Knowing which ewes and does are pregnant, and ideally how many young each carries, lets a keeper feed according to need, rebreed or sell barren animals and prepare for lambing or kidding. Before imaging was available the options were non-return to oestrus, abdominal palpation late in gestation and a series of laboratory and mechanical methods of limited accuracy or convenience (Ishwar, 1995). Real-time B-mode ultrasound replaced them. Early Scottish and Australian work showed that an experienced operator could diagnose pregnancy and count fetuses in large numbers of sheep quickly and with few errors (White et al., 1984; Fowler and Wilkins, 1984), and commercial scanning services grew up in countries with large flocks.

Those services depend on a skilled contractor, a handling crate and flocks mated over a known period. Smallholder and pastoral flocks have none of these. Males often run with females all year, mating dates are unknown, and animals are caught by hand. What has changed is the equipment: hand-held probes that display on a mobile device are now inexpensive enough for a district veterinary office. An instrument is not a method, however. This article proposes a written protocol for using such scanners to diagnose pregnancy in sheep and goats in the field. It is offered as a design for others to test, and no field results are reported.

The protocol draws on what is known about the two scanning approaches, the timing of ultrasonic signs of pregnancy, and the sources of error.

The protocol makes three design choices. It uses the transabdominal approach only, because it needs no rectal entry, suits both species and all body sizes, and can be taught quickly. It accepts a later earliest date of diagnosis in exchange. And it allows the operator to answer "uncertain", with a fixed rule for what follows. Table 1 sets out the steps.

The protocol assumes an operator who has been taught on animals of known status. A reasonable scheme is supervised scanning of pregnant and non-pregnant animals at several stages until the trainee's classifications match those of the instructor, followed by periodic checks. The teaching should cover the false images as carefully as the true ones: the bladder, loops of intestine containing fluid, the full rumen on the left side, and the fluid-filled uterus without a fetus. Agreement between two operators scanning the same animals independently can be expressed with the kappa statistic and interpreted on the conventional scale of Landis and Koch (1977). The final test of any operator is the record of births, and services using the protocol should collect lambing and kidding outcomes from a sample of flocks each season and compare them with the scan sheets.

Field use raises practical matters that clinic-based descriptions pass over. Screens are hard to read in sunlight, so scanning is done in shade or under a hood. Batteries, probe cables and phones fail in dust and heat, and a spare power source belongs in the kit. Probes are wiped and disinfected between flocks to avoid carrying infection on equipment. Keepers should be told beforehand what the three results mean, and in particular that an uncertain result is a request to return and not a failure of the test. Where a service is offered for a fee, the cost of the return visit should be included in the first.

The use made of the result should be agreed with the keeper before the visit. In a flock that lambs once a year the main gains are separating barren females for sale or remating and directing scarce supplementary feed to those carrying young, above all those carrying more than one. In a continuously mated village flock the gain is smaller for each animal but still real: a female sold for slaughter in the belief that she is barren may be pregnant, and a scan before sale prevents that loss. Handling should be calm and brief, with heavily pregnant animals never cast or lifted, and the keeper should restrain the animals personally wherever possible, since familiar handling reduces struggling and improves the image.

The protocol has been assembled from published work and has not been evaluated. Its accuracy under field conditions is unknown. A proper evaluation would scan a defined group of females once, according to the protocol, by operators unaware of mating records, and would take birth, abortion or slaughter findings as the reference standard. The design and reporting of such a study should follow the STARD statement for diagnostic accuracy research (Bossuyt et al., 2015), with sensitivity and specificity given separately for each stage of gestation and each operator, and with the uncertain results reported and not discarded.

Several limits are already clear. Pregnancies younger than about six weeks will be missed or deferred, which matters in systems where early rebreeding is the aim. Losses after a positive scan will appear as false positives against birth records. The gestational age cannot be estimated without fetal measurements and breed-specific growth curves, of the kind published for some dairy breeds (González de Bulnes et al., 1998), and these are not part of the protocol. Counting fetuses demands more skill than detecting pregnancy, and image quality from low-cost probes has not been compared formally with that of the cart-based machines used in the founding studies.

Portable scanners make pregnancy diagnosis possible in flocks that commercial scanning never reached. Their usefulness there depends on a method simple enough to teach and strict enough to keep errors visible. The protocol proposed in this article uses one transabdominal technique, fixed image criteria, an explicit uncertain category and a rescan rule, and it ties the operator's record to later births. A field accuracy study against birth outcomes is the necessary next step before it is recommended for routine use.

To be declared by the author(s) before publication.

To be declared by the author(s) before publication.