Deletions and gains of chromosomal segments detectable cytogenetically have long been recognized as valuable tools for AML classification and prognostication. Segmental amplifications and deletions can be reliably detected at the whole genome level using array-based CGH. In this study, we utilized a novel dense tiling path array consisting of 386,165 unique isothermal (Tm=76°C) oligomers (average length 51 nt) spaced evenly at ~6Kb intervals across the genome. We analyzed 144 adult de novo AML samples: 64 had normal karyotypes, and 80 had 1 or 2 clonal, balanced or unbalanced abnormalities (samples with ≥3 clonal chromosome aberrations were excluded). Similar numbers of FAB M0/1, M2, M3, and M4 cases were included, and all samples had >30% blasts. Bone marrow-derived tumor DNA or control DNA derived from the blood of 23 normal individuals (matched for age and ethnicity, and with no history of cancer) was co-hybridized with a control pool of DNAs derived from the blood of 4 healthy young males. Of the 23 gains and losses detected cytogenetically in >20% of metaphases, 22 (96%) were also detected by CGH. Of the 20 copy number changes present in ≤20% of metaphases, CGH detected only 7 (35%). CGH identified X chromosome number correctly in all samples. Further, a number of previously described segmental copy number polymorphisms (CNPs; Sebat et al, Science2004;305:525) were identified in both the AML samples and the normal control population. Using very stringent criteria to define abnormal segments (≥8 consecutive oligomers with log2 values of at least +/− 0.5), we identified 47 independent loci in the AML samples that had abnormal segments that were not apparent cytogenetically (mostly due to small size). Thirteen of these were present in multiple AML samples (range 2–22), but were also present in at least one cancer-free control sample, or were previously identified in normal individuals. These clearcut CNPs tended to be small (<200 Kb). Thirty-four abnormal segments were found in only one AML sample. Of these, 19 were present in ≥1 of our cancer-free controls, and are thus newly defined CNPs. The remaining 15 loci were present in one AML sample each, but were not found in any cancer-free control samples; 6/15 were from samples with normal cytogenetics, and 9 were from samples with abnormal karyotypes. These 15 loci may represent rare CNPs, or novel microdeletions and amplifications that are relevant for AML pathogenesis. The sizes of the novel amplifications (N=5, sizes of 0.255, 0.572, 0.165, 0.077, and 0.062 Mb) and deletions (N= 10, sizes of 0.475, 0.124, 0.496, 7.56, 0.116, 0.997, 0.062, 0.160. 0.479, 0.356 Mb) tended to be larger than those of known CNPs. All 15 loci contained putative or annotated genes, several of which have been implicated in cancer pathogenesis. Further validation of these segments is ongoing, and additional studies with even higher resolution arrays (containing 1.5 million oligos) are in progress to assess the importance of these novel regions for AML pathogenesis.