Present status of time-lapse monitoring

Morphology has been the method of embryo assessment for over 40 years and represents the main approach for embryo selection during MAR cycles. However, the standard evaluation at specific time points has limitations, mainly associated with the subjectivity of the embryologist, and the missing critical events potentially harmful to the embryo’s viability. Morphological assessment provides only a snap-shot of embryo development in that specific time, thereby omitting what happens during the intervals between the two observations (Cruz et al., Reference Cruz, Garrido, Herrero, Perez-Cano, Munoz and Meseguer2012). Furthermore, the embryo grade may change notably within a short amount of time (Wong et al., Reference Wong, Loewke, Bossert, Behr, De Jonge, Baer and Reijo Pera2010; Meseguer et al., Reference Meseguer, Herrero, Tejera, Hilligsoe, Ramsing and Remohi2011). In contrast, the TLM allows the embryologist not only to analyze embryo morphology and dynamic changes during the in vitro embryo development, but also provides stable culture conditions (Wong et al., Reference Wong, Loewke, Bossert, Behr, De Jonge, Baer and Reijo Pera2010; Meseguer et al., Reference Meseguer, Rubio, Cruz, Basile, Marcos and Requena2012; Basile et al., Reference Basile, Nogales, Bronet, Florensa, Riqueiros, Rodrigo, Garcia-Velasco and Meseguer2014; Aparicio-Ruiz et al., Reference Aparicio-Ruiz, Basile, Perez, Bronet, Remohi and Meseguer2016).

Although pioneering research on TLM has been reported since 1929 and in the late 1990s (Lewis and Gregory, Reference Lewis and Gregory1929; Payne et al., Reference Payne, Flaherty, Barry and Matthews1997), the technology became commercially available for embryology laboratory only in 2009. The large numbers of published articles on the use of TLM in human embryology suggest an active application of this novel technology in embryology laboratories worldwide. However, at this time, very few data are available on the global use of TLM. Scotland is a unique country, as the government had provided funding to permit all public assisted conception units (NHS-Scotland, National Health Service https://www.scot.nhs.uk) within the state to invest in TLM. In addition to this specific case, few papers are available reporting the worldwide use and implementation of the TLM in MAR treatments. One paper was published by Dolinko and colleagues (Reference Dolinko, Farland, Kaser, Missmer and Racowsky2017), and showed the results of a survey of 294 IVF units in the USA.

The authors reported that only 162 units responded, with 35 laboratories announcing that they used at least one time-lapse system. An analogous report have been published by a French team, Boueilh and co-workers (Reference Boueilh, Reignier, Barrière and Fréour2018). Amongst the 78 respondents, 30 centres reported using TLM clinically. Although these surveys provide interesting information on TLM use in two different countries, it is not enough to make a conclusion of the worldwide use of this technology. It would be really interesting to obtain a global picture of the time-lapse approach to evaluate its current use in IVF practice.

Time-lapse monitoring and embryo assessment: from fertilization to blastocyst formation

The identification of the embryo with the best implantation potential and high viability to generate a successful pregnancy remains an elusive goal in MAR cycles. Here, in this paragraph, we will investigate whether the use of TLM and morphokinetic embryo assessment might help in achieving this goal. Time-lapse observations have been used to define new or poorly described concepts of human embryology such as the fertilization process (Coticchio et al., Reference Coticchio, Mignini Renzini, Novara, Lain, De Ponti, Turchi, Fadini and Dal Canto2018), the duration of the first three cell cycles (Wong et al., Reference Wong, Loewke, Bossert, Behr, De Jonge, Baer and Reijo Pera2010, Meseguer et al., Reference Meseguer, Herrero, Tejera, Hilligsoe, Ramsing and Remohi2011), the early compaction stage (Iwata et al., Reference Iwata, Yumoto, Sugishima, Mizoguchi, Kai, Iba and Mio2014) and finally blastocyst formation (Marcos et al., Reference Marcos, Perez-Albala, Mifsud, Molla, Landeras and Meseguer2015, Sciorio et al., Reference Sciorio, Herrer Saura, Thong, Esbert Algam, Pickering and Meseguer2020a, Reference Sciorio, Thong and Pickering2020b).

Recently, Coticchio and collaborators thoroughly investigated the fertilization event and illustrated several unknown characteristics including the cytoplasmic halo (appearance and disappearance), pronuclei (PN) fading, time from PN fading (tPNf) and the first cleavage, and those novel features were used to predict embryo quality on day-3 (Coticchio et al., Reference Coticchio, Mignini Renzini, Novara, Lain, De Ponti, Turchi, Fadini and Dal Canto2018). Another prospective study analyzed the correlation between tPNf and live birth in 159 embryos. The pronuclei morphology of 46 embryos that resulted in live birth was compared with that of 113 embryos that resulted in no live birth. The results reported that, in comparison with embryos resulting in no live birth, tPNf occurred significantly later in embryos resulting in live birth and never earlier than 20 h 45 min (Azzarello et al., Reference Azzarello, Hoest and Mikkelsen2012). A further study noted that erratic PN movement within the cytoplasm and delayed pronuclei fading were indicative of compromised embryo development (Athayde Wirka et al., Reference Athayde Wirka, Chen, Conaghan, Ivani, Gvakharia, Behr, Suraj, Tan and Shen2014). The authors in this retrospective multicentre trial identified four atypical phenotypes, including abnormal syngamy, abnormal first cytokinesis, abnormal cleavage and chaotic cleavage, and correlated their relationship with embryo viability and implantation potential. The conclusion was that embryos exhibiting atypical phenotypes showed a significantly lower developmental potential compared with the control group (Athayde Wirka et al., Reference Athayde Wirka, Chen, Conaghan, Ivani, Gvakharia, Behr, Suraj, Tan and Shen2014). Wong and co-workers (Reference Wong, Loewke, Bossert, Behr, De Jonge, Baer and Reijo Pera2010) suggested that blastocyst stage could be predicted with high sensitivity by the timing of the early developmental stage: including the first cytokinesis (0–33 min), the time interval between the end of the first mitosis and the initiation of the second, which is the duration of 2-cell stage (7.8–14.3 h), and the time interval between the second and third mitoses (0–5.8 h duration of the 3-cell stage).

Lemmen and colleagues established that embryos resulting in successful pregnancies displayed not only a significantly higher cleavage synchrony but also a higher synchrony in nuclear appearance at the 2-cell stage compared with non-implanting embryos (Lemmen et al., Reference Lemmen, Agerholm and Ziebe2008). Using morphokinetics assessment it has been possible to demonstrate associations between various cleavage stage events and the embryo’s ability to reach the blastocyst stage (Wong et al., Reference Wong, Loewke, Bossert, Behr, De Jonge, Baer and Reijo Pera2010; Cruz et al., Reference Cruz, Garrido, Herrero, Perez-Cano, Munoz and Meseguer2012).

Meseguer and collaborators analyzed large data sets on transferred embryos generated by ICSI, and the results displayed that the timing of the cleavage to five cells was the most predictive parameter for embryo viability and implantation (Meseguer et al., Reference Meseguer, Herrero, Tejera, Hilligsoe, Ramsing and Remohi2011). Subsequently, the same group in a retrospective multicentre study performed in 10 IVF clinics compared pregnancy outcomes between time-lapse culture (n = 1390 cycles) and a standard incubator (n = 5915) and they reported a relative 20% improvement in pregnancy rates in the TLM (Meseguer et al., Reference Meseguer, Rubio, Cruz, Basile, Marcos and Requena2012). The authors associated the elevated clinical pregnancy obtained in the TLM group to a combination of both stable culture conditions and the use of morphokinetic parameters for embryo selection.

Similar results were confirmed by the same group in a prospective randomized controlled trial 2 years later (Rubio et al., Reference Rubio, Galan, Larreategui, Ayerdi, Bellver, Herrero and Meseguer2014). The introduction of more physiological culture conditions for in vitro human embryos, has resulted in the routine culture and transfer of embryos at the blastocyst stage (Gardner and Schoolcraft, Reference Gardner and Schoolcraft1999; De Vos et al., Reference De Vos, Van Landuyt, Santos-Ribeiro, Camus, Van de Velde, Tournaye and Verheyen2016). Within countries undertaking only single embryo transfer, an active reduction in the numbers of embryos being transferred has been witnessed.

Additionally, the transfer of a single blastocyst prevents many adverse medical conditions associated with multiple pregnancies (Sullivan et al, Reference Sullivan, Wang, Hayward, Chambers, Illingworth, McBain and Norman2012; De Vos et al., Reference De Vos, Van Landuyt, Santos-Ribeiro, Camus, Van de Velde, Tournaye and Verheyen2016). Furthermore, blastocyst transfer provides increased implantation rates compared with transfers at the cleavage stage, but this outcome needs to be correlated with the possible detrimental epigenetic effects associated with extended in vitro culture (Kirkegaard et al., Reference Kirkegaard, Agerholm and Ingerslev2012). In this context, TLM has been applied to predict blastocyst formation and implantation potential based on novel morphokinetic parameters noted at cleavage stage (Dal Canto et al., Reference Dal Canto, Coticchio, Mignini Renzini, De Ponti, Novara, Brambillasca, Comi and Fadini2012).

Kirkegaard and collaborators reported that cleavage from the 2- to 8-cell stages appears progressively earlier in embryos that will generate a blastocyst and will implant. The authors suggested that top quality blastocysts could be predicted within the first 2 days of in vitro culture by the short duration of the first cleavage, and duration of the 3-cell stage (Kirkegaard et al., Reference Kirkegaard, Kesmodel, Hindkjaer and Ingerslev2013). In the same way, Hashimoto and collaborators (Reference Hashimoto, Kato, Saeki and Morimoto2012) showed that better quality blastocysts presented significantly shorter times for synchrony between the 3- and 4-cell stages. Recently, Motato and colleagues (Reference Motato, de los Santos, Escriba, Ruiz, Remohi and Meseguer2016) analyzed the morphokinetic parameters in 7483 embryos and identified two features linked to blastocyst formation: the time of morula formation (81.28–96.0 h after ICSI), and the time of transition from 5- to 8-cell embryos (≤ 8.78 h). Finally, spontaneous blastocyst collapse during in vitro embryo development has been suggested as a novel marker of embryo viability and implantation potential. Retrospective studies have reported that blastocysts exhibiting collapse during development are less likely to implant and generate a pregnancy compared with embryos that do not (Marcos et al., Reference Marcos, Perez-Albala, Mifsud, Molla, Landeras and Meseguer2015, Sciorio et al., Reference Sciorio, Herrer Saura, Thong, Esbert Algam, Pickering and Meseguer2020a, Reference Sciorio, Thong and Pickering2020b). It has been reported that annotation of collapse(s) events may improve the embryo assessment at blastocyst stage. Summaries of the main atypical features identified with the TLM are given in Tables 1 and Table 2 including some recent papers published on the time-lapse technology.

Table 1. Atypical phenotypes observed with time-lapse monitoring

Table 2. Some studies published from 2010 that have used time-lapse technology

Correlation between TLM and embryos aneuploidy

Aneuploidy is the occurrence of the wrong number of chromosomes in a cell, for example 45 or 47 chromosomes instead of the normal 46. Aneuploidy is an important concern in in vitro human embryos obtained from MAR treatments: the transfer of aneuploid embryos may result in implantation failure, miscarriage or birth of an offspring with a range of potential abnormalities (Sciorio and Dattilo, Reference Sciorio and Dattilo2020).

The conventional procedure to investigate aneuploidy in human embryos is termed preimplantation genetic testing for aneuploidy (PGT-A), previously called preimplantation genetic screening (PGS), which consists in an IVF cycle in which embryos are biopsied and screened for chromosomal abnormalities prior to replacement into the uterus. The procedure was first introduced by Handyside (Handyside et al., Reference Handyside, Kontogianni, Hardy and Winston1990). However, PGT-A is an expensive technology and is not allowed in some countries, and there remains some debate regarding its cost–effectiveness, the invasiveness of the procedure and the clinical efficiency (Sermon et al., Reference Sermon, Capalbo, Cohen, Coonen, De Rycke, De Vos, Delhanty, Fiorentino, Gleicher and Griesinger2016; Sciorio and Dattilo, Reference Sciorio and Dattilo2020).

It has been hypothesized that TLM could be used to identify embryo aneuploidy, therefore providing a cheaper, faster and less invasive evaluation approach. Several studies have correlated morphokinetic parameters using TLM with the probability of selecting chromosomally normal embryos. It was supposed that cell division length has to be within an optimum range to overcome all the cellular processes preceding cytokinesis (Davies et al., Reference Davies, Christopikou, Tsorva, Karagianni, Handyside and Mastrominas2012; Campbell et al., Reference Campbell, Fishel, Bowman, Duffy, Sedler and Hickman2013a, Reference Campbell, Fishel, Bowman, Duffy, Sedler and Thornton2013b; Montag, Reference Montag2013; Swain, Reference Swain2013; Chawla et al., Reference Chawla, Fakih, Shunnar, Bayram, Hellani and Perumal2015).

Davies and co-workers (Reference Davies, Christopikou, Tsorva, Karagianni, Handyside and Mastrominas2012) found that aneuploidy embryos showed delays on the first two cleavages as well as prolonged transitions between 2- and the 4-cell stages. The author also noted that irregular divisions and an asynchronous PN disappearance were higher in abnormal embryos compared with the normal group. Chavez and colleagues (Reference Chavez, Loewke, Han, Moussavi, Colls, Munne, Behr and Reijo Pera2012) investigated the relationship between genetic status and morphokinetic parameters, and demonstrated that euploid embryos have definite timing at the first cell divisions up to the 4-cell stage.

Chawla and associates (Reference Chawla, Fakih, Shunnar, Bayram, Hellani and Perumal2015) assessed several morphokinetic features including timings of the extrusion of second polar body, pronuclei appearance and fading, time of first division, second and third cleavages duration in 460 embryos to discriminate abnormal embryos. The results showed that morphokinetic parameters differed significantly for euploid and aneuploidy embryos (Chawla et al., Reference Chawla, Fakih, Shunnar, Bayram, Hellani and Perumal2015). Campbell and collaborators using the TLM tried to develop a model to identify embryos aneuploidies. They found the time of early blastulation and the timing of full blastocyst formation were relevant features for embryo euploidy (Campbell et al., Reference Campbell, Fishel, Bowman, Duffy, Sedler and Hickman2013a, Reference Campbell, Fishel, Bowman, Duffy, Sedler and Thornton2013b).

Basile and colleagues investigated the differences in the time of cleavage between chromosomally normal and abnormal embryos to identify chromosomally normal embryos. The authors showed that normal and abnormal embryos had different kinetic behaviours and, based on that, they proposed an algorithm as a non-invasive tool to increase the likelihood of selecting genetically normal embryos (Basile et al., Reference Basile, Nogales, Bronet, Florensa, Riqueiros, Rodrigo, Garcia-Velasco and Meseguer2014). A comprehensive review on the value of TLM as a tool to identify and select euploid embryos has been published recently by Reignier and co-workers. They concluded that, despite several studies showing significant differences in morphokinetic parameters between euploid and aneuploid embryos, none of them found adequate evidence to recommend the clinical use of TLM in identifying embryo aneuploidies. Consequently, selection of embryos using time-lapse technology should not be considered as a replacement for PGT-A (Reignier et al., Reference Reignier, Lammers, Barrière and Fréour2018).

Different TLM systems

At present, there are several commercially available time-lapse systems. During the selection process of a TLM model, the clinic should take in consideration some practical aspects, including size and space of each system, the cost and the laboratory workload. Generally, all systems necessitate the use of a digital inverted microscope with a camera to collect embryo images at specific times. Some models contain an incubator equipped with a built-in camera, while other systems comprise a camera that is placed in a traditional large-box incubator (Kirkegaard et al., Reference Kirkegaard, Agerholm and Ingerslev2012; Chen et al., Reference Chen, Tan, Suraj, Pera and Shen2013; Campbell and Fishel, Reference Campbell and Fishel2015).

Although all TLM systems available at this time use an oil overlay on culture microdrops, there are differences in the way the embryos are cultured, and all systems need a specific culture dish, supplied by the manufacturer. Some models provide an individual culture set-up, in which the dish has a designed number of wells, each holding one embryo (Chen et al., Reference Chen, Tan, Suraj, Pera and Shen2013; Campbell and Fishel, Reference Campbell and Fishel2015; Racowsky et al., Reference Racowsky, Kovacs and Martins2015). However, some culture dishes permit sharing of culture medium between compartments and are designed for group culture, allowing exchange of soluble components. This may represent an important concern when deciding which specific model to purchase. However, each system uses a different light source and differs in the way the embryos are brought into the field of view, some without movement of the embryos versus others in which there is a constant movement of the culture dish. Few systems use bright field technology allowing the assessment of both kinetic parameters and embryo morphology.

Other models apply dark field technology supporting the determination of kinetic parameters, but give limited information on the morphological features. However, other aspects influencing a decision might include the nature of the computer software used for visualization and analysis, and the options for annotation, which may be manual or automated. Technical characteristics of some TLM systems available at this time are reported in Table 3.

Table 3. Comparisons of technical characteristics of some TLM systems at this time available

Time-lapse monitoring and its effect on embryo culture

Human embryo culture is associated with numerous physical and chemical stressors (Wale and Gardner, Reference Wale and Gardner2016), which might create a hostile environment for the preimplantation developing embryo. Embryo culture using TLM allows culture in a stable environment, avoiding the exposure of the embryo to the non-physiological conditions such as pH and temperature changes, or altered gases concentrations (CO₂ and O₂). The culture medium used represents an important factor for embryo culture.

Over the past few decades, we have noticed several improvements in the culture systems, mainly linked to medium composition. Two approaches have been suggested: the ‘sequential and the single-step’. The first is proposed to ensure that the physiological conditions of the human embryo are similar to that of the in vivo environment, when it would move from the oviduct to the uterus (Barnes et al., Reference Barnes, Crombie, Gardner, Kausche, Lacham-Kaplan, Suikkari, Tiglias, Wood and Trounson1995).

In contrast, the single-step medium is based on the concept that it is of benefit to supply all metabolic nutrients required, and that the embryo will use them according to its demand (Summers et al., Reference Summers, Bhatnagar, Lawitts and Biggers1995). Several studies have been performed in conventional incubators with the aim to establish which system is superior, but the results remain inconclusive (Sfontouris et al., Reference Sfontouris, Martins, Nastri, Viana, Navarro, Raine-Fenning, van der Poel, Rienzi and Racowsky2016; Werner et al., Reference Werner, Hong, Franasiak, Forman, Reda, Molinaro, Upham and Scott2016). The concern emerges of whether the use of TLM might find small parameters variations between sequential and single-step media.

One of the first studies to analyzing the dynamics of early development between embryos cultured in single and sequential medium using TLM was published by Ciray and associates. A randomized study was performed on 446 oocytes, which were divided between single and sequential medium produced by the same manufacturer and cultured in the same time-lapse incubator. The result found that in single-step medium, fading of PN and cleavage up to 5-cell stage took place significantly earlier compared with embryos cultured in sequential medium. In implanted embryos, t2 and t4 were significantly shorter with the single-step medium. However, the clinical outcome rates did not differ between the two groups (Ciray et al., Reference Ciray, Aksoy, Goktas, Ozturk and Bahceci2012). Similar findings were reported by Kazdar and collaborators (Reference Kazdar, Brugnon, Bouche, Jouve, Veau, Drapier, Rousseau, Pimentel, Viard and Belaud-Rotureau2017).

Conversely, other studies were not able to found any morphokinetic differences between the two approaches (Basile et al., Reference Basile, Morbeck, Garcia-Velasco, Bronet and Meseguer2013; Sfontouris et al., Reference Sfontouris, Kolibianakis, Lainas, Venetis, Petsas, Tarlatzis and Lainas2017). Therefore, current data have been unable to show a clear superiority of either single-step nor sequential medium in terms of clinical pregnancies adopting either with standard incubations or TLM. As mentioned previously, the use of TLM prevents embryo exposure to environmental conditions and thereby emulates in vivo conditions. It has been well reported that steady gas concentrations (for CO₂ and O₂) are extremely important for embryo development and viability (Sciorio and Smith, Reference Sciorio and Smith2019). It has been established that the oxygen concentration of the mammalian female reproductive tract is between 2 and 8% (Fischer and Bavister, Reference Fischer and Bavister1993).

Exposure of embryos to atmospheric oxygen concentrations is correlated with an increased production of reactive oxygen species, which might modify embryo metabolism and gene expression (Fischer and Bavister, Reference Fischer and Bavister1993; Rinaudo et al., Reference Rinaudo, Giritharan, Talbi, Dobson and Schultz2006; Wale and Gardner, Reference Wale and Gardner2012; Sciorio and Smith, Reference Sciorio and Smith2019). There is large amounts of evidence suggesting that embryo culture in 5%, rather than ambient, oxygen leads to improved pregnancy outcomes (Meintjes et al., Reference Meintjes, Chantilis, Douglas, Rodriguez, Guerami, Bookout, Barnett and Madden2009; Bontekoe et al., Reference Bontekoe, Mantikou, van Wely, Seshadri, Repping and Mastenbroek2012). A recent prospective randomized multicentre study performed on 1563 oocytes, confirmed that inclusion of antioxidants in the culture medium increased embryo viability, implantation and ongoing pregnancy rates significantly, possibly through the reduction of oxidative stress (Gardner et al., Reference Gardner, Kuramoto, Tanaka, Mitzumoto, Montag and Yoshida2020). The advantage of lower oxygen tension levels is included in the use of TLM.

Future research

Although TLM has been proposed since the 1929 (Lewis and Gregory, Reference Lewis and Gregory1929), the technology became commercially available only about a decade ago, therefore in comparison with other technical advancements made in cell biology, time-lapse might be considered in its childhood, and, as such, the technology applied could be further improved. Looking to the future, it will be expected that some developments correlated with image collection are likely to become available.

Development of fluorescence and confocal microscopy associated with the time-lapse and allowing the morphokinetic observation of organelles and chromosomes has already been proposed (Holubcova et al., Reference Holubcova, Blayney, Elder and Schuh2015; Patel et al., Reference Patel, Tan, Hartshorne and McAinsh2015), as well as fluorescence live-cell imaging of human embryos (Hashimoto et al., Reference Hashimoto, Nakano, Yamagata, Inoue, Morimoto and Nakaoka2016). Furthermore, one concern of TLM is the difficulty to assess and interpret the huge amount of data collected, which offers the opportunity for evolution of artificial intelligence (AI) and the use of higher-powered computers to analyze the considerable amount of images, and to identify a specific parameter that might correlate with embryo viability and pregnancy outcomes.

In that context, software programs are being used as automatic alternatives to standardize time-lapse annotations (Yeung et al., Reference Yeung, Downing, Fei-Fei and Milstein2018). Unlike other medical fields, ART has not yet explored the advantages of AI for automated embryo evaluation and selection. It has been hypothesized that an AI approach trained through exposure to thousands of embryo images and videos would later permit identification and prediction of embryo quality without mediation. This might be beneficial to reduce human error and standardize the annotation, and will allow embryologists concentrate on different tasks.

A study was performed by Khosravi and colleagues that used AI and TLM and, by analyzing clinical data for 2182 embryos and about 50,000 images, they developed a model that was able to predict blastocyst quality with an AUC of >0.98 (Khosravi et al., Reference Khosravi, Kazemi, Zhan, Malmsten, Toschi, Zisimopoulos, Sigaras, Lavery, Cooper and Hickman2019). In another recent retrospective trial, a deep learning approach has been used to automatically annotate 10,638 embryo videos from eight different IVF units across four countries. The results showed that deep learning model was able to predict fetal heartbeat pregnancy from time-lapse videos with an AUC of 0.93 (Tran et al., Reference Tran, Cooke, Illingworth and Gardner2019). These are retrospective studies, and further trials, including prospective randomized controlled trials, are required to evaluate the clinical significance of AI in IVF. However these trials demonstrated that the deep learning model and AI have an high level of predictability of embryo viability and implantation (Khosravi et al., Reference Khosravi, Kazemi, Zhan, Malmsten, Toschi, Zisimopoulos, Sigaras, Lavery, Cooper and Hickman2019; Tran et al., Reference Tran, Cooke, Illingworth and Gardner2019). Of course, before the AI approach would be clinically used, it will be required to pass vigorous clinical validation process.