카테고리 없음

[논문리뷰] Comparison of two rapid automated analysis tools for large FTIR microplastic datasets

한국분석과학연구소 2023. 3. 24. 10:38

FPA-FT-IR을 사용하여 미세플라스틱을 분석하는 분광기에서 자동화된 SW는 매우 중요합니다. 관련하여 범용적으로 사용되는 SW에 대한 평가 결과를 논문으로 발간된 자료가 있어서 공유드립니다.

Comparison of two rapid automated analysis tools for large FTIR microplastic datasets

대규모 FTIR을 위한 두 가지 신속한 자동 분석 도구 비교 미세플라스틱 데이터 세트

Analytical and Bioanalytical Chemistry

https://doi.org/10.1007/s00216-023-04630-w

One of the biggest issues in microplastic (MP, plastic items <5 mm) research is the lack of standardisation and harmonisation in all felds, reaching from sampling methodology to sample purifcation, analytical methods and data analysis.

This hampers comparability as well as reproducibility among studies. Concerning chemical analysis of MPs, Fourier-transform infrared (FTIR) spectroscocopy is one of the most powerful tools. Here, focal plane array (FPA) based micro-FTIR (µFTIR) imaging allows for rapid measurement and identifcation without manual preselection of putative MP and therefore enables large sample throughputs with high spatial resolution. The resulting huge datasets necessitate automated algorithms for data analysis in a reasonable time frame. Although solutions are available, little is known about the comparability or the level of reliability of their output. For the frst time, within our study, we compare two well-established and frequently applied data analysis algorithms in regard to results in abundance, polymer composition and size distributions of MP (11–500 µm) derived from selected environmental water samples: (a) the siMPle analysis tool (systematic identifcation of MicroPlastics in the environment) in combination with MPAPP (MicroPlastic Automated Particle/fbre analysis Pipeline) and (b) the BPF (Bayreuth Particle Finder). The results of our comparison show an overall good accordance but also indicate discrepancies concerning certain polymer types/clusters as well as the smallest MP size classes. Our study further demonstrates that a detailed comparison of MP algorithms is an essential prerequisite for a better comparability of MP data.

미세플라스틱(MP, 플라스틱 제품 <5 mm) 연구에서 가장 큰 문제 중 하나는 샘플링 방법론에서 샘플 정제, 분석 방법 및 데이터 분석에 이르는 모든 분야에서 표준화 및 조화가 부족하다는 것입니다.

이는 연구 간의 재현성뿐만 아니라 비교 가능성을 방해합니다. MP의 화학적 분석과 관련하여 푸리에 변환 적외선(FTIR) 분광법은 가장 강력한 도구 중 하나입니다. 여기에서 초점면 배열(FPA) 기반 마이크로 FTIR(µFTIR) 이미징을 통해 추정 MP를 수동으로 사전 선택하지 않고도 신속한 측정 및 식별이 가능하므로 높은 공간 분해능으로 많은 샘플 처리량이 가능합니다. 그 결과 거대한 데이터 세트는 합리적인 시간 프레임에서 데이터 분석을 위한 자동화된 알고리즘을 필요로 합니다. 솔루션을 사용할 수 있지만 출력의 비교 가능성 또는 신뢰성 수준에 대해서는 알려진 바가 거의 없습니다. 처음으로 우리 연구에서 우리는 선택된 환경 물 샘플에서 파생된 MP(11–500 µm)의 존재비, 폴리머 구성 및 크기 분포의 결과와 관련하여 잘 확립되고 자주 적용되는 두 가지 데이터 분석 알고리즘을 비교합니다. ) MPAPP(MicroPlastic Automated Particle/fbre analysis Pipeline) 및 (b) BPF(Bayreuth Particle Finder)와 결합된 단순 분석 도구(환경에서 MicroPlastics의 체계적 식별). 비교 결과는 전반적으로 일치하지만 특정 폴리머 유형/클러스터 및 가장 작은 MP 크기 클래스와 관련된 불일치도 나타냅니다. 우리의 연구는 MP 알고리즘의 상세한 비교가 MP 데이터의 더 나은 비교 가능성을 위한 필수 전제 조건임을 보여줍니다.

두 가지 FPA-µFTIR 데이터 분석 알고리즘

www.simple-plastics.eu

: 무료로 액세스 가능)

이미지 분석 도구 MPAPP(MicroPlastic Automated Particle/fibre analysis Pipeline)와 함께 Primpke 등이 게시한 스크립트 포함.

다운로드 가능: https://drive.google.com/drive/folders/1fWIGp7MgJZJcy7NWI5Vri0eUYJw0Qvrz?usp=share_link)

Bayreuth Particle Finder(BPF) Hufnagl et al.

: Epina ImageLab Engine(www.imagelab.at)의 통합 모듈

#Purency GmbH에서 개발한 Purency Microplastics Finder의 예비 버전

상업적으로 이용 가능합니다(https://www.purency.ai/microplastics-finder).

두 접근 방식의 주요 차이점은 siMPle이 인스턴스 기반 기계 학습 접근 방식이며 두 가지 유사성 측정을 사용하는 이중 데이터베이스 검색에 의존한다는 것

HQI는 Pearson 상관관계[48]를 통해 계산

반면에 BPF는 모델 기반 기계 학습 접근 방식

여기서는 스펙트럼 디스크립터와 RDF(Random Decision Forest) 분류기의 조합이 적용

두 파이프라인 모두 필터의 무작위로 미리 선택된 하위 영역에서 얻은 결과를 외삽하는 동안 발생할 수 있는 편향을 피하면서 전체 필터의 분석을 허용

다양한 환경 매트릭스에서 MP를 분석하는 다양한 연구에 자주 적용

예를 들어 #BPF 는 Frei et al.과 같은 담수 환경에서 MP 오염에 초점을 맞춘 다양한 연구에 적용

#siMPle 은 또한 광범위한 응용 분야를 찾습니다. 음용수[61], 폐수[62] 또는 해수[63]의 분석에서. MPAPP 또는 그 이전 버전 APA(자동 입자 분석, Primpke et al.[2] 이후)와 함께 siMPle은 강 지표수[64, 65], 폐수 방류수[53, 66] 및 심해에 대한 최근 연구에 적용

Efects of MP size on spectral quality and automated analysis

스펙트럼 품질 및 자동 분석에 대한 MP 크기의 영향

This study provides a detailed comparison of two MP analysis pipelines, i.e. siMPle/MPAPP and BPF. It is to be noted that this comparison was conducted with environmental samples, containing an unknown amount and composition of MP items. Thus, it remains unclear which results of the aforementioned analysis pipelines are closer to the actual MP occurrence. However, this approach was chosen to shed some light on the complexity of the analysis of environmental samples. Working with internal standards or spiked

samples has multiple limitations in terms of representativeness due to the limited availability of commercially available reference material in regard to polymer types, shapes and sizes. Additionally, the preparation of spiked MP samples is challenging, especially using S-MP<100 µm [70].

As this size range, however, appears to be the most abundant in environmental samples and poses an increased ecotoxicological risk, it is especially important to represent this size class. The aim of our study, however, was to demonstrate the complexity of the analysis of environmental samples in its full spectrum, containing all relevant polymer types, shapes and sizes, as well as the interplay of matrix residues with aged microplastic being the real challenge for both algorithms.

Furthermore, a well-established approach in other domains of chemometrics for comparing algorithms is the use of expert-annotated training data [71, 72]. However, within the considered size range, experimental difculties (e.g. the handling of small virgin MP items) make it very challenging to ensure a correct assignment of the ground truth of spectra, which is why, as a frst step, this study was

limited to a relative comparison of results. Nevertheless, our study highlights the similarities and diferences in results obtained with both tools which are essential for further eforts towards method optimisation and harmonisation.

Despite of the use of state-of-the-art methods, our results, however, also underline the uncertainties in MP analysis concerning polymer identifcation, especially at the lower end of the detection limit in terms of size. Depending on morphology and thickness of MPs in a size range below 50 µm, IR-spectra can be infuenced by efects such as difraction and Mie scattering [42], in tandem with low intensity of

the original polymer signal. Thus, the interplay of these phenomena potentially results in low quality spectra and low signal-to-noise ratios. Due to a more conservative, timeintensive approach with BPF, possibly such low-quality polymer spectra were manually rejected. While for sample set A between 17 and 50% (mean±SD=37%±9%) of all items identifed as MP were accepted after manual re-inspecting

the data, only between 0 and 39% (mean±SD=13%±15%) of all MP hits were accepted for sample set B. For siMPle/MPAPP, such a process was performed former to this study in previous work on surface water samples collected in the North Sea [69], setting the minimum threshold to be reached per polymer type. Nevertheless, during application of siMPle/MPAPP, a fraction of such low quality spectra may have been included automatically. Although the problem of classifcation of MPs with low quality spectra at the lower end of the detection limit of IR spectroscopy has been demonstrated in our case on the example of two automated analysis algorithms, it also holds true for other automated classifcation solutions or pure manual classifcation. The severity of this phenomenon may difer between studies as it also depends on the specifcations of the respective µFTIR system, flter type etc. used for MP sample measurement.

이 연구는 두 개의 MP 분석 파이프라인, 즉 siMPle/MPAPP 및 BPF에 대한 자세한 비교를 제공합니다. 이 비교는 MP 항목의 양과 구성을 알 수 없는 환경 샘플을 사용하여 수행되었다는 점에 유의해야 합니다. 따라서 앞서 언급한 분석 파이프라인의 어떤 결과가 실제 MP 발생에 더 가까운지 명확하지 않습니다. 그러나 이 접근법은 환경 샘플 분석의 복잡성을 밝히기 위해 선택되었습니다. 내부 표준 또는 스파이크 작업

샘플은 폴리머 유형, 모양 및 크기와 관련하여 상업적으로 이용 가능한 참조 자료의 제한적인 가용성으로 인해 대표성 측면에서 여러 가지 제한이 있습니다. 또한, 특히 S-MP<100μm[70]를 사용하여 스파이킹된 MP 샘플을 준비하는 것은 어렵습니다.

그러나이 크기 범위는 환경 샘플에서 가장 풍부한 것으로 나타나고 생태 독성 위험이 증가하므로 이 크기 등급을 나타내는 것이 특히 중요합니다. 그러나 우리 연구의 목적은 모든 관련 폴리머 유형, 모양 및 크기를 포함하는 전체 스펙트럼에서 환경 샘플 분석의 복잡성을 입증하고 매트릭스 잔류물과 노화된 미세 플라스틱의 상호 작용이 진정한 도전 과제임을 입증하는 것이었습니다. 두 알고리즘.

또한, 알고리즘을 비교하기 위한 화학 측정의 다른 영역에서 잘 확립된 접근 방식은 전문가 주석 훈련 데이터를 사용하는 것입니다[71, 72]. 그러나 고려된 크기 범위 내에서 실험적 어려움(예: 작은 처녀 MP 항목 취급)으로 인해 스펙트럼의 실측값을 올바르게 지정하는 것이 매우 어렵습니다.

결과의 상대적인 비교로 제한됩니다. 그럼에도 불구하고 우리의 연구는 방법 최적화 및 조화를 향한 추가 노력에 필수적인 두 도구로 얻은 결과의 유사점과 차이점을 강조합니다.

그러나 최첨단 방법의 사용에도 불구하고 우리의 결과는 특히 크기 측면에서 검출 한계의 하한에서 폴리머 식별과 관련된 MP 분석의 불확실성을 강조합니다. 50μm 미만의 크기 범위에 있는 MP의 형태와 두께에 따라 IR 스펙트럼은 낮은 강도와 함께 회절 및 Mie 산란[42]과 같은 영향을 받을 수 있습니다.

원래 폴리머 신호. 따라서 이러한 현상의 상호 작용은 잠재적으로 낮은 품질의 스펙트럼과 낮은 신호 대 잡음비를 초래합니다. BPF에 대한 보다 보수적이고 시간 집약적인 접근 방식으로 인해 이러한 저품질 폴리머 스펙트럼이 수동으로 거부되었을 수 있습니다. 샘플 세트 A의 경우 MP로 식별된 모든 항목의 17~50%(평균±SD=37%±9%)가 수동 재검사 후 승인되었습니다.

데이터에서 모든 MP 적중의 0 ~ 39%(평균±SD=13%±15%)만이 샘플 세트 B에 대해 허용되었습니다. siMPle/MPAPP의 경우 이러한 프로세스는 표면에 대한 이전 작업에서 이 연구 이전에 수행되었습니다. 북해에서 수집한 물 샘플[69], 폴리머 유형별로 도달해야 하는 최소 임계값 설정. 그럼에도 불구하고 siMPle/MPAPP를 적용하는 동안 이러한 저품질 스펙트럼의 일부가 자동으로 포함되었을 수 있습니다. IR 분광법의 검출 한계 하한에서 낮은 품질의 스펙트럼으로 MP를 분류하는 문제가 우리의 경우 두 가지 자동 분석 알고리즘의 예에서 입증되었지만 다른 자동 분류 솔루션 또는 순수한 수동 분류에도 적용됩니다. . 이 현상의 심각도는 MP 샘플 측정에 사용되는 각각의 μFTIR 시스템, 필터 유형 등의 사양에 따라 달라지므로 연구마다 다를 수 있습니다.

Potential reasons for diferences in MP abundance, polymer composition and size distributions

MP 풍부도, 폴리머 구성 및 크기 분포의 차이에 대한 잠재적인 이유

Future implications

By comparing two currently well-established and frequently applied MP analysis tools, this study can act as a basis for future harmonisation and standardisation eforts in MP analysis. In general, BPF and siMPle/MPAPP showed similar results, with some discrepancies likely caused by matrix efects, and others explainable by the chemical characteristics of certain polymers which could be improved by a broader measurement range including the fngerprint region. On the whole, both pipelines are rapid and generate a detailed data output and therefore show great potential for a broad application in MP assessments.

This study further underlines the importance of QA/QC, e.g. mplemented by manual counter-checking by experts, in order to allow for the generation of high-quality datasets and underlines the importance of purifcation approaches that reduce the present sample matrix efectively.

Additionally, our study also shows that all studies on MP contamination should be interpreted with caution, especially with respect to smaller size classes, since it remains unclear for all currently applied methods how correct the generated results are with respect to the actual occurrence of MP in the environment. Keeping this in mind, as a fnal consequence, we have to admit that even by the use of state-of-the-art methodology, the determination of the real environmental MP number is still a challenge which needs to be addressed by further research eforts. In order to have a clearer picture of how close the obtained results are to the actual numbers, it may perhaps be benefcial to work with spiked samples — although one

has to be aware of the limitations in regard to available polymer types, shapes and sizes. Nevertheless, through the current ongoing development and improvement of the here applied analysis tools, both usability and reliability are being enhanced, e.g. by adaptations of underlying reference databases (siMPle) or optimised follow-up versions (“Purency Microplastic Finder” derived from BPF). These improvements and further optimisations will lead to analysis tools

that — in the best case — produce data with high reliability without additional manual re-evaluation eforts.

향후 영향

본 연구는 현재 잘 정립되고 자주 적용되는 두 가지 MP 분석 도구를 비교함으로써 MP 분석에서 향후 조화 및 표준화 노력의 기초가 될 수 있습니다. 일반적으로 BPF와 siMPle/MPAPP는 유사한 결과를 보였는데, 약간의 불일치는 매트릭스 효과로 인해 발생할 수 있으며, 다른 일부는 지문 영역을 포함한 더 넓은 측정 범위로 개선될 수 있는 특정 폴리머의 화학적 특성으로 설명할 수 있습니다. 전반적으로 두 파이프라인은 빠르고 자세한 데이터 출력을 생성하므로 MP 평가에서 광범위한 응용 프로그램에 대한 큰 잠재력을 보여줍니다.

이 연구는 QA/QC의 중요성을 더욱 강조합니다. 고품질 데이터 세트의 생성을 허용하고 현재 샘플 매트릭스를 효과적으로 줄이는 정화 접근 방식의 중요성을 강조하기 위해 전문가의 수동 카운터 확인으로 구현됩니다.

또한, 우리의 연구는 또한 MP 오염에 대한 모든 연구는 특히 소규모 클래스와 관련하여 주의해서 해석해야 한다는 것을 보여줍니다. 환경. 마지막 결과로 이것을 염두에 두고 최신 방법론을 사용하더라도 실제 환경 MP 수의 결정은 여전히 추가 연구 노력을 통해 해결해야 하는 과제임을 인정해야 합니다. . 얻은 결과가 실제 수치와 얼마나 가까운지 보다 명확하게 파악하려면 스파이크 샘플로 작업하는 것이 도움이 될 수 있습니다.

사용 가능한 폴리머 유형, 모양 및 크기와 관련된 제한 사항을 알고 있어야 합니다. 그럼에도 불구하고 여기에 적용된 분석 도구의 현재 지속적인 개발 및 개선을 통해 유용성과 신뢰성이 모두 향상되고 있습니다. 기본 참조 데이터베이스(siMPle) 또는 최적화된 후속 버전(BPF에서 파생된 "Purency Microplastic Finder")의 조정을 통해. 이러한 개선 및 추가 최적화는 분석 도구로 이어질 것입니다.

최선의 경우 추가적인 수동 재평가 없이 높은 신뢰성으로 데이터를 생성합니다.

Supplementary Information The online version contains supplementary

material available at https://doi.org/10.1007/s00216-023-04630-w.

***************************************************************************

#미세플라스틱 #시험분석 기술서비스 및 #시험표준개발 전문 #국제공인시험기관 (#KOLAS )

- 시험/분석: 환경(수질/토양/대기/폐기물), 식품(음료/벌꿀/주류/소금/어패류/해조류), #이물분석, #식품이물

#세탁폐수 (#미세섬유 )/#세탁망 /#필터, #화장품 /#치약 /#생활화학제품, 각종 #식품용기 (티백/젖병/종이컵 등), #불량분석

- 표준개발: ISO/TC61/SC14, TC38, TC147/SC2&SC6 (Microplastics) Korean Delegate

IEC/TC 111/WG 3 & JWG 14 Co-Convenor

IEC 62321-3-2(Halogen),-10 (PAHs), -13(BPA), -14(SCCP/MCCP) Project leader

- 분석장비: #TED-GC-MS, #micro-FT-IR, micro-#Raman, #ICP-MS, XRF, Combustion-IC 등

02-6951-1116

***************************************************************************