Role of magnetic resonance imaging in characterization of central nervous system lesions in pediatric patients with leukemia and post-treatment complications

Leukemia is one of the most common fatal diseases in pediatric oncology. Recently, advances in drug therapy have improved the prognosis of acute leukemia with event-free survival of up to 60%; however, complications and adverse effects of the disease and anti-leukemic treatment have also increased. The CNS complications of leukemia can be classified into those that developed directly or indirectly from the underlying leukemic process and those that can be related to antileukemic therapy. MRI had improved early detection of CNS complications and proper management. The study aims to characterize the MRI findings caused by the leukemic involvement of CNS structures and treatment-associated CNS complications and assess its value in early management and avoidance of long-term side effects. The patient’s age ranged from 2 to 18 years with different types of leukemia classified regarding the time of presentation as pretreatment, during treatment phases, and post-treatment. Different MRI abnormalities were recorded and clinically correlated. The neurological complications of leukemia have common presenting symptoms but varying imaging abnormalities. To reach the correct diagnosis, the presenting signs, symptoms, and laboratory data must be considered along with the radiologic findings. A diagnostic algorithm using conventional, post-contrast MRI, MR venography, along with diffusion-weighted MRI was of great value in early detection and differentiation of different CNS lesions detected in pediatric patients with leukemia and post-treatment CNS complications.


Background
Leukemia is a heterogeneous group of disorders, caused by a neoplastic transformation in blood cell precursors during their differentiation in bone marrow [1]. It is the most prevalent malignancy in children and represents nearly 30% of all malignancies, with 75% of cases corresponding to acute lymphoblastic leukemia (ALL). The prevalence of all forms of leukemia is more in children under 5 years of age and decreases significantly with advancing age. Diagnosis is made through a wide range of clinical and laboratory findings [2]. The central nervous system (CNS) complications of leukemia were rarely seen due to the rapid course of the disease. But now, therapeutic advances have improved the prognosis of acute leukemia with a survival rate up to 60% [3]. The prophylactic therapy of the CNS includes intrathecal methotrexate, high-dose chemotherapy, radiotherapy, or a combination of any of these. However, patients with ALL can present with neurological complications as a result of this therapy and the underlying systemic effects of leukemia [4]. The neurological complications of leukemia can be divided into (a) infiltration of the CNS by leukemic cells or (b) secondary complications due to the disease, and/or its therapy. The disease itself may involve the leptomeninges, brain parenchyma, or cerebral vasculature. CNS complications related to treatment include white matter lesions, small-vessel calcifications, cerebrovascular disorders, secondary tumors and infections, atrophic brain changes, endocrinopathies, and/or neurocognitive deficits caused by the "late effects" of their treatment [5].
Diagnosing CNS lesions in leukemia is often burdensome due to a variable set of causes; there are no pathognomonic imaging features for CNS lesions in leukemia, and the histological verification of a brain biopsy remains the gold standard for diagnosis. As many CNS lesions in leukemia are curable, early diagnosis is important for their proper management. MRI is known for its superior soft-tissue imaging; it was observed that MRI aided the characterization of CNS lesions caused by the leukemic involvement of CNS structures and treatment-associated CNS complications and thus will lead to early management and prevention or minimizing the side effects of the disease [6]. The purpose of this study is to analyze the MRI imaging features of these complications and interpreting these scans in view of clinical data to assess its value in early management.

Study patients
This prospective study was performed during the period from April 2018 to June 2020 which included 50 child patients with leukemia (30 males and 20 females) presented with neurological manifestations. The inclusion criteria were pediatric age group (up to 18 years) and confirmed cases of leukemia presented with neurological manifestations, while the exclusion criteria were patient over 18 years old and pediatric patient with leukemia but with no neurological manifestations. Patients included in the study were subjected to history taking, including age, sex, type of leukemia, clinical picture, and stage of treatment.

Image acquisition
All patients included in this study underwent: (1) Conventional MRI 1.5 T, superconducting MR imager (Magnotom ESPREE 1.5 T, Erlangen, Semines, Germany). All the cases were examined in the supine position with standard circularly polarized head coil using the following sequences: In 20 cases of the studied patients, we needed deep sedation or general anesthesia to minimize motion artifacts during MRI examinations.

Interpretation of MRI findings
The MRI findings in all cases were interpreted by two consultants of radiology in our department; each had more than 10 years of experience, conjoint reading.
The study was done after the approval of the Research and Ethical Committee. Informed consent from each child's guardian was obtained before performing the study.
Regarding the time of development of CNS manifestations, 76% of the studied patients had CNS manifestation during treatment phases, while 12% at the time of presentation before the start of treatment and 12% were relapsed disease post-end of treatment (Fig. 2). Regarding the phase of treatment at study time, 48.6% were at the induction phase, 38% at the maintenance phase, while 13.5% at the consolidation phase (Table 1).
The frequency of different MRI abnormalities regarding to treatment phase was depicted in Table 2.
PRES lesions were common at the induction phase (80%), while a small percentage at the consolidation and the maintenance phase in 10% per each. PRES lesions were identified in FLAIR images, and their distribution in anatomical regions was made (Fig. 4). The most commonly involved typical regions were the occipital lobes (90%), and this was followed by the parietal lobes (80%), frontal lobe (50%), and cerebellum (10%) (Fig. 5). The other atypical regions involved were the basal ganglia, thalamus, and brainstem in 10% per each (Fig. 6). All patients with cerebral lesions had bilateral distribution apart from one patient who had unilateral distribution affecting the left frontal lobe only. Follow-up study of that patient after a few days revealed new lesions appeared at both occipital lobes. All patients were normotensive apart from two patients who were hypotensive. In DWI, most patients showed facilitated diffusion (80%) and two patients (20%) had a restricted DWI pattern.
Hemorrhage was detected in two patients; one patient who was at the maintenance phase showed left frontal intra-axial hematoma, and the other patient who was at the time of presentation had multiple parenchymal hemorrhagic foci.
Cerebral venous sinus thrombosis was detected in 17 patients. It was seen mainly at the maintenance phase (53%) and at the induction phase (41%); only 5% was seen at the consolidation phase ( Table 3). The superior sagittal sinus was the commonest involved (76%), followed by the transverse sinuses which were involved in 41%. The left transverse sinus was more commonly involved in six cases, while the right transverse sinus was involved in one case (Fig. 7). Hemorrhage Mineralizing angiopathy --1 - Cerebral infection represented 4% of the studied cases. In the current study, two patients developed a cerebral infection with fungal aspergillosis seen at the induction phase. The first patient showed multiple cerebral masses, and the second patient had a focal lesion at the midbrain. In both cases, the lesions elicited an intermediate signal at T1WI with a marginal high signal. At T2WI, the core of the mass became low signal. At the DWI, the lesions showed marginal diffusion restriction, while the gradient images showed blooming of the center of the lesions. The first patient died within a few days, and the other improved with antifungal treatment. Both cases had concomitant chest fungal infection with a positive galactomannan test at the serum (Fig. 8).
Leukemic infiltration was seen at 11 cases, classified regarding the type of leukemia to seven cases with ALL (63%), two cases with Burkitt's leukemia (18%), one case with AML (9%), and one case with CML (9%) ( Table 4).
As regards to the site of infiltration, the MRI findings were classified as retinal, cerebral, meningeal, orbital, skull, and maxillofacial bone infiltration (Table 5).
Three cases (27%) had retinal infiltration presented with enhancing retinal nodules, elicited intermediate signal in T1WI, and slightly low signal in T2WI with postcontrast enhancement and diffusion restriction in DWI. Two cases (18%) had cerebral infiltration: the first case presented with multiple foci of abnormal parenchymal signal and enhancement seen at the thalami, brain stem, and cerebellum with no diffusion restriction, and the second case showed the right parieto-occipital and the right basal ganglia area of high FLAIR signal with postcontrast enhancement. Three cases (27%) had meningeal infiltration: one at presentation had dural thickening and enhancement seen at the frontal, temporal, and parasellar regions, the other presented with late CNS relapse had a subdural collection with cranial and spinal meningeal irregular thickening with enhancement, while the last one had intraventricular enhancing masses (Fig. 9). All cases had CSF analysis revealed blast cells in CSF. One case (9%) had an enhancing left orbital mass at presentation proved by biopsy as leukemic infiltration (Fig. 10). Two cases (18%) with skull bones and maxillofacial bone infiltration showed diffuse infiltrative marrow signal with periosteal reaction and periosteal soft tissue tumefaction (Fig. 11).
MTX toxicity was detected in five cases (10%); 60% at the consolidation phase, one 20% at the maintenance phase, and 20% at the re-induction phase post-relapse. Most cases showed a rather symmetrical periventricular, centrum semi-oval foci of restricted diffusion with no abnormality at T1-T2 or FLAIR (Fig. 12).
Leuko-encephalopathy was detected in 2% at the maintenance phase presented with a periventricular sheet of a high signal at FLAIR images, classified as grade I regarding the US National Cancer Institute grading of leukoencephalopathy. Isolated atrophic brain changes were detected in 2%, presented with late hematological relapse and had prominent cerebral ventricle, extra axial CSF spaces, and mild volume loss. Mineralizing angiopathy (2%) was seen at the maintenance phase. This patient was treated for ALL. CT showed bilateral subcortical calcific foci seen at the graywhite matter interface. In MRI, the calcifications were not clearly visible. In the follow-up studies, the patient developed cortical laminar necrosis appeared as a cortical bright signal at T1WI. Those patients were treated with intrathecal chemotherapy with no history of cranial radiation.

Discussion
Leukemia is considered one of the most common childhood cancers. Long ago, the central nervous system complications of leukemia were rarely seen because the disease was rapid and fatal. Latterly, advances in imaging techniques and treatment methods have prolonged survival, but unfortunately, the frequency of CNS complications has increased. We divide the CNS complications of leukemia into those that result directly or indirectly from the underlying leukemic process and those related to treatment side effects [5,7].
We prospectively evaluated the cranial MRI of 50 patients proved with leukemia ranging in age from 2 to 18 years and having CNS manifestations. Male represented 60% in our study, while females represented 40%.
ALL represented 86% of our cases, while AML represented 6%. Mixed phenotype leukemia and CML each represented 2%, while Burkitt's leukemia represented 4%. This coordinated with previous studies of Verma et al. [2] and Terwilliger et al. [7] which stated that ALL is the most common type of leukemia.
In the current study, vascular disorders were the most common complication (58%), of which (34%) had sinovenous thrombosis, 20% had PRES, and 4% had a parenchymal hemorrhage.
Cerebral venous sinus thrombosis was considered the most common complication seen commonly in the maintenance phase 53% followed by the induction phase 41% and a few percentages 5% in the consolidation phase. The study performed by Malhotra et al. [8] and Porto et al. [9] said that cerebro-venous thrombosis was  most common in the induction phase. Both studies have reported that venous sinuous thrombosis commonly happened post-exposure to L-asparaginase therapy in induction and maintenance phases. Treatment with Lasparaginase leads to the depletion of plasma proteins involved in both coagulation and fibrinolysis and has been linked to cerebrovascular complications including cortical infarcts, dural sinus thrombosis, intracerebral hemorrhage, and hemorrhagic infarcts. Cerebrovascular thrombosis or hemorrhage can also occur during antileukemic treatment as a result of leukocytosis, thrombocytopenia, sepsis, and coagulopathy [9]. This explains the sinovenous thrombosis in one of our patients in the consolidation phase apparently not related to asparaginase therapy.
For the diagnosis of venous thrombosis, the combination of conventional MR imaging with MRV could detect venous sinus thrombosis in all cases in our study. This agrees with multiple studies [10,11] which stated that the combination of MR and MRV is now the method of choice because of its capability to reveal the absence of flow in the cerebral veins even with the absence of typical findings of brain infarcts. MRV can demonstrate the thrombus in the acute phase by the absence of flow which can be mistaken in conventional images for flowing blood.
PRES is a clinico-neuro-radiologic disease with distinct MR imaging findings, accompanied by clinical manifestations ranging from headache, altered mental status, seizures, and vision loss, to the loss of consciousness. In the current study, most cases diagnosed with PRES (70%) were detected early in the induction phase, while 10% in the consolidation phase, 10% in the maintenance phase, and 10% in the re-induction phase post-relapse. This agreed with Bianca et al. [12] and Raman [13] which have reported that cases with PRES tend to occur early in the induction phase. In the present study, MRI examinations showed typical features of PRES with the most frequently affected areas of the brain being the posterior regions, the occipital and posterior parietal regions (80%), followed by the frontal regions (60%), and the least commonly affected regions such as the cerebellum, thalami, basal ganglia, and pons which represented 10% per each. This matched with the study performed by Appachu et al. [14] and in accordance with the findings of the previous studies by Raman [13], Kastrup et al. [15], and McKinney [16]. Although hypertension is a risk factor for PRES, the blood pressure can be normal, particularly in the settings of chemotherapy and immunosuppressive therapy [13]. This is strongly supported by the study results, as most of the studied cases were normotensive apart from only two cases which had low blood pressure.
In the present study, we had two cases with hemorrhage: one had a frontal hematoma, and the other had parenchymal microbleeds. The MRI accurately detected parenchymal hematoma and microbleeds in the two cases. T2* images show high sensitivity in the detection of parenchymal microbleeds. This agreed with the study performed by Charidimou [17] that said the introduction of blood-sensitive MRI sequences, including T2*-GRE and susceptibility-weighted imaging (SWI), has enabled the accurate detection of cerebral microbleeds (CMBsO) (defined radiologically as small, rounded, homogeneous, hypointense lesions not seen with conventional spin-echo sequences). Also, the study by Beavers et al. [18] stated that GRE images were the most sensitive for the detection of retinal hemorrhages.
Another complication was cerebral infection. It represented 4% and was detected in two cases of the present study. Both cases had cerebral aspergillosis: one presented with multiple intra-axial masses, and the other had a solitary lesion at the midbrain. Both cases had AML in the early induction phase. The diagnosis of aspergillosis was confirmed by biopsy and detecting galactomannan antigens both in CSF and serum. One of the two patients with multiple masses unfortunately died, and the other patient improved with antifungal Skull and maxillofacial -- Orbital mass ---1 - treatment. A study performed by Pandian et al. [19] stated that one of the major challenges in AML induction chemotherapy is the risk of death named induction mortality. One of the reasons which place AML patients at risk of acquiring various infections even before initiating induction chemotherapy could be the variable delay in seeking treatment. It is a well-known fact that the incidence and severity of infection depend on the duration of neutropenia. When the neutropenia is prolonged for more than 5 weeks, the chance of infection can be as high as 100%. The signal criteria show characteristic low signal foci with blooming artifacts in the wall and in the core of the lesions at gradient images corresponded to other prior studies by Marzolf et al. [20], Finelli et al. [21], and Luthra [22] which have assumed that this signal could be the result of either ferromagnetic fungal deposits (iron, manganese, magnesium, zinc) or the presence of methemoglobin (in the capsule wall and/or the macrophages) or the presence of free radicals produced by the macrophages. Previous clinical studies by Marzolf et al. [20] and Gartner et al. [23] had suggested that doctors should exclude the presence of coexisting pulmonary or sinus infections. In the current study, CT examinations of the chest and sinuses were performed in both cases to exclude pulmonary and sinus aspergillosis, and they were positive.
Among the studied patients, 11 cases (22%) had CNS infiltration with leukemia whether cerebral, retinal, leptomeningeal, or craniofacial bone leukemic infiltration. In the present study, CNS infiltration was common in cases with ALL (63%), followed by Burkitt's leukemia (18%), and then CML and AML (each represented 9%). This correlated with multiple studies by Charles et al.  [24] and Del Principe et al. [25] which stated that CNS involvement in patients with AML is considerably less common than CNS involvement by acute lymphoblastic leukemia (ALL) in both adults and children.
In the present study, meningeal infiltration was detected in three cases. The first case showed enhanced dural thickening and subdural collection which were reported as inflammatory versus infiltration. CSF cytology revealed blast cells in CSF. The second showed multiple meningeal-based enhanced small masses at the brain and spine which were reported as infiltration. The last one showed diffuse ependymal enhancement and intraventricular masses. These cases were proved by CSF analysis and showed regression in the follow-up studies post-treatment with chemotherapy.
Cerebral parenchymal involvement was detected in two cases and appeared as multiple enhancing foci and illdefined areas seen at the thalami, brain stem, and cerebellum which were reported as inflammatory versus infiltration. Further CSF analysis proved leukemic infiltration.
Among the cases of leukemic infiltration, there was one case with Burkitt's leukemia that had a left orbital mass, proved by biopsy as leukemic infiltration. The MRI findings showed imaging characteristics that were compatible with those described by other studies [26][27][28][29].
In the present study, MRI had high sensitivity in the detection of retinal involvement in all three cases presented by focal retinal enhancement and thickening with enhanced nodules and diffusion restriction in DWI. This was proved by ophthalmic examination which showed subretinal fluid collection suggestive of infiltration. The follow-up studies showed marked improvement postinduction of chemotherapy. These findings corresponded to a study performed by Green et al. [30] and Pflugrath [31].
In the current study, MRI had detected bone marrow infiltration of the skull and maxillofacial bones in two cases by alteration of the marrow signal, cortical expansion, erosions, and associated soft tissue masses that showed post-contrast enhancement and diffusion restriction in DWI. This agreed with a study performed by Eisa [32] and Cao et al. [33] which said that MRI with DWI is a useful and non-invasive tool for detecting skull bone infiltration in ALL children before treatment and also the normalization of marrow signal at complete remission after therapy. Toxicity related to high-dose methotrexate was detected in five patients presented by bilateral basal ganglia, periventricular regions, and centrum semi-oval areas with diffusion restriction and low ADC value. Subtle or no abnormal FLAIR signal was noted. These findings were similar to those described by Cruz-Carreras et al. [34] and Tando [35], as well as a study by Michael et al. [36] which concluded that DWI has the potential role for early detection of methotrexate white matter injury.
Periventricular leukoencephalopathy was detected in one case at the maintenance phase, presented by a periventricular deep white matter sheet of high FLAIR signal which was constant in follow-up studies and one case with atrophic brain changes post-end of therapy manifested by decreased cerebral volume and widened extraaxial CSF spaces. Both cases had no history of cranial irradiation. In a study performed by Porto et al. [9] said that atrophy of the brain was believed to be a late finding after irradiation related to a diffuse white matter injury, but other studies have suggested that the atrophic changes may be related to chemotherapy, with cranial irradiation playing a lesser role.
Mineralizing angiopathy was present in one case, which appeared as multiple calcific foci at the gray-white mater interface; this case had history of cranial radiation before the development of these calcifications. Calcifications were seen easily in non-contrast CT, while in MRI showed a faint bright signal at T1WI and a low signal at T2WI. This agreed with a study performed by Rossi [37] which stated that the best diagnostic radiologic tool for mineralizing microangiopathy is CT, which shows calcium deposition mainly localized in the basal ganglia and subcortical white matter.

Conclusion
The neurological complications of leukemia have common presenting symptoms but varying imaging abnormalities. To reach the correct diagnosis, the presenting signs, symptoms, and laboratory data must be considered along with the radiologic findings. A diagnostic algorithm using conventional, post-contrast MRI, MR venography, along with diffusion-weighted MRI was of great value in early detection and differentiation of different CNS lesions detected in pediatric patients with leukemia and post-treatment CNS complications.