This research was intended to be an exploratory research of the applicability of WB-DWI-MRI. It was a prospective clinical study, in which the patients were followed up till a clarification of the nature of the lesion had been reached. So, the researchers are naturally blind to the actual nature of the lesion. In most similar studies, comparisons are made with other concurrently done scans that have their own inherent accuracies. Additionally, for this study to be as beneficial and reproducible as possible, we tried to simulate the actual clinical scenario that the radiographer is exposed to in his/her daily activities. So, we combined the data from the conventional MRI to that of DWI-MRI and did not try to compare them to each other, which does not happen in the clinical scenarios.
We did not restrict our inclusion to a specific disease or organ, either primary or secondary. Also, to get the highest yield from our limited research, we deliberately did not investigate the role of WB-MRI in assessment of tumor response.
In this study, to get the most accurate calculations, we divided the LN and skeletal groups to the smallest possible segments or regions. For example, ribs, we divided in 24 ribs; so every single rib is an independent region. Meanwhile, we did not divide organs into further smaller segments.
LN lesions
Regarding evaluating malignant metastatic or lymphomatous LNs, the relatively low sensitivity of WB-DWI-MRI in detection of regional nodal metastases in this study can be explained by the fact that metastases can be present in non-enlarged LN and not all enlarged nodes are malignant. Changing the size criteria towards larger or smaller cut-offs will influence sensitivity and specificity as a low threshold for cut-off values would increase the sensitivity, but specificity would decrease. This can explain the variation in the sensitivity of WB-DWI-MRI in other studies which ranges from 60 to 90% [4, 5]. The most accepted criterion for malignant LN involvement in conventional MRI sequences is that a short diameter greater than 8–10 mm [6]. A normal LN already has a relatively long T2 relaxation time and a restricted diffusion, due to their high cellularity, thus adding ADC values enhanced the discrimination between malignant and benign LNs. In this study, the combined usage of WB-DWI-MRI with the conventional sequences improved the detectability of LN metastasis. Seber et al. [7] in their study on differentiating benign from malignant LN infiltration from wide range of tumor types involving different body locations, found that adding DWI (utilizing an ADC cut-off value of 0.8 × 10−3 mm2/s) to conventional MRI findings increased the accuracy from 80.6 to 91.6%.
There is a wide variability of accuracies within different sites in the study. The highest accuracy was found in porta-hepatis LNs. Whereas the lowest accuracy was found in mediastinal LNs where image quality may be affected by pulsation artifacts. The later finding was contrary to Sigovan et al.’s findings [8] where they showed high sensitivity, specificity, and accuracy (90.9%, 83%, and 85%, respectively) for DWI in differentiating benign from malignant enlarged mediastinal LNs.
One of the confusing pitfalls is splenic hilar lymphadenopathy with an accessory spleen, as both may be hyper-intense on STIR and DWI sequences. The same issue was reported by Massani et al. [9].
In our study, there were no differences in DWI values between Hodgkin disease and non-Hodgkin’s lymphoma. Interestingly, Sabri et al. [10], in their study on malignant mediastinal LN involvement, detected a significant difference between the ADC values of Hodgkin disease and non-Hodgkin lymphoma.
Skeletal system lesions
In this study, WB-DWI-MRI identified 1137 bony lesions in 984 regions. The sensitivity, specificity, and accuracy were 88%, 94%, and 92%, respectively. Similar to other studies, we noted that the sole evaluation of bone lesions by DWI, STIR, or T2-WI makes high false-positive results of non-tumor lesions as in anemia, smoking, spondylodiscitis, osteomyelitis, bone infarctions, hemangiomas, and fractures which cause increased signal due to bone marrow edema [11,12,13]. The use of ADC maps and T1-WI in these occasions could decrease the false-positive results. Sagittal STIR of whole spine was included for evaluation of all the vertebrae to improve the accuracy of lesion localizations due to a higher spatial resolution [14]. False-negative results were found in the areas prone to motion artifacts as the sternum and clavicles. They were more obvious in conventional sequences than in WB-DWI [15, 16].
Brain lesions
WB-DWI-MRI only provide information about tumor cellularity, and this may explain its low sensitivity of 78% and high specificity 95% in this study as it could correctly identify 13 lesions of the brain in 9 patients. This was evident in other studies as well and is believed to be due to inherent high signal of the brain [17]. In suspected brain lesions, FLAIR sequence was exceptionally added to confirm the diagnosis, which is capable to elucidate areas of vasogenic edema around the tumor [18]. Ahmed and Mokhtar [19] reported similar finding to ours in their study on cerebral lesions. They found that there are mixed diffusion changes in most of the lesions in their study, and there was overlap in the measured ADC values between neoplastic and non-neoplastic lesions. On the other hand, Berghoff et al. [20] reported that ADC values correlated with survival and recurrence after surgical resection. Additionally, Lee et al. [21] reported that it correlated with survival after radiosurgery. Another interesting finding was reported by Zakaria et al. [22], that ADC changes at the tumor edge may indicate a more aggressive phenotype.
Pulmonary lesions
WB-MRI identified 54 lesions in 17 patients. Sensitivity, specificity, and accuracy were 64%, 88%, and 76%, respectively. This was lower than that of other conventional imaging as CT in many studies [23]. This can be attributed to difficulties in detection of small metastatic lung nodules (less than 8 mm). Free-breathing DWI acquisition in this study led to reduced accuracy of lesion detection (especially lung bases) due to respiratory movements and cardiac motion [23]. Also, lesions located at air-tissue interfaces were more difficult to be identified. Similarly, Regier et al. [24] reported sensitivity of 97% for nodules larger than 10 mm, but for 6–9 mm nodules, the sensitivity decreased to 86%, and for nodules 5 mm or smaller, it decreased to 43.8%. Similar finding was reported by Liu et al. [25] in assessing pulmonary metastases in patients with renal cell carcinoma. The use of ADC mapping added a value of discrimination between benign and malignant nodules [26]. A recent meta-analysis reported sensitivity and specificity of DWI to be 82.8% and 80.1%, respectively [27].
Liver lesions
Regarding hepatic metastases, WB-MRI had an accuracy of 100%. Thanks to the use of combined DWI with ADC together with anatomical localization by conventional MRI, the detection rate of malignant focal hepatic lesion was high. These results were also reported in another study [28]. Another explanation to these high results in this study as whole organ is considered a single region. As there was usually more than one lesion in the liver, if only one of them was correctly detected, the researchers would consider this scan a true-positive; the others will not be reported. WB-MRI is comparable with contrast-enhanced CT or contrast-enhanced MRI but with advantage that it does not require contrast medium [29]. Of particular interest, Abou-Khadrah and Bedeer [30] in a prospective study showed that DWI is beneficial in small hepatocellular carcinomas smaller than 2 cm where they reported high sensitivity, specificity, and accuracy of combined conventional MRI and diffusion (92.8%, 71.4%, and 93.3%, respectively).
Study limitations
There are some limitations to the study. It included a heterogenous group of patients with different pathologies, so, each pathology was represented by a small number of patients. Our aim was to do an exploratory study for the value of WB-DWI-MRI in detecting metastases, by including different types of malignancies and lymphomas in a single study. To minimize the effect of small study size, we deliberately did not investigate the role of WB-MRI in assessment of tumor response. Another limitation to the study is that histological examination to confirm the nature of the lesion was possible in only a small number of lesions. This study was conducted in clinical setting, and biopsies were requested by the treating physicians for the clinical requirements not for the sake of the study. This study was prospectively conducted and included only those patients who could be followed up by different radiological examinations for periods that were deemed sufficient to prove that a specific radiological finding is metastatic or not base upon its biological behavior.
Conclusion
The diagnostic performance of WB-DWI-MRI combined with conventional MRI and ADC mapping is variable among different anatomical sites. It has good performance in diagnosis of some organs as liver, bone marrow, and some LNs as porta-hepatis. It has a less diagnostic performance in the lung, and LNs located in cervical, mediastinum, supraclavicular, and mesenteric regions.