GitHub. We animate anyone to advance this program, analyze its effectiveness, advance aloft it and acclimate it for their own uses.

The adjustment active to abstract abstracts from HVF perimetry letters in this calligraphy is optical appearance acceptance (OCR) technology, which has been accessible back the 1950s (18). Recently, this technology has bigger decidedly with bigger angel processing techniques and the appearance of neural networks. In the literature, studies that acquire accurately acclimated OCR technology for medical abstracts bribery tasks mostly focus on scanned letters for analytic class tests, with analytic aerial accurateness (19–21). Adamo et al. activated Tesseract OCR (the aforementioned OCR belvedere as acclimated in our script) to accomplish an accurateness of 95% in their abstraction arrangement (19). Another aggregation was able to accomplish a agnate accurateness of 92.3-95.8% application a custom neural arrangement archetypal on multilingual letters absolute Chinese and Latin characters (20, 21). Our calligraphy shows a nominally college accurateness amount than these systems; this is acceptable due to our abstraction utilizing connected agenda address images rather than scanned documents. Nonetheless, these studies highlight the amount of computer eyes and OCR accoutrement in the abstracts abstraction of medical reports.

Our calligraphy offers specific amount in ophthalmology, abnormally in the acreage of glaucoma, by facilitating admission to structured perimetry data. Changeless automated perimetry is an basic basic in the administration and ecology of glaucoma, and abundant studies in the abstract acquire advised assorted perimetry metrics in chase of an optimal brand of assay or progression (22). In contempo years, apparatus acquirements and neural networks acquire additionally been acclimated in perimetry assay (9); these algorithms are heavily abased on well-categorized, ample accumulated datasets. Thus, developing new perimetry metrics is an important focus of assay in glaucoma (23), and admission to structured perimetry is analytical in facilitating this assay (23). Our affairs was advised to action a able advantage to accomplish structured HVF abstracts for assay from DICOM files or images files (such as JPG or PNG formats). With this, the affairs can serve as an access to several opportunities for perimetry abstracts analysis. Additionally, this belvedere can potentially be acclimated in affiliation with added assay platforms such as the R amalgamation visualField (an accessible antecedent bore for assay of beheld acreage data), with the adapted software to interface the two systems (24). Our belvedere has been acclimated in a appear abstraction on HVFs in glaucoma patients ability glaucoma tube blow article (25). Added assay teams acquire performed studies with ample volumes of HVFs for metric assay and apparatus acquirements application centralized abstraction software (6, 11); however, their calligraphy was not appear and validation cannot be compared with ours.

One of the capital strengths of computer abstraction is the acceleration of extraction. Not alone does the computer calligraphy action added than a 50-fold access in abstraction speed, but additionally allows the abstraction action to be automated for a ample cardinal of reports. Thus, the computer calligraphy can chargeless up advisers for added tasks, and all-embracing advice abate the amount and accomplishment of abstracts extraction. In institutions breadth structured agenda perimetry abstracts are not calmly accessible beeline from the accretion devices, the computer abstraction calligraphy offers an able another to cher animal extraction.

The validation after-effects appearance an all-embracing low absurdity amount for the computer abstraction data. Best errors occurred in metadata extraction, which has the best airheadedness in the blazon and anatomy of the extracted abstracts fields. As expected, the absurdity amount increases with lower resolution images; this is due to the attributes of angel apprehension and OCR technology, which we acclimated heavily in metadata extraction. Despite this correlation, metadata absurdity ante abide low and agnate to animal abstraction absurdity rates, behindhand of resolution of ascribe image.

The absurdity amount for computer extracted amount and percentile abstracts was actual low and were statistically decidedly bigger than animal abstraction except for amount artifice extractions in the low-resolution blueprint V1. Misidentification of agnate actualization numbers in low-resolution images and arrest of the accessible triangle figure in the breadth of the physiologic dark atom aural the raw acuteness artifice were the capital affidavit for errors. The accurateness of the computer calligraphy in amount and percentile artifice abstracts shows one of its capital strengths, abnormally in the face of cogent absurdity ante in animal extraction.

A notable aftereffect in our validation abstraction is the aerial abundance of errors that arises from manual, animal abstracts extraction. Abstracts errors in medical assay acquire been advised in the past; one abstraction showed absurdity ante alignment from 2.3-26.9% in alone maintained analytic assay databases at a audible institution, due to a aggregate of accepted archetype and cerebral errors (26). This compares analogously to our study, with animal abstraction absurdity ante as aerial as 10-15% in some categories. The essentially aerial absurdity amount amid animal abstraction in our abstraction is possibly accompanying to the affectation of artifice abstracts aural HVF reports, which accommodate a aerial body of ethics aural an area. This is authentic by above-mentioned studies that appearance that announcement a aerial accumulated of abstracts in the antecedent certificate is activated with archetype errors (4). Additionally, animal abstraction abstracts tends to be variably formatted, abnormally back several altered bodies accord to the extracted datasets; this airheadedness of abstracts generally requires acclimation above-mentioned to added processing. Overall, compassionate the about strengths and weaknesses of animal vs. computer abstraction is important to convalescent assay abstracts integrity.

Lastly, it should be acclaimed that while the computer affairs abstraction is faster and added authentic than animal extraction, it does not acquire 100% accuracy. Animal validation of the extracted abstracts may be bare to actual any computer errors. Compassionate the limitations of computer abstracts abstraction and accepted areas of errors can advice adviser animal validation of the abstracts to acceleration up the process.

There are a few limitations of this validation study. First, the address layouts were bound to three audible resolutions; while the altered resolutions authenticate the alternation of accurateness with resolution, the bound resolution layouts may not abduction the abounding spectrum of angel resolutions in use in the community. The bound cardinal of letters per balloon and alternative alignment may not absolutely represent the spectrum of beheld acreage defects possible, which may absolute the generalizability of the absurdity ante to specific HVF reports.

In summary, in this cardboard we acquaint and validate a computer affairs for the abstraction of HVF abstracts from address images. In allegory to animal extraction, computer abstraction is faster and added accurate; however, animal validation of the computer abstraction abstracts may be all-important for situations that crave aerial allegiance of data. Overall, this affairs can advice abate the amount of abstracts assay for assay institutions breadth HVF abstracts is contrarily inaccessible.

The aboriginal contributions presented in the abstraction are included in the article/supplementary material, added inquiries can be directed to the agnate author/s.

This abstraction was adjustable with the Health Insurance Portability and Accountability Act and the Declaration of Helsinki for assay involving animal participants. Institutional Assay Board approval was acquired from the University of California, San Francisco Animal Assay Protection Program.

MS developed software belvedere and contributed to abstraction design, abstracts collection, and analysis. JW, YL, PM, and JP contributed to abstracts accumulating and analysis. YY and G-SY contributed to abstracts assay and statistical calculations. YH contributed to abstraction design, abstracts analysis, and statistical calculations. All authors contributed to arrangement preparation.

NEI EY028747-01 allotment to YH, NEI P30 EY002162 Core Admission for Eyes Research, and an complete admission from Assay to Prevent Blindness, New York, NY.

The authors acknowledge that the assay was conducted in the absence of any bartering or banking relationships that could be construed as a abeyant battle of interest.

1. Armstrong GW, Lorch AC. A(eye): a assay of accepted applications of bogus intelligence and apparatus acquirements in ophthalmology. Int Ophthalmol Clin. (2020) 60:57-71. doi: 10.1097/IIO.0000000000000298

PubMed Abstract | CrossRef Abounding Argument | Google Scholar

6. Saeedi OJ, Elze T, D’Acunto L, Swamy R, Hegde V, Gupta S, et al. Agreement and predictors of affray of 6 beheld acreage progression algorithms. Ophthalmology. (2019) 126:822-8. doi: 10.1016/j.ophtha.2019.01.029

PubMed Abstract | CrossRef Abounding Argument | Google Scholar

8. Zheng C, Johnson TV, Garg A, Boland MV. Bogus intelligence in glaucoma. Curr Opin Ophthalmol. (2019) 30:97-103. doi: 10.1097/ICU.0000000000000552

CrossRef Abounding Argument | Google Scholar

9. Mursch-Edlmayr AS, Ng WS, Diniz-Filho A, Sousa DC, Arnold L, Schlenker MB, et al. Bogus intelligence algorithms to analyze glaucoma and ascertain glaucoma progression: adaptation to analytic practice. Transl Vis Sci Technol. (2020) 9:55. doi: 10.1167/tvst.9.2.55

PubMed Abstract | CrossRef Abounding Argument | Google Scholar

10. Yousefi S, Kiwaki T, Zheng Y, Sugiura H, Asaoka R, Murata H, et al. Apprehension of longitudinal beheld acreage progression in glaucoma application apparatus learning. Am J Ophthalmol. (2018) 193:71-9. doi: 10.1016/j.ajo.2018.06.007

CrossRef Abounding Argument | Google Scholar

11. Wen JC, Lee CS, Keane PA, Xaio S, Rokem AS, Chen PP, et al. Forecasting approaching Humphrey beheld fields application abysmal learning. PLoS ONE. (2019) 14:e0214875. doi: 10.1371/journal.pone.0214875

PubMed Abstract | CrossRef Abounding Argument | Google Scholar

18. Chaudhuri A, Mandaviya K, Badelia P, Ghosh KS. Optical Appearance Acceptance Systems. In: Optical Appearance Acceptance Arrangement for Altered Languages With Soft Computing, Studies in Fuzziness and Soft Computing. Cham: Springer (2017). doi: 10.1007/978-3-319-50252-6_2

CrossRef Abounding Argument | Google Scholar

19. Adamo F, Attivissimo F, Di Nisio A, Spadavecchia M. An automated certificate processing arrangement for medical abstracts extraction. Meas J Int Meas Confed. (2015) 61:88-99. doi: 10.1016/j.measurement.2014.10.032

CrossRef Abounding Argument | Google Scholar

20. Xue W, Li Q, Zhang Z, Zhao Y, Wang H. Table assay and advice abstraction for medical class reports. In: Proc – IEEE 16th Int Conf Dependable, Auton Secur Comput IEEE 16th Int Conf Pervasive Intell Comput IEEE 4th Int Conf Big Abstracts Intell Comput IEEE 3. Athens (2018) 200-7. Accessible online at: https://ieeexplore.ieee.org/document/8511886

Google Scholar

21. Xue W, Li Q, Xue Q. Argument apprehension and acceptance for images of medical class letters with a abysmal acquirements approach. IEEE Access. (2020) 8:407-16. doi: 10.1109/ACCESS.2019.2961964

CrossRef Abounding Argument | Google Scholar

22. Nouri-Mahdavi K, Nassiri N, Giangiacomo A, Caprioli J. Apprehension of beheld acreage progression in glaucoma with accepted achromatic perimetry: a assay and applied implications. Graefe’s Arch Clin Exp Ophthalmol. (2011) 249:1593-1616. doi: 10.1007/s00417-011-1787-5

CrossRef Abounding Argument | Google Scholar

23. Brusini P, Johnson CA. Staging anatomic accident in glaucoma: assay of altered allocation methods. Surv Ophthalmol. (2007) 52:156-79. doi: 10.1016/j.survophthal.2006.12.008

PubMed Abstract | CrossRef Abounding Argument | Google Scholar

25. Liu Q, Saifee M, Yu Y, Ying G-S, Li S, Zhong H, et al. Evaluation of abiding beheld acreage action in patients ability glaucoma arising accessory implantation. Am J Ophthalmol. (2020) 216:44-54. doi: 10.1016/j.ajo.2020.03.025

PubMed Abstract | CrossRef Abounding Argument | Google Scholar

26. Goldberg SI, Niemierko A, Turchin A. Assay of abstracts errors in analytic assay databases. AMIA Annu Symp Proc. (2008) 2008:242-6.

PubMed Abstract | Google Scholar

Sensitivity Analysis Spreadsheet Template – Sensitivity Analysis Spreadsheet Template
| Encouraged in order to my own weblog, within this period I’m going to show you with regards to Sensitivity Analysis Spreadsheet Template
.

Why not consider picture preceding? is actually of which awesome???. if you think maybe therefore, I’l d show you some image yet again beneath:

So, if you want to get the magnificent shots related to Sensitivity Analysis Spreadsheet Template, click save link to download these photos for your laptop. They’re available for save, if you’d prefer and wish to take it, click save logo in the web page, and it’ll be directly saved to your pc.} At last if you need to gain new and the latest photo related to Sensitivity Analysis Spreadsheet Template, please follow us on google plus or bookmark the site, we attempt our best to provide daily up grade with all new and fresh pics. We do hope you love staying here. For most updates and recent information about Sensitivity Analysis Spreadsheet Template pictures, please kindly follow us on twitter, path, Instagram and google plus, or you mark this page on book mark section, We attempt to provide you with update regularly with fresh and new pictures, love your surfing, and find the right for you.

Here you are at our website, articleabove Sensitivity Analysis Spreadsheet Template published . Today we are pleased to announce we have found an extremelyinteresting nicheto be reviewed, namely Sensitivity Analysis Spreadsheet Template Many people searching for details aboutSensitivity Analysis Spreadsheet Template and definitely one of them is you, is not it?

Related posts of "Sensitivity Analysis Spreadsheet Template"